├── .dockerignore ├── .env.example ├── .gitignore ├── CONTRIBUTING.md ├── Dockerfile.backend ├── LICENSE ├── README.md ├── compose.yaml ├── data ├── pitch_deck.md ├── playbook.md ├── the_pmarca_blog_archive.md └── vc_funding.md ├── docker ├── agent.py ├── main.py └── tools.py ├── environment.yml ├── frontend ├── Dockerfile ├── frontend.py └── requirements.txt ├── logo.png ├── scripts ├── agent.py ├── main.py └── tools.py ├── setup.ps1 ├── setup.sh ├── shell ├── conda_env.sh └── run.sh ├── start_services.ps1 ├── start_services.sh └── workflow.png /.dockerignore: -------------------------------------------------------------------------------- 1 | qdrant/ 2 | script/ 3 | .mypy_cache/ 4 | -------------------------------------------------------------------------------- /.env.example: -------------------------------------------------------------------------------- 1 | GROQ_API_KEY="gsk_***" -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | qdrant/ 2 | */__pycache__/ 3 | .mypy_cache/ 4 | .env 5 | scripts/.env 6 | ragcoon.code-workspace 7 | -------------------------------------------------------------------------------- /CONTRIBUTING.md: -------------------------------------------------------------------------------- 1 | # Contributing to `RAGcoon` 2 | 3 | Do you want to contribute to this project? Make sure to read this guidelines first :) 4 | 5 | ## Issue 6 | 7 | **When to do it**: 8 | 9 | - You found bugs but you don't know how to solve them or don't have time/will to do the solve 10 | - You want new features but you don't know how to implement them or don't have time/will to do the implementation 11 | 12 | > ⚠️ _Always check open and closed issues before you submit yours to avoid duplicates_ 13 | 14 | **How to do it**: 15 | 16 | - Open an issue 17 | - Give the issue a meaningful title (short but effective problem description) 18 | - Describe the problem following the issue template 19 | 20 | ## Traditional contribution 21 | 22 | **When to do it**: 23 | 24 | - You found bugs and corrected them 25 | - You optimized/improved the code 26 | - You added new features that you think could be useful to others 27 | 28 | **How to do it**: 29 | 30 | 1. Fork this repository 31 | 2. Commit your changes 32 | 3. Submit pull request (make sure to provide a thorough description of the changes) 33 | 34 | ### Thanks for contributing! -------------------------------------------------------------------------------- /Dockerfile.backend: -------------------------------------------------------------------------------- 1 | FROM condaforge/miniforge3 2 | 3 | WORKDIR /app/ 4 | COPY ./docker/*.py /app/ 5 | COPY ./data /app/data/ 6 | COPY ./shell/ /app/shell/ 7 | COPY ./environment.yml /app/ 8 | 9 | RUN bash /app/shell/conda_env.sh 10 | 11 | CMD ["bash", "/app/shell/run.sh"] -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | MIT License 2 | 3 | Copyright (c) 2025 Clelia Astra Bertelli 4 | 5 | Permission is hereby granted, free of charge, to any person obtaining a copy 6 | of this software and associated documentation files (the "Software"), to deal 7 | in the Software without restriction, including without limitation the rights 8 | to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 9 | copies of the Software, and to permit persons to whom the Software is 10 | furnished to do so, subject to the following conditions: 11 | 12 | The above copyright notice and this permission notice shall be included in all 13 | copies or substantial portions of the Software. 14 | 15 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 | IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 | FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE 18 | AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 | LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 20 | OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 21 | SOFTWARE. -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | 2 |

RAGcoon🦝

3 | 4 |

Agentic RAG to help you build your startup

5 | 6 |
7 |

If you find RAGcoon userful, please consider to donate and support the project:

8 | GitHub Sponsors Badge 9 |
10 |
11 |
12 | ragcoon Logo 13 |
14 | 15 | ## Install and launch🚀 16 | 17 | The first step, common to both the Docker and the source code setup approaches, is to clone the repository and access it: 18 | 19 | ```bash 20 | git clone https://github.com/AstraBert/ragcoon.git 21 | ``` 22 | Once there, you can choose one of the two following approaches: 23 | 24 | ### Docker (recommended)🐋 25 | 26 | > _Required: [Docker](https://docs.docker.com/desktop/) and [docker compose](https://docs.docker.com/compose/)_ 27 | 28 | - Add the `groq_api_key` in the [`.env.example`](.env.example)file and modify the name of the file to `.env`. Get this key: 29 | 30 | + [On Groq Console](https://console.groq.com/keys) 31 | 32 | ```bash 33 | mv .env.example .env 34 | ``` 35 | 36 | - Launch the Docker application: 37 | 38 | ```bash 39 | # If you are on Linux/macOS 40 | bash start_services.sh 41 | # If you are on Windows 42 | .\start_services.ps1 43 | ``` 44 | 45 | Or, if you prefer: 46 | 47 | ```bash 48 | docker compose up qdrant -d 49 | docker compose up frontend -d 50 | docker compose up backend -d 51 | ``` 52 | 53 | You will see the frontend application running on `http://localhost:8001`/ and you will be able to use it. Depending on your connection and on your hardware, the set up might take some time (up to 30 mins to set up) - but this is only for the first time your run it! 54 | 55 | ### Source code🖥️ 56 | 57 | > _Required: [Docker](https://docs.docker.com/desktop/), [docker compose](https://docs.docker.com/compose/) and [conda](https://anaconda.org/anaconda/conda)_ 58 | 59 | - Add the `groq_api_key` in the [`.env.example`](.env.example)file and modify the name of the file to `.env` in the `scripts` folder. Get this key: 60 | + [On Groq Console](https://console.groq.com/keys) 61 | 62 | ```bash 63 | mv .env.example scripts/.env 64 | ``` 65 | 66 | - Set up RAGcoon using the dedicated script: 67 | 68 | ```bash 69 | # For MacOs/Linux users 70 | bash setup.sh 71 | # For Windows users 72 | .\setup.ps1 73 | ``` 74 | 75 | - Or you can do it manually, if you prefer: 76 | 77 | ```bash 78 | docker compose up qdrant -d 79 | conda env create -f environment.yml 80 | conda activate ragcoon 81 | ``` 82 | 83 | - Now launch the frontend application 84 | 85 | ```bash 86 | gunicorn frontend:me --bind 0.0.0.0:8001 87 | ``` 88 | 89 | - And then, from another terminal window, go into `scripts` and launch the backend: 90 | 91 | ```bash 92 | uvicorn main:app --host 0.0.0.0 --port 8000 93 | ``` 94 | 95 | You will see the application running on `http://localhost:8001` and you will be able to use it. 96 | 97 | ## How it works 98 | 99 | ![Workflow](workflow.png) 100 | 101 | The main workflow is handled by a Query Agent, built upon the ReAct architecture. The agent exploits, as a base LLM, the latest reasoning model by Qwen, i.e. [`QwQ-32B`](https://console.groq.com/docs/model/qwen-qwq-32b), provisioned by [Groq](https://groq.com/). 102 | 103 | 1. The question coming from the frontend (developed with [Mesop](https://google.github.io/mesop/) - running on `http://localhost:8001`) is sent into a POST request to the [FastAPI](https://fastapi.tiangolo.com/)-managed API endpoint on `http://localhost:8000/chat`. 104 | 2. When the Agent is prompted with the user's question, it tries to retrieved relevant context routing the query to one of three query engines: 105 | - If the query is simple and specific, it goes for a direct hybrid retrieval, exploiting both a dense ([`Alibaba-NLP/gte-modernbert-base`](https://huggingface.co/Alibaba-NLP/gte-modernbert-base)) and a sparse ([`Qdrant/bm25`](https://huggingface.co/Qdrant/bm25)) retriever 106 | - If the query is general and vague, it first creates an hypothetical document, which is embedded and used for retrieval 107 | - If the query is complex and involves searching for nested information, the query is decomposed into several sub-queries, and the retrieval is performed for all of them, with a summarization step in the end 108 | 3. The agent evaluates the context using [`llama-3.3-70B-versatile`](https://github.com/meta-llama/llama-models/blob/main/models/llama3_3/MODEL_CARD.md), provisioned through Groq. If the context is deemed relevant, the Agent proceeds, otherwise it goes back to retrieval, trying a different method. 109 | 4. The agent produces a candidate answer 110 | 5. The agent evaluates the faithfulness and relevancy of the candidate response, in light of the retrieved context, using [LlamaIndex evaluation methods](https://docs.llamaindex.ai/en/stable/module_guides/evaluating/) 111 | 6. If the response is faithful and relevant, the agent returns the response, otherwise it gets back at generating a new one. 112 | 113 | ## Contributing 114 | 115 | Contributions are always welcome! Follow the contributions guidelines reported [here](CONTRIBUTING.md). 116 | 117 | ## License and rights of usage 118 | 119 | The software is provided under MIT [license](./LICENSE). 120 | -------------------------------------------------------------------------------- /compose.yaml: -------------------------------------------------------------------------------- 1 | services: 2 | qdrant: 3 | image: qdrant/qdrant 4 | ports: 5 | - 6333:6333 6 | - 6334:6334 7 | volumes: 8 | - ./qdrant/storage:/qdrant/storage 9 | frontend: 10 | build: 11 | context: ./frontend/ 12 | dockerfile: Dockerfile 13 | ports: 14 | - 8001:8001 15 | backend: 16 | build: 17 | context: ./ 18 | dockerfile: Dockerfile.backend 19 | ports: 20 | - 8000:8000 21 | secrets: 22 | - groq_key 23 | 24 | secrets: 25 | groq_key: 26 | environment: GROQ_API_KEY -------------------------------------------------------------------------------- /data/pitch_deck.md: -------------------------------------------------------------------------------- 1 | # THE FUNDAMENTAL BOOK 2 | 3 | # Build a Winning Pitch Deck 4 | --- 5 | # Table of Content 6 | 7 | - Introduction 8 | - The Pitch Deck 9 | - Elements 10 | - Story 11 | - Design 12 | - Delivery 13 | - Pitching Exercises 14 | - Investors 15 | - Action Plan 16 | - Vocabulary 17 | 18 | Text: Emma Davidson 19 | 20 | Illustrations: Ryan Ceazar Santua 21 | 22 | Idea, design and curation by the passionate Team of BaseTemplates 23 | 24 | © 2020 BaseTemplates.com 25 | 26 | Please don’t share this book, rip off any content or imagery, or otherwise try to use it for your own gain. If you do share or write about it somewhere (which we actively encourage) please be sure to give BaseTemplates appropriate credit and a link. 27 | 28 | 2 of 75 29 | --- 30 | # Introduction 31 | 32 | Startup fundraising is not easy. Maintaining your enthusiasm and passion for the seed of an idea while you rebound from one investor to the next is exhausting. It is a process that's killed hundreds of startups. At BaseTemplates, we aim to help founders and startup teams launch their ventures successfully with one powerful, affordable pitch deck solution. 33 | 34 | But providing you with just a template? Well, that's like giving you a plane without teaching you to fly. So, we created this helpful book to walk you through the basics, introducing the concept of the pitch deck, its advantages and uses, as well as providing advice and support on planning, design, pitching and choosing investors. Your brilliant idea is just a pitch deck away from becoming a successful startup. Let’s make it happen! 35 | 36 | Max & Julian, 37 | Founders of BaseTemplates 38 | 39 | Made with ♥ by basetemplates.com 40 | 41 | 3 of 75 42 | --- 43 | # Chapter 1 44 | 45 | # The Pitch Deck 46 | 47 | Startup financing is changing rapidly and it is no longer entirely dependent on angel investors and venture capitals. With crowdfunding platforms and new financial technology emerging all the time, entrepreneurs now have more options for financing than ever before. But one thing remains certain, whatever your seeding round looks like, your business should start life as a pitch deck. 48 | --- 49 | # Chapter 1. The Pitch Deck 50 | 51 | # Purpose of a Pitch Deck 52 | 53 | A pitch deck is a short presentation, a visual breakdown of your business model. It is usually created in PowerPoint or Keynote and delivered in a PDF format. Your pitch should tell a story; it should excite potential investors, giving them the information they need to decide whether to risk their capital on your startup. 54 | 55 | Creative and flexible, a pitch deck can be quickly revised, making it easy to adjust the business model to the needs of a changing market or tailor the pitch to specific investors. 56 | 57 | # The Pitch Deck vs. The Business Plan 58 | 59 | Made with ♥ by basetemplates.com 60 | 61 | 5 of 75 62 | --- 63 | # Chapter 1. The Pitch Deck 64 | 65 | The traditional business plan was once the gold standard for seeding rounds. It provided investors with detailed forecasts, carefully planning every aspect of a startup launch and predicting returns years into the future. 66 | 67 | Shorter than a business plan, a pitch deck delivers only the most relevant information. It is not a detailed plan of growth but a brief introduction to a business that's full of potential. You'll only see financial forecasts, competitor, and market research in around two-thirds of pitch decks. Instead, investor pitches focus on the idea, the problem that idea solves, and the team that will make it a reality. 68 | 69 | When you are creating your first pitch deck, it's important to understand the difference between planning your business and pitching it. A pitch focuses on the possibilities; a plan on the details. Pitches adapt to changes in the market; plans try to predict those changes. 70 | 71 | Your pitch deck should paint a picture of the world your product will create. 72 | 73 | Made with ♥ by basetemplates.com 74 | 75 | 6 of 75 76 | --- 77 | # Chapter 1. The Pitch Deck 78 | 79 | # Your Venture's First Prototype 80 | 81 | A pitch deck isn't just for securing funding. Writing your ideas down, defining your business and summarizing all that you want to achieve in less than 20 slides forces you to consider how practical your idea really is. It highlights any initial problems, gives you an opportunity to find solutions, and strengthens your business model. In the early stages of a startup, there is no faster or cheaper way to build your business than a pitch deck. 82 | 83 | A pitch deck is a prototype for your venture. It provides enough detail for other people to quickly understand your idea, allowing your friends, family, and colleagues to offer relevant advice, identify obstacles and suggest improvements. This type of presentation can be swiftly and easily adapted to incorporate new ideas, so when you do come to launch your business, you have a strong prototype and the best possible chance of success. 84 | 85 | Made with ♥ by basetemplates.com 86 | 7 of 75 87 | --- 88 | # Chapter 2 89 | 90 | # Elements 91 | 92 | “You have a very short amount of time to make a first impression. If you've got a long rambling slide deck… you're done.” 93 | 94 | - Naval Ravikant, Co-founder AngelList 95 | --- 96 | # Chapter 2. Elements 97 | 98 | It is important to understand your audience and deliver a pitch that meets their expectations. Some investors will want information on market size and opportunity; others will appreciate customer acquisition data and milestones. To address this, most startups have a basic pitch deck template that includes the essential slides – business model, problem, solution and team – with additional deck slides added to meet the needs of individual investors. 99 | 100 | The best pitch decks use less than 20 slides. The order each slide appears in your presentation is your choice, but most ventures will place their team pages at the beginning or end of the pitch deck. However, experts like Reid Hoffman (LinkedIn) recommend that you open your pitch with the investment thesis and leave the team page until the end. 101 | 102 | To help you understand each pitch deck element, we will be pitching some existing (and a few fictional) BaseTemplates products to you. Each slide you see has been created with our amazing Pitch Deck Template. 103 | 104 | Made with ♥ by basetemplates.com 105 | 106 | 9 of 75 107 | --- 108 | # Chapter 2. Elements 109 | 110 | # Essential Slides 111 | 112 | While every pitch deck is unique, there are ten slides that should feature in every investor pitch: Title, Summary, The Problem, The Solution, The Business Model, Market Size, Customer Acquisition, Competition, The Team and The Ask. Let’s have a closer look at each one. 113 | 114 | Made with ♥ by basetemplates.com 115 | 10 of 75 116 | --- 117 | # Chapter 2. Elements 118 | 119 | # Title Slide 120 | 121 | The title slide is your first and last impression. It will be displayed behind you while you introduce yourself and could be used as a backdrop when you thank your audience for their time, so it is likely to be the most viewed slide in your pitch. Title slides must be eye-catching and memorable, keep the design simple and make sure it includes your logo. 122 | 123 | The best examples of title slides are engaging and straightforward. They deliver a visual overview of the business model and use a catchy tagline to state the mission of the company. While not essential, we always recommend including a date on your title slide as well, since this makes it easier to identify different versions of your pitch deck. 124 | 125 | |Pitch Deck Template|Pitch Deck Template| 126 | |---| 127 | |the easiest way to build your pitch deck and raise money|the easiest way to build your pitch deck and raise money| 128 | |Darrin Warner|dwarner@gmail.com| 129 | |Pitch Deck Template|Pitch Deck Template| 130 | |the easiest way to build your pitch deck and raise money|the easiest way to build your pitch deck and raise money| 131 | |Darrin Warner|d.warner@rgmail.com| 132 | |Pitch Deck Template|Pitch Deck Template| 133 | |the easiest way to build your pitch deck and raise money|the easiest way to build your pitch deck and raise money| 134 | |Pitch Deck Template|Pitch Deck Template| 135 | |Darrin Warner|dwarner@gmail.com| 136 | 137 | All slides built with our Pitch Deck Template 138 | 139 | Made with ♥ by basetemplates.com 140 | 141 | 11 of 75 142 | --- 143 | # Chapter 2. Elements 144 | 145 | # Summary/Overview 146 | 147 | The summary slide provides a brief outline of your business. It should detail the problem you see and your solution. This is a very short slide that is designed to make investors curious about your business. Don't be afraid to make big claims; investors want to see that you are passionate and invested in bringing your idea to life. Keep it simple, explain what your business does in 30 words or less. 148 | 149 | # Summary 150 | 151 | Pitch Deck Template is a presentation template that helps founders to create a visually impressive and effective pitch deck quickly and easily. 152 | 153 | |Strong Team|Raising $500,000|5850,000 Revenue|S250M Market Opportunity| 154 | |---|---|---|---| 155 | |Founded in 2014|1200 customers|15850,000 revenue in 2018| | 156 | |Darrin Warner|Katherine Kim|Ending "Quarter ARR|900.000 $| 157 | |CEO|CTO|2 675.000|S250M Market Opportunity| 158 | |Phillip Mann|Sylvia Estrada|450.000|225.000 $| 159 | |CCO|CMO| | | 160 | 161 | 2014 2015 2016 2017 2018 162 | 163 | All slides built with our Pitch Deck Template 164 | 165 | Made with ♥ by basetemplates.com 166 | --- 167 | # Chapter 2. Elements 168 | 169 | # The Problem 170 | 171 | This is one of the most important slides in your deck. It shows investors how relevant your product is and provides a glimpse into your market. You need your audience to agree with this slide, so keep the design simple and tell them a story while they're looking at it. A tale about a person who benefits from using your product will highlight its relevance. 172 | 173 | If your business doesn't solve a problem – many restaurants and consumer goods shops won't – then you may want to swap the problem deck slide for the opportunity slide instead. 174 | 175 | |The Problem|The Problem| 176 | |---| 177 | |Creating a great pitch deck is time consuming|costly process| 178 | |Lack of Content|Lack of Ideas| 179 | |Investors Are Not Interested|Founders Do Not Know| 180 | |The Problem|The Problem| 181 | |Creating a great pitch deck is time consuming|costly process| 182 | |Lack Of Content|Lack Of Storylines| 183 | |Lack Of Ideas|Lack Of Design| 184 | |Lack Of Font|Lack Of Design| 185 | 186 | All slides built with our Pitch Deck Template 187 | 188 | Made with ♥ by basetemplates.com 189 | 190 | 13 of 75 191 | --- 192 | # Chapter 2. Elements 193 | 194 | # The Solution 195 | 196 | By now, your investors should understand the significance of the problem. They should be invested in the idea of helping you fix it, so the solution you present must be innovative, exciting and memorable. If you have a prototype, now is the time to show it off. If you are pitching a concept, include videos, pictures and illustrations. Encourage investors to ask questions and make sure your audience fully understands your solution. If investors in one pitch struggle to understand the concept, be sure to adjust the slide so that it is clearer for your next pitch. 197 | 198 | # The Solution 199 | 200 | |The Solution|Pitch Deck Template| 201 | |---|---| 202 | |presentation template that helps founders create a visually impressive and effective pitch deck quickly and easily|presentation template that helps founders create a visually impressive and effective pitch deck quickly and easily| 203 | |Content Direction|Editable Layouts| 204 | |5990|Apresentation template that helps founders create a visually impressive and effective pitch deck quickly and easily| 205 | |The Solution|Pitch Deck Template| 206 | |presentation template that helps founders create a visually impressive and effective pitch deck quickly and easily|presentation template that helps founders create a visually impressive and effective pitch deck quickly and easily| 207 | |Content Direction|Visual Direction| 208 | |Editable Layouts|Impress Investors| 209 | | |Raise Money| 210 | | |Grow Your Company| 211 | |All slides built with our Pitch Deck Template|All slides built with our Pitch Deck Template| 212 | 213 | Made with ♥ by basetemplates.com 214 | 215 | 14 of 75 216 | --- 217 | # Chapter 2. Elements 218 | 219 | # The Business Model 220 | 221 | The business model deck slide describes to investors how you will make money. This slide doesn't need to include detailed data and financial forecasts, but it should show that you have used revenue, earnings and profit projections to help shape your business model. 222 | 223 | The best examples of these slides are not data-heavy. They use simple statistics like revenue or market share to highlight how profitable the business could be. The slides themselves will lack detailed facts and figures, but be prepared for investors to ask about the statistics and calculations you used to produce them and make sure you know these answers ahead of your pitch. 224 | 225 | |We charge|$49 per template and $500 for a custom pitch deck| 226 | |---|---| 227 | |10,000 Template Downloads|$490,000| 228 | |1,000 Custom Requests|$500,000| 229 | | |$1,090,000 Revenue| 230 | 231 | # Business Model 232 | 233 | We charge $49 per template and $500 for a custom pitch deck based on monthly subscription plans. 234 | 235 | |Base|Business|Enterprise| 236 | |---|---|---| 237 | |$25/month|$50/month|$100/month| 238 | |Up to 6 users|Up to 9 users|Up to 20 users| 239 | |1GB storage|5GB storage|20GB storage| 240 | |Basic Analytics|Pro Analytics|Dedicated Support| 241 | 242 | All slides built with our Pitch Deck Template 243 | 244 | Made with ♥ by basetemplates.com 245 | --- 246 | # Chapter 2. Elements 247 | 248 | # The Market/The Opportunity (Why Now) 249 | 250 | The opportunity and market slides provide similar information, and you are unlikely to need both in the same pitch. The opportunity slide describes how your business will operate within your industry, the market slide describes the current market. Both can be used to explain why now is the best time to launch and might include market trends, the size of the market and growth potential. The purpose of these slides is to demonstrate that your business has a future and that you know the market and opportunities better than anyone else. 251 | 252 | # The Market 253 | 254 | |World|USA| 255 | |---|---| 256 | |S1B|SSOOM| 257 | |SSOOM|S1OOM| 258 | |Presentation Templates|S65M| 259 | |Europe|S8oM| 260 | | |S150M| 261 | |S250M|S10M| 262 | |Our Market Opportunity|(10% of available market)| 263 | 264 | # The Market 265 | 266 | |Presentation templates online sales growth|USA| 267 | |---|---| 268 | |50 SM|Presentation Templates Sales| 269 | |32 SM|2022| 270 | |25 SM|S50 million| 271 | |32 SM|expected growth| 272 | |40 SM|average price for presentation template| 273 | |2018|S18 million| 274 | |current revenue|S500| 275 | |2014|average price for work on custom pitch deck| 276 | |2015| | 277 | |2016| | 278 | |2017| | 279 | |2018| | 280 | 281 | All slides built with our Pitch Deck Template 282 | 283 | Made with ♥ by basetemplates.com 284 | --- 285 | # Chapter 2. Elements 286 | 287 | # Customer Acquisition (Go To Market) 288 | 289 | Think of this slide as a description of your 'go-to' market. This is where you explain who your customers are and how well you know them. Describe your market, advertising channels, and location. Don't forget to show investors how many potential customers you are targeting. Some founders will use this slide to demonstrate the size or value of the potential market; others might include the costs associated with customer acquisition. The important thing to remember is that the best examples can be read and understood in three seconds or less. Maps, diagrams and headline statistics all work well and grab your audience's attention. 290 | 291 | # Customer Acquisition 292 | 293 | |Online|Founders|Offline| 294 | |---|---|---| 295 | |Social Media|20-60|Sponsorships| 296 | |SEO|Higher Income|Print Ads| 297 | |PPC| |Events| 298 | |Affiliates| |Direct Mail| 299 | 300 | # Customer Acquisition 301 | 302 | |Founders|20 to 60 years old|X| 303 | |---|---|---| 304 | |higher income|Launch|Scale| 305 | |SEO|TV Ad| | 306 | |Webinars|Events| | 307 | |Social Media|Print Ad| | 308 | |SEO & Social Media|Email Campaign|PPC| 309 | |News/Media/PR|Sponsorship| | 310 | |Blogs & Blogging| | | 311 | 312 | All slides built with our Pitch Deck Template 313 | 314 | Made with ♥ by basetemplates.com 315 | 316 | 17 of 75 317 | --- 318 | # Chapter 2. Elements 319 | 320 | # Competition 321 | 322 | The purpose of this deck slide is to demonstrate that you know your industry and can provide an honest assessment of how your company fits into that industry. Include information on other businesses and services that will be competing for the same customers. Don't ignore major competitors to make your business look stronger, instead, focus on your strengths and highlight your competitive advantages. Investors want to see that you have considered every potential obstacle and have a plan to overcome it. 323 | 324 | Keeping a competition slide simple can be difficult, especially if you have many competitors. Plotting the key features of your product on an x and y-axis and using it to compare your services with those offered by your competitors is a tested format. 325 | 326 | # Competition 327 | 328 | |Our product is unique and doesn't have direct competitors|Our sustainable competitive advantages| 329 | |---|---| 330 | |All In One|Google| 331 | |Google|Pitch Deck Template| 332 | |Marketplace|Store| 333 | |Easy To Create|Only Real Cases| 334 | |Unique Services|Google| 335 | |Thousands of low quality templates| | 336 | 337 | # Competition 338 | 339 | |Our product is unique and doesn't have direct competitors|Our product is unique and doesn't have direct competitors| 340 | |---|---| 341 | |PDT|Google| 342 | |Google|Agency| 343 | |All InOneConcept|Template Store| 344 | |Google|Google| 345 | |Thousands of low quality templates| | 346 | 347 | All slides built with our Pitch Deck Template 348 | 349 | Made with ♥ by basetemplates.com 350 | 351 | 18 of 75 352 | --- 353 | # Chapter 2. Elements 354 | 355 | # The Team 356 | 357 | The team slide provides an introduction to each of the key team members. Investors want to see relevant skills and experience; they need to have confidence in your team and your ability to make the venture a success. Keep bios short and to the point (you can always provide an accompanying document with longer introductions) and highlight any key skills that are essential to the business. Some of the best examples of team deck slides are short. They include professional headshots of the main team members and brief introductions – sometimes as simple as three or four key pieces of information i.e. name, position, and biggest professional achievement. Don't be afraid to be creative with this slide; it needs to make an impression. 358 | 359 | # Team 360 | 361 | |We have background, proven track record and vision to succeed|Darrin Warner|Katherine Kim|Sylvia Estrada| 362 | |---|---|---|---| 363 | | |CEO|CCO|CTO| 364 | |Stanford Business School Graduates|Computer Science MSDegree|CPA & Masters in Psychology| | 365 | 366 | # Team 367 | 368 | |We have background, proven track record and vision to succeed|Darrin Warner|Phillip Mann|Katherine Kim|Sylvia Estrada| 369 | |---|---|---|---|---| 370 | | |CEO|CCO|CTO|CMO| 371 | | | | | | | 372 | 373 | # Team 374 | 375 | |Darrin Warner|Katherine Kim| 376 | |---|---| 377 | |CEO|CTO| 378 | |Stanford Business School Graduates|Computer Science MSDegree| 379 | 380 | # Team 381 | 382 | |Tony Woods|Darla Ferguson|Kerry Black|Jordan Elliott| 383 | |---|---|---|---| 384 | |Atlas Ventures|Presentation Expert|AnotherWorld|Atlas Ventures| 385 | 386 | All slides built with our Pitch Deck Template 387 | 388 | Made with ♥ by basetemplates.com 389 | 390 | 19 of 75 391 | --- 392 | # Chapter 2. Elements 393 | 394 | # The Ask 395 | 396 | This slide is about defining what you need from investors and what you will provide in return. Data oriented pitches might include acquisition targets, plans for initial public offerings and details about how the company can be sold. A concept pitch deck is more likely to focus on the business and how you will use investments to drive growth. Whatever style you chose, the ask slide should clearly state what size and type of investment you are looking for and what that investment will deliver (i.e. the milestones you can achieve with it or the resources it will buy). Remember that money isn't the only contribution your audience can make to your startup. Advice, friendship, and expertise are just as critical to making your business a success. 397 | 398 | # The Ask 399 | 400 | We are raising seed round of $500,000 for 10% 401 | 402 | |CAPEX|CAPEX| 403 | |---| 404 | |Product Development|11 %| 405 | |Marketing and Sales|33 %| 406 | |Personnel|43 %| 407 | 408 | |Product Dev:|$150,000| 409 | |---|---| 410 | |New Hires|$250,000| 411 | |Office & Other|$100,000| 412 | 413 | # The Ask 414 | 415 | We are looking for $500k in funds to finish development of our product; release a beta, launch new offices, and hire key staff. 416 | 417 | |Security:|Qualified Conversion:| 418 | |---|---| 419 | |Convertible Promissory Note|259 to the next Financial Round| 420 | |Amount:|Note Conversion at Maturity:| 421 | |Up to $500,000|$5,000,000 Valuation| 422 | |Quarterly Interest:|12% beginning March 2019| 423 | 424 | |$150,000|Beta Launch| 425 | |---|---| 426 | |$250,000|Open New Offices| 427 | |$100,000|Product Development| 428 | | |Public Launch| 429 | 430 | All slides built with our Pitch Deck Template 431 | 432 | Made with ♥ by basetemplates.com 433 | 434 | 20 of 75 435 | --- 436 | # Chapter 2. Elements 437 | 438 | # Optional Slides 439 | 440 | When deciding which slides are essential to your pitch deck, consider whether you are pitching data or a concept. A concept pitch is best suited to a new and innovative product; you are pitching a vision of the future and your plan to make it a reality. These decks might include additional slides such as the vision, mission, and product. 441 | 442 | If you are on your second or third seeding round, or are pitching a business model that already has data to support it, then a data pitch could be more appropriate. This gives investors the opportunity to judge your business based on the strength of the available data and will include additional slides such as financials and traction. 443 | 444 | Made with ♥ by basetemplates.com 445 | 21 of 75 446 | --- 447 | # Chapter 2. Elements 448 | 449 | # Contact 450 | 451 | Strictly speaking, the contact slide is optional. Many of the best pitch deck examples don't include one and instead choose to put an email address on the summary slide or in the footer of other deck slides. While investors will already have your contact information, it may not be readily available to every member of the investment team. If an executive is looking over your pitch deck after your meeting, a contact slide makes it easy for them to ask questions and seek additional information. We recommend including at least two forms of contact information (for example, an email address and a phone number) on the contact slide. This is also the perfect place to showcase any social media accounts the business has. 452 | 453 | |Pitch Deck Template|Pitch Deck Template| 454 | |---|---| 455 | |the easiest way to build your pitch deck and raise money|the easiest way to build your pitch deck and raise money| 456 | |basetemplates.com|basetemplates.com| 457 | |investors@basetemplates.com|investors@basetemplates.com| 458 | 459 | # Thank You 460 | 461 | Darrin Warner 462 | 463 | dwaner@gmail.com 464 | 465 | All slides built with our Pitch Deck Template 466 | 467 | Made with ♥ by basetemplates.com 468 | 469 | 22 of 75 470 | --- 471 | # Chapter 2. Elements 472 | 473 | # The Product 474 | 475 | The product slide is very similar to the solution slide and can be helpful for further explaining your product, particularly if you don't yet have a prototype. 476 | 477 | Use this slide to describe your product in more detail but don't repeat information from the solution slide. Instead, focus on manufacturing timetables, supplier relationships, variations in the product (i.e. size or color) and any patents you own or have pending on the design. These features all add value to your business and demonstrate that your product is viable and competitive. 478 | 479 | # The Product 480 | 481 | # How it works 482 | 483 | Download Template 484 | Add Your Content 485 | Save and Present 486 | 487 | # Pitch Deck Template 488 | 489 | Experience is a key 490 | 491 | Great Visuals 492 | 493 | 53901 599 494 | 495 | # Pitch Deck Template 496 | 497 | # All in One Template 498 | 499 | # The Product 500 | 501 | # How it works 502 | 503 | Sign Up 504 | Set Your Goal 505 | Go 506 | 507 | now 508 | 509 | future 510 | 511 | future 512 | 513 | # Download App 514 | 515 | Template Based 516 | 517 | Standalone App 518 | 519 | Machine Learning & AI 520 | 521 | All slides built with our Pitch Deck Template 522 | 523 | Made with ♥ by basetemplates.com 524 | 525 | 23 of 75 526 | --- 527 | # Chapter 2. Elements 528 | 529 | # Traction 530 | 531 | The traction slide exists to tell investors how successful you already are. Its purpose is to make them feel like your idea will succeed with or without their backing. This is the part of the pitch deck where you show that your assumptions about the market are correct and demonstrate how efficient your team already is. 532 | 533 | # Traction 534 | 535 | | | |Our MRR grew 6 times in the past 12 month| | |We move forward confidently| | 536 | |---|---|---|---|---|---|---| 537 | |$180,000|$140,000| | | | | | 538 | |$135,000|$585,000 total visitors in Google| | | | | | 539 | |$90,000| | | |gross revenues| | | 540 | |$45,000| | | |in 2018| | | 541 | |$0| | | | |template downloads| | 542 | | |1-Nov|1-Jan|1-Mar|1-May|1-Jul|1-Sept| 543 | | | | | | |average template order|average custom order| 544 | 545 | # Traction 546 | 547 | Some of our numbers and customers since product launch 548 | 549 | |13,500|1,500|$585,000| 550 | |---|---|---| 551 | |template downloads|custom requests|total revenue| 552 | 553 | Downloads 554 | 555 | | | |Google|Google|Google|Google|Google|Google| 556 | |---|---|---|---|---|---|---|---| 557 | |53,960| |90%| | | | | | 558 | 559 | repeat customers 560 | 561 | All slides built with our Pitch Deck Template 562 | 563 | Made with ♥ by basetemplates.com 564 | --- 565 | # Chapter 2. Elements 566 | 567 | # Milestones 568 | 569 | A milestone slide celebrates your most important achievements. It shows investors that you are already creating momentum and driving your business forward. 570 | 571 | For those in the early stages of developing their startup, the milestone slide can be used to provide a road map, showing investors and your staff what the next step is. This is an easy way to plan ahead and shows how you intend to make your venture a success. 572 | 573 | |Milestones|Milestones|Milestones|Milestones|Milestones|Milestones|Milestones|Milestones| 574 | |---|---| 575 | |Where and when we are going|Raise Money|New Hires|Gain Traction|Where and when we are going|Raise Money|May|Offline Ad| 576 | |Q1 2018|Q1 2019|Q1 2019|Q1 2019|October 2018|Beta Launch|April|New Service| 577 | |Beta Launch|Public Launch|December|New Hires|March|New Features|January 2018|February| 578 | |Milestones|Milestones|Milestones|Milestones|Milestones|Milestones|Milestones|Milestones| 579 | |Where and when we are going|Raise Money|2019|2020|2021|2022|Where and when we are going|Raise Money| 580 | |Beta Launch|Public Launch|September 2017|October|New Hires|November|New Features|December| 581 | |Offline Ad Campaign|Public Launch|March 2018|New Features|April|New Service|June|IPO| 582 | 583 | All slides built with our Pitch Deck Template 584 | 585 | Made with ♥ by basetemplates.com 586 | --- 587 | # Chapter 2. Elements 588 | 589 | # Financials 590 | 591 | Financials normally include a balance sheet, income, and cash flow statements. Around 58% of pitch decks include a financials slide, which is of more interest to investors if you have launched or have already conducted a seeding round. Businesses just starting out are not usually expected to display financials. 592 | 593 | Financial deck slides do not take a standard form. Some startups will use them to showcase how much they raised during their first seeding round or to list potential buyers for an exit strategy. Others might skip this slide altogether and include an income statement on their traction slide instead, highlighting their venture's success so far. 594 | 595 | # Financials 596 | 597 | # Our income statement projection 598 | 599 | # Our financial projections are more than realistic 600 | 601 | | |2015|2016|2017|2018|2019| 602 | |---|---|---|---|---|---| 603 | |Revenue|800,000|600,000|3,200,000|6,400,000|12,800,000| 604 | |Cost of Goods Sold|200,000|400,000|800,000|1,600,000|3,200,000| 605 | |Gross Profit|600,000|1,200,000|2,400,000|4,800,000|9,600,000| 606 | |Total Expenses|114,000|228,000|456,000|912,000|1,824,000| 607 | |Earning Before Taxes|486,000|972,000|1,944,000|3,888,000|7,776,000| 608 | |Taxes|145,800|291,600|583,200|1,166,400|2,332,800| 609 | |Net Profit|340,200|680,400|1,360,800|2,721,600|5,443,200| 610 | 611 | # Financials 612 | 613 | # Our financial projections are more than realistic 614 | 615 | |Users|Sales| 616 | |---|---| 617 | |2016|40,000| 618 | |2017|3,000,000| 619 | |2018|2,000,000| 620 | |2019|1,000,000| 621 | |2020| | 622 | 623 | All slides built with our Pitch Deck Template 624 | 625 | Made with ♥ by basetemplates.com 626 | 627 | 26 of 75 628 | --- 629 | # Chapter 2. Elements 630 | 631 | # Exit Strategies 632 | 633 | Some companies never recover from the loss of a founding CEO and investors want to know how you plan to ensure your business can survive without you. The exit strategy slide details plans for exit options such as acquisition or IPO, and might include a list of potential buyers. 634 | 635 | # Exit Strategy 636 | 637 | |Google|Google|Google|Acquisition| 638 | |---|---|---|---| 639 | |Why?|Why?|Why?| | 640 | |Allows Livelan (PaloAlto Software) to offer an easy solution for pitch deck creation|Allows Prezi to attract new group of users|Allows Office (Microsoft) to offer an easy solution for pitch deck creation| | 641 | |How?|How?|How?| | 642 | |Pitch Deck Template could be transferred from PPT format into online|Pitch Deck Template could be transferred from PPT format into online|Pitch Deck Template could be supplied free or paid through Office Store| | 643 | 644 | # Exit Strategy 645 | 646 | |Google|Google|Others| 647 | |---|---|---| 648 | |Allows Livelan (PaloAlto Software) to offer easy but comprehensive solution for pitch deck creation|Allows Prezi to attract new group of customers by offering comprehensive solution for pitch deck creation|Google| 649 | |Google|Google|Google| 650 | |Allows Google Docs (Google) to offer easy but comprehensive solution for pitch deck creation|Google| | 651 | 652 | All slides built with our Pitch Deck Template 653 | 654 | Made with ♥ by basetemplates.com 655 | 656 | 27 of 75 657 | --- 658 | # Chapter 2. Elements 659 | 660 | # The Mission and The Vision 661 | 662 | The mission statement is a short, simple slide at the beginning or end of the pitch deck that explains why your company exists. At the prototype stage, the mission slide provides direction for you and your team. During a seeding round, it provides investors with insight into the idea that is driving your venture forward. 663 | 664 | A vision slide is a great addition to any concept pitch. It describes the world your product will help to create or your long-term plan for your business. It shows investors that you have a firm idea of how the business will grow. 665 | 666 | It's 2018. 667 | 668 | # Our Vision 669 | 670 | To become SquareSpace for Presentations 671 | 672 | # Our Mission 673 | 674 | To help founders to create visually impressive and effective pitch deck quickly and easily 675 | 676 | All slides built with our Pitch Deck Template 677 | 678 | Made with ♥ by basetemplates.com 679 | 680 | 28 of 75 681 | --- 682 | # Chapter 2. Elements 683 | 684 | # Partnerships 685 | 686 | It takes more than money to launch a startup, investors want to see who is involved in the project. The partnership slide is not a common slide, but it can be very valuable. 687 | 688 | Include information on any industry experts you are consulting with, as well as strategic relationships with contractors or suppliers that might lower your overheads or help you break into new markets. If you already have backers invested, or have backers from previous successful funding rounds, then include this information as well. If you are associated with a prestigious accelerator program, then now would be the time to showcase that relationship. 689 | 690 | The important thing with this deck slide is to only include partners that offer a strategic benefit; including every family member who has invested will draw focus from the more powerful partners who will drive your success and impress investors. Don't forget to check with your partners – particularly financial backers – that they are happy for you to publicize their involvement. 691 | 692 | # Partnerships 693 | 694 | |Present|Future|Future| | | 695 | |---|---|---|---| 696 | | | |Google|Google| 697 | |Google|Google|Google| | 698 | |Google|Google|More than 7000 incubators worldwide| | 699 | 700 | All slides built with our Pitch Deck Template 701 | 702 | Made with ♥ by basetemplates.com 703 | 704 | 29 of 75 705 | --- 706 | # Chapter 3 707 | 708 | # Story 709 | 710 | “That's what storytellers do. We restore order with imagination. We instill hope again and again.” 711 | 712 | - Walt Disney, Founder The Walt Disney Company 713 | --- 714 | # Chapter 3. Story 715 | 716 | A pitch is about more than the slides you choose and how well they are presented. If you're going to turn your pitch deck into a valuable asset, you need a story. 717 | 718 | # Why Is The Story Important? 719 | 720 | When you're looking for startup funding, remember, stories are easier to recall than facts and figures. A great story will create an emotional link between your business and your investors, highlight the effectiveness of your solution and show them how well you know your audience. 721 | 722 | Your narrative should explain each of your slides, linking them together to tell the story of your venture. You can use characters or elements from your story to answer investors' questions or add personality to a reading deck (we'll discuss these in more detail in next chapter). 723 | 724 | Made with ♥ by basetemplates.com 725 | 726 | 31 of 75 727 | --- 728 | # Chapter 3. Story 729 | 730 | # How To Tell Your Story 731 | 732 | A pitch deck story should have the same elements as a good book – action, excitement, success, and failure. You should include as many details as you can and work hard to get the audience on your side. By the end of your pitch, investors should care about your venture. 733 | 734 | We've provided three examples of a pitch deck story. Each highlights different features of the enterprise and can be tailored to meet the needs of particular investors. 735 | 736 | # The Hero’s Journey 737 | 738 | This is a classic story that follows a tried and tested format: 739 | 740 | 1. Our hero is living a normal life, unaware that anything is wrong. 741 | 2. He encounters a problem that cannot be fixed with the tools available to him. 742 | 3. He sets about building a solution no one has thought of before. 743 | 4. He finds his purpose and dedicates himself to solving that problem. It's a goal that continues to motivate him to this day. 744 | 745 | This is the tale of why you decided to launch your startup. It is personal; it explains where you have come from and what is motivating you. It is the best format for innovative concept pitches, new products and 746 | --- 747 | # Chapter 3. Story 748 | 749 | charities, since it creates a bond between you and your audience, highlighting your experience and showcasing your passion for the project. 750 | 751 | # BaseTemplates: A Hero's Journey 752 | 753 | The BaseTemplates story begins with a student named Max. Touring industry conferences, watching hundreds of presentations, Max noticed two things; ugly slide decks are everywhere, and even the best speeches are ruined by poor presentations. So he set to work fixing it. 754 | 755 | As a freelance Presentation Designer, Max solved the problem one deck at a time. He turned boring slides full of bullet-points into entertaining presentations. He helped startups deliver clear messages, and made great looking pitches for industry leaders. But one man can only do so much, and ugly pitch decks were still everywhere he looked. 756 | 757 | Then he had an idea. What if he could create a smart template, specific to every business and able to meet the needs of individuals? A template that worked perfectly on both PowerPoint and Keynote, with great slide design and helpful visual assets? Today he is launching BaseTemplates, the all-in-one solution that's set to put an end to ugly presentations everywhere. 758 | 759 | # The Customer’s Tale 760 | 761 | This story focuses on your product from your customer's perspective. It is an excellent way to secure startup funding since it shows investors. 762 | 763 | Made with ♥ by basetemplates.com 764 | 765 | 33 of 75 766 | --- 767 | # Chapter 3. Story 768 | 769 | that you know your market and highlights the value of your product to real people. The basic format of the customer's tale: 770 | 771 | 1. Introduce someone with a problem. 772 | 2. Describe his everyday life and the options he explored to fix his problem. 773 | 3. Explain how he discovered your product and how it solves the issue. 774 | 4. Show investors the difference your product has made to this customer. 775 | 776 | Another excellent choice for concept decks, the customer's tale also works very well for those pitching consumer goods and tech startups. It demonstrates your advantage over competitors. 777 | 778 | # BaseTemplates: The Customer's Tale 779 | 780 | Hans and Katie spent two years developing Legion - a plugin that accurately translates email content into three languages. They funded the beta tests themselves and developed a working prototype to take to investors. They developed an excellent network during their testing phase and secured meetings with three angel investors quickly. But explaining their concept and capturing the attention of investors was difficult, and they struggled to secure funding - until a contact introduced them to BaseTemplates. 781 | 782 | Within a day, they had completely re-designed their presentation. The advanced 783 | 784 | Made with ♥ by basetemplates.com 785 | 786 | 34 of 75 787 | --- 788 | # Chapter 3. Story 789 | 790 | Visual effects, icons, images and layouts meant that their deck now accurately showcased their startup. This was the deck investors expected from them, and it was this deck that helped them close their seeding round five weeks later having secured the $150,000 they needed. 791 | 792 | # The Industry Point of View 793 | 794 | This is the tale of your industry and why it needs you. This story shows investors that your startup has a role to play and demonstrates how well you know your market and current industry trends. The industry point of view: 795 | 796 | 1. The industry has always operated according to a set of assumptions based on the environment it grew up in. 797 | 2. Technological, economic or social changes mean that these assumptions simply don't hold true anymore, causing problems for the big players in the industry. 798 | 3. This problem creates a unique opportunity for a new business to step in and take advantage of these changes. 799 | 4. This is the perfect time for that business – your business – to launch. Demonstrate to investors that you understand the impact your product will have. They need to know your venture will be a success with or without them. 800 | 801 | Made with ♥ by basetemplates.com 802 | 803 | 35 of 75 804 | --- 805 | # Chapter 3. Story 806 | 807 | This format works especially well for data pitches and ground-breaking startups. It demonstrates your competitive edge, how your business will fit into the wider industry, and the size of your potential customer base. 808 | 809 | # BaseTemplates: An Industry Point of View 810 | 811 | Today, there are over 1000 online stores selling presentation templates. Each store sells hundreds of templates, and they all look the same; fancy slide designs, fonts, and animations. But how many are really useful? Most founders don't know which template is best suited to their business. Some templates don't work properly on Mac; others require customers to buy additional design elements. In short, building engaging, professional presentations with them is hard. 812 | 813 | That's where BaseTemplates comes in. An all-in-one template that meets the needs of every business or presentation, with great design and visual assets at no extra cost, and easy customization in both PowerPoint or Keynote. Now, presentation creators don't need to download additional resources and can start work on their deck the moment they purchase it. The Content Direction feature provides step-by-step advice to create an engaging deck, whatever your goal, and the company provides support and assistance, as well as regular updates, so every customer knows they're getting the best value for money. 814 | 815 | Figures collected by the Global Entrepreneurship Monitor (GEM) suggest that 150 million businesses are currently seeking funding. BaseTemplates not only provides these founders with the only all-in-one template to create professional pitch decks, but it also supplies presentation creators – academics, industry leaders and business executives – all over the world with a comprehensive solution to create the best presentations. 816 | 817 | Made with ♥ by basetemplates.com 818 | 819 | 36 of 75 820 | --- 821 | # Chapter 4 822 | 823 | # Design 824 | 825 | “Design is not just what it looks like and feels like, design is how it works.” 826 | 827 | - Steve Jobs, Co-Founder, Apple Inc. 828 | 829 | Made with ♥ by basetemplates.com 830 | 831 | 37 of 75 832 | --- 833 | # Chapter 4. Design 834 | 835 | # Does Pitch Deck Design Really Matter? 836 | 837 | The impressions we form in the first few seconds matter. Most investors will judge your pitch deck (and you) on appearance without even realizing it. This means that different colors, shapes and font styles all have a big impact on the success of your pitch. 838 | 839 | Investors see hundreds of pitch presentations a year, and design is one of the easiest ways to ensure they remember yours. Good design enhances your story and makes the key points of the presentation stand out, communicating your story more clearly and effectively. 840 | 841 | # Design Tips 842 | 843 | Investors need to understand your slides within seconds of looking at them, so the amount of written content you include is critical. Always 844 | 845 | Made with ♥ by basetemplates.com 846 | 847 | 38 of 75 848 | --- 849 | # Chapter 4. Design 850 | 851 | Choose images, charts and figures over writing and, if you do have to include written material, use as few words as possible. 852 | 853 | Every deck slide should have the same style: margins, color scheme, font size, and visual assets. These small details ensure your presentation looks smart, professional and well considered. 854 | 855 | # Page Margins 856 | 857 | Avoid large or very small page margins, as this makes your slides look crowded. We recommend a margin between 1.5cm and 3cm. 858 | 859 | # Color Scheme 860 | 861 | Your color scheme should enhance your deck and complement your logo. Pick two or three main colors for your deck and use shades of the same color for variation. A highlighter color to showcase important content is also a great idea. 862 | 863 | # Font Size 864 | 865 | Choose no more than two font sizes; one for headings and another for everything else. Make sure all fonts can be seen from across a room (we recommend at least 50px) but are not so large that words are split between one line and the next. 866 | 867 | # Visual Assets 868 | 869 | Make sure you use high resolution images and avoid any that would not occur naturally. Charts, graphs, and diagrams must 870 | 871 | Made with ♥ by basetemplates.com 872 | 873 | 39 of 75 874 | --- 875 | # Chapter 4. Design 876 | 877 | Complement your color scheme and fit neatly onto the slide. Avoid 3D images, which can be difficult to interpret. 878 | 879 | # Slide Transitions and Animations 880 | 881 | Avoid using transitions and animations, unless you have slides showing a complex process or data, in which case introducing elements one at a time can be helpful. 882 | 883 | # Deck Formats 884 | 885 | There are two types of pitch decks: presentation decks and reading decks. 886 | 887 | # Presentation Deck 888 | 889 | Contain very little information, it is designed to accompany a verbal pitch and should be used to highlight key points. 890 | 891 | # Reading Deck 892 | 893 | Is not presented to investors in person. It must include enough information to give readers a detailed overview of your business model, team, and solution. 894 | 895 | As you would expect, some design tips work best for pitch presentations while others are better suited to reading decks. Here, we share some of the best design tips for both. 896 | 897 | Made with ♥ by basetemplates.com 898 | 899 | 40 of 75 900 | --- 901 | # Chapter 4. Design 902 | 903 | # Presentation Deck 904 | 905 | When it comes to pitch decks, there is one point every successful founder should know – use as few words as possible. Your deck slides are there to introduce an idea; it is your verbal pitch that will provide the majority of the information. Always use images and charts instead of words and limit yourself to 10 words per slide. 906 | 907 | |Why Now|Exit Strategy| 908 | |---|---| 909 | |Startups|Google| 910 | |VC and Angel|Google| 911 | |Pitch Deck|Google| 912 | |Easy Access|Make business plan| 913 | |Attract new group|Attract new group| 914 | |of customers|of customers| 915 | | |software even more powerful| 916 | 917 | All slides built with our Pitch Deck Template 918 | 919 | Top Tip. Once you have written your presentation deck, delete any word that is not a header. Can you rephrase the header so that you can still understand the slide? 920 | 921 | Made with ♥ by basetemplates.com 922 | 923 | 41 of 75 924 | --- 925 | # Chapter 4. Design 926 | 927 | # Reading Deck 928 | 929 | A reading presentation deck has a lot of work to do. Investors should be able to glance at your deck and understand within minutes what your venture is about. The slides have to say everything you would have said in person in the shortest time possible. 930 | 931 | Each slide has two purposes; to make a point, and to back it up with data. The most effective slides make one point and one point only. Using the same slide to deliver multiple ideas makes it harder to read and understand. 932 | 933 | Select words carefully, each one must convey a meaning. Steer clear of useless adjectives and impressive phrases – investors should be able to see that your customer acquisition strategy is effective, your product exciting, you don't need to waste space on lengthy phrases. 934 | 935 | |The Problem|The Ask| 936 | |---|---| 937 | |Creating great pitch deck is time consuming and costly process|$500,000| 938 | |We are looking for $500,000 in funds to finish development of our product; beta, launch new offices, hire key staff and release| | 939 | |Lack of Content| | 940 | |Founders don’t have enough content for creating a great storyline| | 941 | |Lack of Ideas| | 942 | |Founders don’t have enough ideas for presenting the content| | 943 | |Lack of Design| | 944 | |Founders simply don't have enough relevant skills to create| | 945 | 946 | # Why Now 947 | 948 | Growth. Venture Capital. Angel investors. 949 | 950 | Startups on the rise again, despite historical lows in recent years. 565,000 startups launched each year just in the United States cumulatively last year. 951 | 952 | Raising money from investors is challenging and requires a great pitch deck. PowerPoint and Keynote are the most popular software for pitch deck creation. 953 | 954 | Founders spent more than 50 hours creating a pitch deck. Average person with computer has installed PowerPoint or Keynote. 955 | 956 | All slides built with our Pitch Deck Template. Made with ♥ by basetemplates.com 957 | --- 958 | # Chapter 4. Design 959 | 960 | NO: Our comprehensive customer acquisition strategy delivers effective cross channel returns. 961 | 962 | YES: Average spend per sales lead. 963 | 964 | |Social Media|$0.70| 965 | |---|---| 966 | |Location-based|$1.60| 967 | |Print|$2.60| 968 | 969 | Top Tip: Split your reading deck slides into three, with a title at the top, a 'mission statement' in the middle, and a more detailed explanation (no more than 50 words) at the bottom. 970 | 971 | Made with ♥ by basetemplates.com 972 | 43 of 75 973 | --- 974 | # Chapter 4. Design 975 | 976 | # Useful Resources 977 | 978 | At BaseTemplates, we use a number of resources to help create the best presentations; here are a few of our favorites. Most of them are free, and all are really helpful. Enjoy! 979 | 980 | - Create or find Color Palette 981 | - Find new font 982 | - Free images (limited selection) 983 | - Paid images (industry specific) 984 | - Replace words with icons 985 | - Create Product Screenshots 986 | 987 | Made with ♥ by basetemplates.com 988 | --- 989 | # Chapter 5 990 | 991 | # Delivery 992 | 993 | Don't underestimate the importance of the finishing touches. You've spent hours on your presentation deck, investing just a few more in the delivery stage will ensure investors take note. 994 | --- 995 | # Chapter 5. Delivery 996 | 997 | # Tips for a Printed Deck 998 | 999 | When it comes to the printed deck, presentation is everything. Take your deck slides to a print shop and get them properly bound. This shows investors how serious you are, and means your deck will remain neat and ordered no matter how many people thumb through it. 1000 | 1001 | We recommend steel binding, which is the professional standard. If you can not find a print shop with a steel binder, then spiral binding is the next best option. Avoid comb binding, which can fall apart and do not, under any circumstances, use a staple. 1002 | 1003 | Print your deck onto a heavy weight paper, we recommend over 120gsm, which makes your deck feel like a high-quality product. It is also less likely to cause paper cuts (an added benefit to investors). Each slide should have a page of its own, and you should include an 1004 | 1005 | Made with ♥ by basetemplates.com 1006 | --- 1007 | # Chapter 5. Delivery 1008 | 1009 | additional 'overview' page which displays all the slides together for easy viewing. 1010 | 1011 | # Tips for a Digital Deck 1012 | 1013 | A digital presentation has one crucial difference to the printed deck; it can be viewed on screens as small as five inches and as large as 21. 1014 | 1015 | # PDF 1016 | 1017 | Smaller than Keynote and PowerPoint files, a PDF is the best format to deliver your digital deck in. The layout remains the same across Windows, Apple, and Android operating systems, plus you have the added benefit of a password protection option which gives you some degree of control over who views your deck. 1018 | 1019 | We recommend using docsend.com, an analytics dashboard for documents that lets you keep track of who read your deck and how long they spent on it. It is also useful to host your deck on a cloud service like Dropbox and provide investors with an access link since this allows you to update the deck even after investors have received it. 1020 | 1021 | Startup investor decks take time to create and even longer to practice. 1022 | 1023 | Made with ♥ by basetemplates.com 1024 | 47 of 75 1025 | --- 1026 | # Chapter 6 1027 | 1028 | # Pitching Exercises 1029 | 1030 | “It usually takes me more than three weeks to prepare a good impromptu speech.” 1031 | 1032 | - Mark Twain, Writer and Humorist 1033 | --- 1034 | # Chapter 6. Pitching Exercises 1035 | 1036 | Startup investor decks take time to create and even longer to practice. We've provided a breakdown of some of our favorite pitching exercises to help get you off to a flying start. 1037 | 1038 | # 30 Word Answers 1039 | 1040 | |Success managing|Eating|4-29 words more| 1041 | |---|---|---| 1042 | |Can +|Line +|Pitching| 1043 | |Time|<15 minutes|Requires| 1044 | |A friend who knows nothing about your venture| | | 1045 | |Purpose|Control how your investors view your business and the questions they ask.| | 1046 | 1047 | Made with ♥ by basetemplates.com 1048 | 1049 | 49 of 75 1050 | --- 1051 | # Chapter 6. Pitching Exercises 1052 | 1053 | In 30 words or less, describe your venture. Sounds easy? Well, after those first 30 words, you can only speak again when your viewer asks you a question. And you only have 30 words to answer it with. 1054 | 1055 | This exercise takes some practice, but it is ideal for thinking about the information you need in your pitch. Use your 30 words to encourage your viewer to ask another question, try to shape the direction of their questions until you have included all the key points you want an investor to know. 1056 | 1057 | Optional: When you have finished, ask your viewer to summarize what they think your venture is about. Do they understand your idea? Is your business model clear? What else do they need to know? 1058 | 1059 | d.School Frameworks 1060 | 1061 | Made with ♥ by basetemplates.com 1062 | --- 1063 | # Chapter 6. Pitching Exercises 1064 | 1065 | Time: 30 minutes each 1066 | 1067 | Requires: A pen and a paper 1068 | 1069 | Purpose: Discover how to tell your story 1070 | 1071 | These are exercises from Stanford's renowned d.school, a course for startups and would-be entrepreneurs. 1072 | 1073 | 1. The One Word Summary. 1074 | Find one word that describes everything you want your audience to understand, think, and feel about your venture. 1075 | 1076 | For BaseTemplates, that word might be: Comprehensive 1077 | 2. The 3-Act Pitch. 1078 | Turn your pitch into a three act story, using no more than 30 words for each act. 1079 | 1080 | - Act 1. We meet our hero. 1081 | - Act 2. We discover a problem. 1082 | - Act 3. Our hero finds the solution. 1083 | 1084 | Made with ♥ by basetemplates.com 1085 | 1086 | 51 of 75 1087 | --- 1088 | # Chapter 6. Pitching Exercises 1089 | 1090 | For BaseTemplates, this exercise might look something like this: 1091 | 1092 | Act 1: Max is a presentation designer, he works with startup founders and industry leaders to create beautiful and clear presentations. 1093 | 1094 | Act 2: Every conference Max visits it's the same thing, a parade of boring bullet-point presentations. The problem is, not everyone knows where to find a professional designer to help him improve his pitch. 1095 | 1096 | Act 3: So Max made an affordable template that allows anyone to design an eye-catching presentation. Now there's no excuse for a lousy pitch! 1097 | 1098 | # 3. The Five-Star Review. 1099 | 1100 | Imagine you are your venture's customer and you are writing a 50 word, five-star review of the business, its products or services. What would this person say? 1101 | 1102 | # 4. The Obituary. 1103 | 1104 | This exercise isn't for everyone. Imagine it is 70 (80, 90, 100) years in the future. Your venture was more successful than you could ever have imagined and you are writing your obituary for the Wall Street Journal. Describe your company, the impact it had on the industry and its legacy. 1105 | 1106 | Made with ♥ by basetemplates.com 1107 | --- 1108 | # Chapter 6. Pitching Exercises 1109 | 1110 | # Practice Makes Perfect 1111 | 1112 | Time 1113 | 1114 |     <30 minutes 1115 | 1116 | Requires 1117 | 1118 |     A complete pitch deck and a willing audience 1119 | 1120 | Purpose 1121 | 1122 |     To learn your pitch deck 1123 | 1124 | A good pitch requires practice. Start by running through your pitch on your own. Once you know each of your presentation slides well, you can face a real audience. Ask them to interrupt you, to jump in with questions and demand extra facts and supporting data. Learn how to handle these interruptions without losing the momentum of your pitch. 1125 | 1126 | Optional: Consider recording your presentation. Take note of the places where your pretend investors ask questions and the questions they ask. Could you prevent them from asking by reordering the slides or adding in new ones? Is something in your presentation deck unclear? How can you make your key points clearer? 1127 | 1128 | Made with ♥ by basetemplates.com 1129 | 53 of 75 1130 | --- 1131 | # Chapter 7 1132 | 1133 | # Investors 1134 | 1135 | “Always choose your investors based on who you want to work with, be friends with, and get advice from. Never, ever, choose your investors based on valuation.” 1136 | 1137 | - Jason Goldberg, founder and CEO of Hem and Fab 1138 | --- 1139 | # Chapter 7. Investors 1140 | 1141 | # Types of Investors 1142 | 1143 | For those seeking seed capital, there are five primary funding sources: friends and family, crowdfunding, accelerator programs, angel investors and venture capitals (VCs). Each group invests in startups at different stages and typically contribute varying levels of funding. As such, founders will need to consider which investor is best suited to their venture, its current needs and long-term objectives. 1144 | 1145 | |Venture Capitals|Angels|Self|Friends & Family|Crowdfunding|Accelerators| 1146 | |---|---|---|---|---|---| 1147 | |Seed stages|Seed stages|Seed stages|Seed stages|Seed stages|Seed stages| 1148 | |Early growth|Early growth|Early growth|Early growth|Early growth|Early growth| 1149 | |Late|Late|Late|Late|Late|Late| 1150 | 1151 | Made with ♥ by basetemplates.com 1152 | 1153 | 55 of 75 1154 | --- 1155 | # Chapter 7. Investors 1156 | 1157 | # Self Funded 1158 | 1159 | Founders who support projects themselves maintain complete ownership of their businesses and don't have to accommodate others in the decision-making process. It is also the cheapest way to finance your venture. However, you may miss out on the networking opportunities and expert advice a successful seeding round can generate. Funds for these projects typically come from re-invested profits or 'bootstrapping', where founders sell other services and invest that revenue into the business. 1160 | 1161 | # Friends and Family 1162 | 1163 | Friends and family will most likely invest in your venture to support you, but you should make it clear that they are buying equity in your company, and how much. Make sure they understand the risks involved; most startups fail, and they probably won't get their money back. 1164 | 1165 | Made with ♥ by basetemplates.com 1166 | 1167 | 56 of 75 1168 | --- 1169 | # Chapter 7. Investors 1170 | 1171 | Investments from friends and family are a great way to raise relatively small amounts – around $5,000 – for the initial startup phase. 1172 | 1173 | # Crowdfunding 1174 | 1175 | Platforms like Indiegogo and Kickstarter allow lots of strangers to invest small sums. With five-figure campaigns becoming more and more common, this funding method is growing in popularity among those looking to raise seed capital. Of course, no one wants something for nothing, and founders must choose whether to sell equity or reward backers with perks such as discounted products or early delivery. Campaigns only get funded if they meet their pledge goal. There is a high failure rate but, for those that are successful, there are additional benefits to running a campaign, which can be used to market a product, increase pre-sales, and test pricing. 1176 | 1177 | # Accelerators 1178 | 1179 | Accelerators don't invest in your venture to the same extent that an angel investor or venture capital might, but they do offer a gateway to funding. These closed groups provide selected startups with access to a network of mentors, investors, suppliers, vendors and other useful resources. 1180 | --- 1181 | # Chapter 7. Investors 1182 | 1183 | contacts. They also supply educational services and advice to help refine your business model. 1184 | 1185 | Approved startups typically give up equity in exchange for access to an accelerator's network. When applying to an accelerator group, it is important to research them thoroughly. Many are industry specific, and some are a lot better than others. 1186 | 1187 | # Angel Investors 1188 | 1189 | Angel investors are wealthy individuals or small trusts that personally invest in your venture. They are often successful entrepreneurs and can provide valuable advice and industry contacts as well as seed funding. 1190 | 1191 | An angel investor buys equity in your company in exchange for future profits but, unlike other financing methods, they often take an interest in business operations. As such, it is important to research angels before you approach them. Make sure any angel investors you consider. 1192 | --- 1193 | # Chapter 7. Investors 1194 | 1195 | Partnering with have a good reputation, as well as the skills and contacts to help your business grow. 1196 | 1197 | # Venture Capitals (VCs) 1198 | 1199 | Like angel investors, venture capitals provide seed funds in exchange for equity. Unlike angels, who typically commit relatively small sums (up to $500,000), venture capitals have the resources to invest millions of dollars in your business. 1200 | 1201 | These firms pool money from institutional investors, pension funds and insurance companies and invest in high-risk enterprises. The high-risk nature of their investments mean they are expected to deliver their clients (i.e. the companies that contributed the money) very high returns. As such, VCs often take an active role in the direction of the company and may prioritize revenue over the founder's original vision. 1202 | 1203 | Made with ♥ by basetemplates.com 1204 | 1205 | 59 of 75 1206 | --- 1207 | # Chapter 7. Investors 1208 | 1209 | # Approaching Investors 1210 | 1211 | How you approach investors is critical. Cold calls and emails can ruin your chance of getting funded, seek personal introductions instead. 1212 | 1213 | # Accelerators 1214 | 1215 | You will find information on the accelerator application process on their website or by approaching them directly. Be selective about which groups you approach – submitting multiple applications is time-consuming and looks desperate. 1216 | 1217 | Be sure to research the group before you apply; there are plenty of bad accelerator programs out there. Find out as much information as you can about their network and other startups they have worked with. 1218 | 1219 | When you have found the right program, follow their instructions for joining and make sure your venture meets the program specifications before submitting. 1220 | 1221 | # Angel Investors 1222 | 1223 | Personal introductions are the best way to approach an angel investor. Use LinkedIn and your professional and personal networks to find a common connection who is willing to endorse you. If you don't have someone, then use conferences and industry events to seek an introduction. If all else fails, pick up the phone and try to arrange a meeting. 1224 | 1225 | Made with ♥ by basetemplates.com 1226 | 1227 | 60 of 75 1228 | --- 1229 | # Chapter 7. Investors 1230 | 1231 | Remember, don't ask for money straight away. You are building a two-way relationship, and you need to know that any angel investor you partner with is the right fit for your business. 1232 | 1233 | # Venture Capitals 1234 | 1235 | Again, personal introductions work best. Venture Capitals are huge companies with large networks, so finding a common connection should be easier than with an angel investor. 1236 | 1237 | These companies are regularly approached by startups so, when you do get an introduction, be ready to move quickly. Send a reading deck the same day that you exchange business cards and never go into a presentation pitch without additional documentation ready to send should the VC ask for it. 1238 | 1239 | Made with ♥ by basetemplates.com 1240 | 1241 | 61 of 75 1242 | --- 1243 | # Chapter 8 1244 | 1245 | # Action Plan 1246 | 1247 | “I knew that if I failed I wouldn't regret that, but I knew the one thing I might regret is not trying.” 1248 | 1249 | - Jeff Bezos, founder and CEO of Amazon 1250 | --- 1251 | # Chapter 8. Action Plan 1252 | 1253 | The actions you take to produce your finished pitch deck depends on its purpose. If you are pitching a concept, developing a prototype, or seeking seed capital, then your deck will begin with your initial idea. More established ventures will already have a working understanding of their business model and may choose to skip ahead, bypassing the prototype stage. 1254 | 1255 | # 1. The Prototype 1256 | 1257 | Consider how your business will run day-to-day; how it will make money and who your customers will be. Identify companies that are already producing similar products, who their clients are and the markets they trade in. Consider recent market trends and future predictions and define how your business will operate in this market. Don't worry about the story of your venture, pitch deck design or even which slides you will include at this point. Instead, focus on creating a simple deck that describes you, your idea, the problem it solves, the market, competitors and business model. 1258 | 1259 | Congratulations! You have a prototype business. 1260 | 1261 | # 2. Feedback and Advice 1262 | 1263 | Ask friends, family and industry contacts to take a look at your pitch deck. Do they have any suggestions on how to improve your business. 1264 | 1265 | Made with ♥ by basetemplates.com 1266 | 1267 | 63 of 75 1268 | --- 1269 | # Chapter 8. Action Plan 1270 | 1271 | model? Insight into the market? Would they like to see more information on one slide, or less? This step is about perfecting your business model, use the feedback from your network to adjust your pitch deck slides. 1272 | 1273 | # 3. Define Your Story 1274 | 1275 | The first step to creating your pitch deck is to decide which story it will tell. You will need different stories for different investors so draft two or three; The Hero's Journey, Customers Tale and Industry Point of View we showed you in chapter 3 make excellent starting points. 1276 | 1277 | Write each story down and, once you are happy that it is perfect, split it into sections based on the deck slide types we discussed in part three. By dividing the story in this way, we have identified the slides we need to tell it and the order they should appear in the deck. 1278 | 1279 | Remember, the story will develop as you create your deck, and you may find yourself adding or removing slides at a later date. 1280 | 1281 | Made with ♥ by basetemplates.com 1282 | 1283 | 64 of 75 1284 | --- 1285 | # Chapter 8. Action Plan 1286 | 1287 | # 4. Create Investor Deck 1288 | 1289 | MyCompany 1290 | 1291 | Creating better 1292 | 1293 | Use the information you've gathered in the previous steps to storyboard your pitch deck. Sketch out your story with text, pictures and charts. Choose a color scheme that complements your logo and a font that showcases the personality of your brand. Now you're ready to really design your deck with the help of a professional designer or our Pitch Deck Template. 1294 | 1295 | # 5. Practice Pitching 1296 | 1297 | Practicing your pitch (see chapter 6 for pitching exercises) early will give you confidence in your venture. It also keeps critical information fresh in your mind, so if you do meet an investor by chance, you know exactly what to say to them. 1298 | 1299 | Made with ♥ by basetemplates.com 1300 | 1301 | 65 of 75 1302 | --- 1303 | # Chapter 8. Action Plan 1304 | 1305 | # 6. Approach Investors 1306 | 1307 | Decide which type of investor is best suited to your venture (see chapter 7 for advice) and define your ask. Only approach business investors who are likely to support your startup. Create a list of the top 30 investors, any connections you have in common with them, and the conferences or industry events they are likely to attend. A personal introduction is the best way to approach investors. 1308 | 1309 | # 7. Present Your Deck 1310 | 1311 | Remember, the little things count. Printed decks should be printed on high quality paper and properly bound. Digital decks should be sent as a PDF file (see chapter 5 for more information). When presenting use PDF format as well and avoid native PowerPoint or Keynote formats, unless is really important to use animation features for your slides. 1312 | 1313 | Made with ♥ by basetemplates.com 1314 | 1315 | 66 of 75 1316 | --- 1317 | # Chapter 8. Action Plan 1318 | 1319 | # 8. Get Funded! 1320 | 1321 | Wow - I just wanted to say a HUGE thank you. You just finished reading the whole pitch deck guide. Now you prepared to build your kick-ass deck and get funded. 1322 | 1323 | If you need more specific help or questions remain open, you can give us a mail at hi@basetemplates.com at any time. We're glad to help you on the journey to build your business. We also appreciate any feedback on our guide or templates. 1324 | 1325 | Ready to go? 1326 | 1327 | Start building your deck today with our awesome Pitch Deck Template! 1328 | 1329 | 20% off with code: welcome20 1330 | 1331 | Made with ♥ by basetemplates.com 1332 | 1333 | 67 of 75 1334 | --- 1335 | # Vocabulary 1336 | 1337 | # Startup Fundraising Vocabulary 1338 | 1339 | # B 1 L 1340 | 1341 | # Accelerators 1342 | 1343 | See startup accelerators 1344 | 1345 | # AngelList 1346 | 1347 | An online platform designed to help partnered startups achieve success through a network of investors and advisors. 1348 | 1349 | # Angel Investors 1350 | 1351 | A wealthy individual (or group) who use personal finances to buy equity in promising startups. Often entrepreneurs themselves, angel investors may take an active role in the businesses they support. 1352 | --- 1353 | # Vocabulary 1354 | 1355 | # B 1356 | 1357 | # Balance Sheet 1358 | 1359 | A document detailing the current assets, liabilities and capital of a business. 1360 | 1361 | # Beachhead Market 1362 | 1363 | A small test market with similarities to the priority market. Eg. Singapore is a beachhead market for Asia. 1364 | 1365 | # Bootstrapping 1366 | 1367 | A service offered (often by a founder) for the sole purpose of raising revenue to invest in another venture. 1368 | 1369 | # Business Model 1370 | 1371 | A statement of potential revenue streams, customers, vendors, and suppliers that form a plan of business operations. 1372 | 1373 | # Business Plan 1374 | 1375 | A document that defines the business goals, objectives, wider market trends, and finances and details how they can be achieved. 1376 | 1377 | # C 1378 | 1379 | # Cash Flow Statement 1380 | 1381 | A breakdown of a business’s actual or intended incoming and outgoing cash. 1382 | --- 1383 | # Vocabulary 1384 | 1385 | # Common Stockholder 1386 | 1387 | Equity holders with no say in the operation of a business. Typically receive smaller dividends than preferred stockholders. 1388 | 1389 | # Convertible Note 1390 | 1391 | Often issued to companies that have yet to be valued, a convertible note is a loan that can be converted into a set number of shares or cash equivalent at a later date. 1392 | 1393 | # Cost Breakdown 1394 | 1395 | The individual costs associated with the production of a product. 1396 | 1397 | # Customer Acquisition 1398 | 1399 | Strategies employed to generate sales leads. 1400 | 1401 | # Deal Flow (or Dealflow) 1402 | 1403 | A measure of how frequently investors are approached with pitches and business plans. 1404 | 1405 | # Deck Slide 1406 | 1407 | An individual slide in either a reading or presentation pitch deck. 1408 | 1409 | # Dividends 1410 | 1411 | Payments made to shareholders by a business, typically a share of the profits. 1412 | 1413 | Made with ♥ by basetemplates.com 1414 | 1415 | 70 of 75 1416 | --- 1417 | # Vocabulary 1418 | 1419 | # E 1420 | 1421 | Elevator Pitch 1422 | 1423 | A 30-second description of a business model, product or service designed to showcase and sell it. 1424 | 1425 | Equity 1426 | 1427 | An ownership interest in the business, usually held in the form of common or preferred stock. 1428 | 1429 | Equity Financing Agreement 1430 | 1431 | Selling equity to raise capital for business growth. 1432 | 1433 | Exit Strategy 1434 | 1435 | A plan of how the founding team or investors will leave the business. 1436 | 1437 | # F 1438 | 1439 | Financials 1440 | 1441 | A statement of a business’s financial position. Typically includes a balance sheet, income, and cash flow statements. 1442 | 1443 | Fundraising 1444 | 1445 | Raising money from voluntary sources, either by donations or sales. 1446 | 1447 | Made with ♥ by basetemplates.com 1448 | --- 1449 | # Vocabulary 1450 | 1451 | # G 1452 | 1453 | Go to Market 1454 | See customer acquisition. 1455 | 1456 | # I 1457 | 1458 | Income Statement 1459 | An account of a venture’s income and outgoings over a defined accounting period. 1460 | 1461 | Intellectual Property 1462 | Assets with no physical presence but potential value, such as patents, copyrights, and logos. 1463 | 1464 | Investor 1465 | An individual, group or organization that commits money or resources to a business in exchange for equity, interest repayments or a percentage of future profits. 1466 | 1467 | # L 1468 | 1469 | Liquidation Preference 1470 | The sum of money to be paid to preferred stockholders if the business is liquefied, this payment takes priority over those of common shareholders. 1471 | 1472 | Made with ♥ by basetemplates.com 1473 | 1474 | 72 of 75 1475 | --- 1476 | # Vocabulary 1477 | 1478 | # Logo 1479 | 1480 | A unique emblem associated with a business, its brand, products, and services. 1481 | 1482 | # Patent 1483 | 1484 | A legal license recognizing ownership over intellectual property, typically an invention. Prevents others from using the same design without the permission of the patent holder. 1485 | 1486 | # Pitch Deck 1487 | 1488 | A series of visual slides that accompany a speech or verbal pitch. 1489 | 1490 | # Preferred Stockholder 1491 | 1492 | Equity holders with a priority claim to a business’s assets and dividends. They do not generally have voting rights. 1493 | 1494 | # Presentation Pitch 1495 | 1496 | See pitch deck. 1497 | 1498 | # Profit and Loss Account 1499 | 1500 | See income statement. 1501 | 1502 | # Profit Margin 1503 | 1504 | The amount of income that exceeds the cost of production. 1505 | 1506 | Made with ♥ by basetemplates.com 1507 | 1508 | 73 of 75 1509 | --- 1510 | # Vocabulary 1511 | 1512 | # R 1513 | 1514 | Reading Deck 1515 | 1516 | A collection of slides designed to introduce a venture or concept to potential investors and interested parties. Presented alone without an accompanying pitch or additional documentation. 1517 | 1518 | # S 1519 | 1520 | Seed Capital 1521 | 1522 | The funds used to launch a project. Also called seed money. 1523 | 1524 | Seeding Round 1525 | 1526 | Fundraising activities designed to raise a set amount of capital to launch a business, usually through angel investors. 1527 | 1528 | Series A Round 1529 | 1530 | A business’s first fundraising round after the seeding round, typically pitched to venture capitals for funds to support expansion. 1531 | 1532 | Slide Deck 1533 | 1534 | See pitch deck. 1535 | 1536 | Startup 1537 | 1538 | A young business still in the early stages of development or not yet launched. 1539 | 1540 | Made with ♥ by basetemplates.com 1541 | 1542 | 74 of 75 1543 | --- 1544 | # Vocabulary 1545 | 1546 | # Startup Accelerator 1547 | 1548 | Also called seed accelerators. A members-only group that provides selected startups with access to useful business contacts, including investors, suppliers, and advisors, in exchange for equity. 1549 | 1550 | # Statement of Cashflows. 1551 | 1552 | See cashflow statement. 1553 | 1554 | # Traction 1555 | 1556 | The actions taken so far to generate interest in a business, product, or services and the interest they have generated. 1557 | 1558 | # Venture Capitals 1559 | 1560 | Firms that pool money from clients including institutional investors, pension schemes and insurance companies to invest significant sums (often more than $1,000,000) in high-risk ventures and startups. 1561 | 1562 | Made with ♥ by basetemplates.com 1563 | 1564 | 75 of 75 -------------------------------------------------------------------------------- /data/playbook.md: -------------------------------------------------------------------------------- 1 | # Startup Playbook 2 | ## Written by Sam Altman · Illustrated by Gregory Koberger 3 | 4 | We spend a lot of time advising startups. Though one-on-one advice will always be crucial, we thought it might help us scale Y Combinator if we could distill the most generalizable parts of this advice into a sort of playbook we could give YC and YC Fellowship companies. 5 | Then we thought we should just give it to everyone. 6 | This is meant for people new to the world of startups. Most of this will not be new to people who have read a lot of what YC partners have written—the goal is to get it into one place. 7 | There may be a part II on how to scale a startup later—this mostly covers how to start one. 8 | * [Part I: The Idea](https://playbook.samaltman.com/#idea) 9 | * [Part II: A Great Team](https://playbook.samaltman.com/#team) 10 | * [Part III: A Great Product](https://playbook.samaltman.com/#product) 11 | * [Part IV: Great Execution](https://playbook.samaltman.com/#execution) 12 | * [Closing Thought](https://playbook.samaltman.com/#closing) 13 | 14 | 15 | * [Growth](https://playbook.samaltman.com/#growth) 16 | * [Focus & Intensity](https://playbook.samaltman.com/#focus) 17 | * [Jobs of the CEO](https://playbook.samaltman.com/#ceo) 18 | * [Hiring & Managing](https://playbook.samaltman.com/#hiring) 19 | * [Competitors](https://playbook.samaltman.com/#competition) 20 | * [Making Money](https://playbook.samaltman.com/#money) 21 | * [Fundraising](https://playbook.samaltman.com/#fundraising) 22 | 23 | 24 | Your goal as a startup is to make something users love. If you do that, then you have to figure out how to get a lot more users. But this first part is critical—think about the really successful companies of today. They all started with a product that their early users loved so much they told other people about it. If you fail to do this, you will fail. If you deceive yourself and think your users love your product when they don’t, you will still fail. 25 | The startup graveyard is littered with people who thought they could skip this step. 26 | It’s much better to first make a product a small number of users love than a product that a large number of users like. Even though the total amount of positive feeling is the same, it’s much easier to get more users than to go from like to love. 27 | A word of warning about choosing to start a startup: It sucks! One of the most consistent pieces of feedback we get from YC founders is it’s harder than they could have ever imagined, because they didn’t have a framework for the sort of work and intensity a startup entails. Joining an early-stage startup that’s on a rocketship trajectory is usually a much better financial deal. 28 | On the other hand, starting a startup is not in fact very risky to your career—if you’re really good at technology, there will be job opportunities if you fail. Most people are very bad at evaluating risk. I personally think the riskier option is having an idea or project you’re really passionate about and working at a safe, easy, unfulfilling job instead. 29 | To have a successful startup, you need: a great idea (including a great market), a great team, a great product, and great execution. 30 | 31 | ### Part I - **The Idea** 32 | 33 | One of the first things we ask YC companies is what they’re building and why. 34 | We look for clear, concise answers here. This is both to evaluate you as a founder and the idea itself. It’s important to be able to think and communicate clearly as a founder—you’ll need it for recruiting, raising money, selling, etc. Ideas in general need to be clear to spread, and complex ideas are almost always a sign of muddled thinking or a made up problem. If the idea does not really excite at least some people the first time they hear it, that’s bad. 35 | Another thing we ask is who desperately needs the product. 36 | In the best case, you yourself are the target user. In the second best case, you understand the target user extremely well. 37 | If a company already has users, we ask how many and how fast that number is growing. We try to figure out why it’s not growing faster, and we especially try to figure out if users really love the product. Usually this means they’re telling their friends to use the product without prompting from the company. We also ask if the company is generating revenue, and if not, why not. 38 | If the company doesn’t yet have users, we try to figure out the minimum thing to build first to test the hypothesis—i.e., if we work backwards from the perfect experience, we try to figure out what kernel to start with. 39 | The way to test an idea is to either launch it and see what happens or try to sell it (e.g. try to get a letter of intent before you write a line of code.) The former works better for consumer ideas (users may tell you they will use it, but in practice it won’t cut through the clutter) and the latter works better for enterprise ideas (if a company tells you they will buy something, then go build it.) Specifically, if you are an enterprise company, one of the first questions we’ll ask you is if you have a letter of intent from a customer saying they’ll buy what you’re building. For most biotech and hard tech companies, the way to test an idea is to first talk to potential customers and then figure out the smallest subset of the technology you can build first. 40 | It’s important to let your idea evolve as you get feedback from users. And it’s critical you understand your users really well—you need this to evaluate an idea, build a great product, and build a great company. 41 | As mentioned earlier, startups are really hard. They take a very long time, and consistent intense effort. The founders and employees need to have a shared sense of mission to sustain them. So we ask why founders want to start this particular company. 42 | We also ask how the company will one day be a monopoly. There are a lot of different terms for this, but we use Peter Thiel’s. Obviously, we don’t want your company to behave in an unethical way against competitors. Instead, we’re looking for businesses that get more powerful with scale and that are difficult to copy. 43 | Finally, we ask about the market. We ask how big it is today, how fast it’s growing, and why it’s going to be big in ten years. We try to understand why the market is going to grow quickly, and why it’s a good market for a startup to go after. We like it when major technological shifts are just starting that most people haven’t realized yet—big companies are bad at addressing those. And somewhat counterintuitively, the best answer is going after a large part of a small market. 44 | A few other thoughts on ideas: 45 | We greatly prefer something new to something derivative. Most really big companies start with something fundamentally new (one acceptable definition of new is 10x better.) If there are ten other companies starting at the same time with the same plan, and it sounds a whole lot like something that already exists, we are skeptical. 46 | One important counterintuitive reason for this is that it’s easier to do something new and hard than something derivative and easy. People will want to help you and join you if it’s the former; they will not if it’s the latter. 47 | The best ideas sound bad but are in fact good. So you don’t need to be too secretive with your idea—if it’s actually a good idea, it likely won’t sound like it’s worth stealing. Even if it does sound like it’s worth stealing, there are at least a thousand times more people that have good ideas than people who are willing to do the kind of work it takes to turn a great idea into a great company. And if you tell people what you’re doing, they might help. 48 | Speaking of telling people your idea—while it’s important the idea really excites some people the first time they hear it, almost everyone is going to tell you that your idea sucks. Maybe they are right. Maybe they are not good at evaluating startups, or maybe they are just jealous. Whatever the reason is, it will happen a lot, it will hurt, and even if you think you’re not going to be affected by it, you still will be. The faster you can develop self-belief and not get dragged down too much by haters, the better off you’ll be. No matter how successful you are, the haters will never go away. 49 | What if you don’t have an idea but want to start a startup? Maybe you shouldn’t. It’s so much better if the idea comes first and the startup is the way to get the idea out into the world. 50 | We once tried an experiment where we funded a bunch of promising founding teams with no ideas in the hopes they would land on a promising idea after we funded them. 51 | All of them failed. I think part of the problem is that good founders tend to have lots of good ideas (too many, usually). But an even bigger problem is that once you have a startup you have to hurry to come up with an idea, and because it’s already an official company the idea can’t be too crazy. You end up with plausible sounding but derivative ideas. This is the danger of pivots. 52 | So it’s better not to try too actively to force yourself to come up with startup ideas. Instead, learn about a lot of different things. Practice noticing problems, things that seem inefficient, and major technological shifts. [Work on projects you find interesting](http://blog.samaltman.com/projects-and-companies). Go out of your way to hang around smart, interesting people. At some point, ideas will emerge. 53 | ### Part II - **A Great Team** 54 | 55 | Mediocre teams do not build great companies. One of the things we look at the most is the strength of the founders. When I used to do later-stage investing, I looked equally hard at the strength of the employees the founders hired. 56 | What makes a great founder? The most important characteristics are ones like unstoppability, determination, formidability, and resourcefulness. Intelligence and passion also rank very highly. These are all much more important than experience and certainly “expertise with language X and framework Y”. 57 | We have noticed the most successful founders are the sort of people who are low-stress to work with because you feel “he or she will get it done, no matter what it is.” Sometimes you can succeed through sheer force of will. 58 | Good founders have a number of seemingly contradictory traits. One important example is rigidity and flexibility. You want to have strong beliefs about the core of the company and its mission, but still be very flexible and willing to learn new things when it comes to almost everything else. 59 | The best founders are unusually responsive. This is an indicator of decisiveness, focus, intensity, and the ability to get things done. 60 | Founders that are hard to talk to are almost always bad. Communication is a very important skill for founders—in fact, I think this is the most important rarely-discussed founder skill. 61 | Tech startups need at least one founder who can build the company’s product or service, and at least one founder who is (or can become) good at sales and talking to users. This can be the same person. 62 | Consider these criteria when you’re choosing a cofounder -- it’s one of the most important decisions you’ll make, and it’s often done fairly randomly. You want someone you know well, not someone you just met at a cofounder dating thing. You can evaluate anyone you might work with better with more data, and you really don’t want to get this one wrong. Also, at some point, the expected value of the startup is likely to dip below the X axis. If you have a pre-existing relationship with your cofounders, none of you will want to let the other down and you’ll keep going. Cofounder breakups are one of the leading causes of death for early startups, and we see them happen very, very frequently in cases where the founders met for the express purpose of starting the company. 63 | The best case, by far, is to have a good cofounder. The next best is to be a solo founder. The worse case, by far, is to have a bad cofounder. If things are not working out, you should part ways quickly. 64 | A quick note on equity: the conversation about the equity split does not get easier with time—it’s better to set it early on. Nearly equal is best, though perhaps in the case of two founders it’s best to have one person with one extra share to prevent deadlocks when the cofounders have a fallout. 65 | ### Part III - **A Great Product** 66 | 67 | Here is the secret to success: have a great product. This is the only thing all great companies have in common. 68 | If you do not build a product users love you will eventually fail. Yet founders always look for some other trick. Startups are the point in your life when tricks stop working. 69 | A great product is the only way to grow long-term. Eventually your company will get so big that all growth hacks stop working and you have to grow by people wanting to use your product. This is the most important thing to understand about super-successful companies. There is no other way. Think about all of the really successful technology companies—they all do this. 70 | You want to build a “product improvement engine” in your company. You should talk to your users and watch them use your product, figure out what parts are sub-par, and then make your product better. Then do it again. This cycle should be the number one focus of the company, and it should drive everything else. If you improve your product 5% every week, it will really compound. 71 | The faster the repeat rate of this cycle, the better the company usually turns out. During YC, we tell founders they should be building product and talking to users, and not much else besides eating, sleeping, exercising, and spending time with their loved ones. 72 | To do this cycle right, you have to get very close to your users. Literally watch them use your product. Sit in their office if you can. Value both what they tell you and what they actually do. You should not put anyone between the founders and the users for as long as possible—that means the founders need to do sales, customer support, etc. 73 | Understand your users as well as you possibly can. Really figure out what they need, where to find them, and what makes them tick. 74 | “Do things that don’t scale” has rightfully become a mantra for startups. You usually need to recruit initial users one at a time (Ben Silbermann used to approach strangers in coffee shops in Palo Alto and ask them to try Pinterest) and then build things they ask for. Many founders hate this part, and just want to announce their product in the press. But that almost never works. Recruit users manually, and make the product so good the users you recruit tell their friends. 75 | You also need to break things into very small pieces, and iterate and adapt as you go. Don’t try to plan too far out, and definitely don’t batch everything into one big public release. You want to start with something very simple—as little surface area as possible—and launch it sooner than you’d think. In fact, simplicity is always good, and you should always keep your product and company as simple as possible. 76 | Some common questions we ask startups having problems: Are users using your product more than once? Are your users fanatical about your product? Would your users be truly bummed if your company went away? Are your users recommending you to other people without you asking them to do it? If you’re a B2B company, do you have at least 10 paying customers? 77 | If not, then that’s often the underlying problem, and we tell companies to make their product better. I am skeptical about most excuses for why a company isn’t growing—very often the real reason is that the product just isn’t good enough. 78 | When startups aren’t sure what to do next with their product, or if their product isn’t good enough, we send them to go talk to their users. This doesn’t work in every case—it’s definitely true that people would have asked Ford for faster horses—but it works surprisingly often. In fact, more generally, when there’s a disagreement about anything in the company, talk to your users. 79 | The best founders seem to care a little bit too much about product quality, even for seemingly unimportant details. But it seems to work. By the way, “product” includes all interactions a user has with the company. You need to offer great support, great sales interactions, etc. 80 | Remember, if you haven’t made a great product, nothing else will save you. 81 | ### Part IV - **Great Execution - ** 82 | 83 | Although it’s necessary to build a great product, you’re not done after that. You still have to turn it into a great company, and you have to do it yourself—the fantasy of hiring an “experienced manager” to do all this work is both extremely prevalent and a graveyard for failed companies. You cannot outsource the work to someone else for a long time. 84 | This sounds obvious, but you have to make money. This would be a good time to start thinking about how that’s going to work. 85 | The only universal job description of a CEO is to make sure the company wins. You can do this as the founder even if you have a lot of flaws that would normally disqualify you as a CEO as long as you hire people that complement your own skills and let them do their jobs. That experienced CEO with a fancy MBA may not have the skill gaps you have, but he or she won’t understand the users as well, won’t have the same product instincts, and won’t care as much. 86 | ### Part IV: Execution - **Growth** 87 | 88 | Growth and momentum are the keys to great execution. Growth (as long as it is not “sell dollar bills for 90 cents” growth) solves all problems, and lack of growth is not solvable by anything but growth. If you’re growing, it feels like you’re winning, and people are happy. If you’re growing, there are new roles and responsibilities all the time, and people feel like their careers are advancing. If you’re not growing, it feels like you’re losing, and people are unhappy and leave. If you’re not growing, people just fight over responsibilities and blame. 89 | Founders and employees that are burn out nearly always work at startups without momentum. It’s hard to overstate how demoralizing it is. 90 | The prime directive of great execution is “Never lose momentum”. But how do you do it? 91 | The most important way is to make it your top priority. The company does what the CEO measures. It’s valuable to have a single metric that the company optimizes, and it’s worth time to figure out the right growth metric. If you care about growth, and you set the execution bar, the rest of the company will focus on it. 92 | Here are a couple of examples. 93 | The founders of Airbnb drew a forward-looking graph of the growth they wanted to hit. They posted this everywhere—on their fridge, above their desks, on their bathroom mirror. If they hit the number that week, great. If not, it was all they talked about. 94 | Mark Zuckerberg once said that one of the most important innovations at Facebook was their establishment of a growth group when growth slowed. This group was (and perhaps still is) one of the most prestigious groups in the company—everyone knew how important it was. 95 | Keep a list of what’s blocking growth. Talk as a company about how you could grow faster. If you know what the limiters are, you’ll naturally think about how to address them. 96 | For anything you consider doing, ask yourself “Is this the best way to optimize growth?” For example, going to a conference is not usually the best way to optimize growth, unless you expect to sell a lot there. 97 | Extreme internal transparency around metrics (and financials) is a good thing to do. For some reason, founders are always really scared of this. But it’s great for keeping the whole company focused on growth. There seems to be a direct correlation between how focused on metrics employees at a company are and how well they’re doing. If you hide the metrics, it’s hard for people to focus on them. 98 | Speaking of metrics, don’t fool yourself with vanity metrics. The common mistake here is to focus on signups and ignore retention. But retention is as important to growth as new user acquisition. 99 | It’s also important to establish an internal cadence to keep momentum. You want to have a “drumbeat” of progress—new features, customers, hires, revenue milestones, partnerships, etc that you can talk about internally and externally. 100 | You should set aggressive but borderline achievable goals and review progress every month. Celebrate wins! Talk internally about strategy all the time, tell everyone what you’re hearing from customers, etc. The more information you share internally—good and bad—the better you’ll be. 101 | There are a few traps that founders often fall into. One is that if the company is growing like crazy but everything seems incredibly broken and inefficient, everyone worries that things are going to come unraveled. In practice, this seems to happen rarely (Friendster is the most recent example of a startup dying because of technical debt that I can point to.) Counterintuitively, it turns out that it’s good if you’re growing fast but nothing is optimized—all you need to do is fix it to get more growth! My favorite investments are in companies that are growing really fast but incredibly un-optimized—they are deeply undervalued. 102 | A related trap is thinking about problems too far in the future—i.e. “How are we going to do this at massive scale?” The answer is to figure it out when you get there. Far more startups die while debating this question than die because they didn’t think about it enough. A good rule of thumb is to only think about how things will work at 10x your current scale. Most early-stage startups should put “Do things that don’t scale” up on their wall and live by it. As an example, great startups always have great customer service in the early days, and bad startups worry about the impact on the unit economics and that it won’t scale. But great customer service makes for passionate early users, and as the product gets better you need less support, because you’ll know what customers commonly struggle with and improve the product/experience in those areas. (By the way, this is a really important example—have great customer support.) 103 | There’s a big catch to this—”Do things that don’t scale” does not excuse you from having to eventually make money. It’s ok to have bad unit economics in the early days, but you have to have a good reason for why the unit economics are going to work out later. 104 | Another trap is getting demoralized because growth is bad in absolute numbers even though it’s good on a percentage basis. Humans are very bad at intuition around exponential growth. Remind your team of this, and that all giant companies started growing from small numbers. 105 | Some of the biggest traps are the things that founders believe will deliver growth but in practice almost never work and suck up a huge amount of time. Common examples are deals with other companies and the “big press launch”. Beware of these and understand that they effectively never work. Instead get growth the same way all great companies have—by building a product users love, recruiting users manually first, and then testing lots of growth strategies (ads, referral programs, sales and marketing, etc.) and doing more of what works. Ask your customers where you can find more people like them. 106 | Remember that sales and marketing are not bad words. Though neither will save you if you don’t have a great product, they can both help accelerate growth substantially. If you’re an enterprise company, it’s likely a requirement that your company get good at these. 107 | Don’t be afraid of sales especially. At least one founder has to get good at asking people to use your product and give you money. 108 | Alex Schultz [gave a lecture on growth for consumer products](http://startupclass.samaltman.com/courses/lec06/) that’s well worth watching. For B2B products, I think the right answer is almost always to track revenue growth per month, and remember that the longer sales cycle means the first couple of months are going to look ugly (though sometimes selling to startups as initial customers can solve this problem). 109 | ### Part IV: Execution - **Focus & Intensity** 110 | 111 | If I had to distill my advice about how to operate down to only two words, I’d pick focus and intensity. These words seem to really apply to the best founders I know. 112 | They are relentlessly focused on their product and growth. They don’t try to do everything—in fact, they say no a lot (this is hard because the sort of people that start companies are the sort of people that like doing new things.) 113 | As a general rule, don’t let your company start doing the next thing until you’ve dominated the first thing. No great company I know of started doing multiple things at once—they start with a lot of conviction about one thing, and see it all the way through. You can do far fewer things than you think. A very, very common cause of startup death is doing too many of the wrong things. Prioritization is critical and hard. (Equally important to setting the company’s priorities is setting your own tactical priorities. What I’ve found works best for me personally is a pen-and-paper list for each day with ~3 major tasks and ~30 minor ones, and an annual to-do list of overall goals.) 114 | While great founders don’t do many big projects, they do whatever they do very intensely. They get things done very quickly. They are decisive, which is hard when you’re running a startup—you will get a lot of conflicting advice, both because there are multiple ways to do things and because there’s a lot of bad advice out there. Great founders listen to all of the advice and then quickly make their own decisions. 115 | Please note that this doesn’t mean doing everything intensely—that’s impossible. You have to pick the right things. As Paul Buchheit says, find ways to get 90% of the value with 10% of the effort. The market doesn’t care how hard you work—it only cares if you do the right things. 116 | It’s very hard to be both obsessed with product quality and move very quickly. But it’s one of the most obvious tells of a great founder. 117 | I have never, not once, seen a slow-moving founder be really successful. 118 | You are not different from other startups. You still have to stay focused and move fast. Companies building rockets and nuclear reactors still manage to do this. All failing companies have a pet explanation for why they are different and don’t have to move fast. 119 | When you find something that works, keep going. Don’t get distracted and do something else. Don’t take your foot off the gas. 120 | Don’t get caught up in early success—you didn’t get off to a promising start by going to lots of networking events and speaking on lots of panels. Startup founders who start to have initial success have a choice of two paths: either they keep doing what they’re doing, or they start spending a lot of time thinking about their “personal brand” and enjoying the status of being a founder. 121 | It’s hard to turn down the conferences and the press profiles—they feel good, and it’s especially hard to watch other founders in your space get the attention. But this won’t last long. Eventually the press figures out who is actually winning, and if your company is a real success, you’ll have more attention than you’ll ever want. The extreme cases—early-stage founders with their own publicists—that one would think only exist in TV shows actually exist in real life, and they almost always fail. 122 | Focus and intensity will win out in the long run. (Charlie Rose once said that things get done in the world through a combination of focus and personal connections, and that’s always stuck with me.) 123 | ### Part IV: Execution - **Jobs _of the_ CEO** 124 | 125 | Earlier I mentioned that the only universal job description of the CEO is to make sure the company wins. Although that’s true, I wanted to talk a little more specifically about how a CEO should spend his or her time. 126 | A CEO has to 1) set the vision and strategy for the company, 2) evangelize the company to everyone, 3) hire and manage the team, especially in areas where you yourself have gaps 4) raise money, and 5) set the execution quality bar. 127 | In addition to these, find whatever parts of the business you love the most, and stay engaged there. 128 | As I mentioned at the beginning, it’s an intense job. If you are successful, it will take over your life to a degree you cannot imagine—the company will be on your mind all the time. Extreme focus and extreme intensity means it’s not the best choice for work-life balance. You can have one other big thing—your family, doing lots of triathlons, whatever—but probably not much more than that. You have to always be on, and there are a lot of decisions only you can make, no matter how good you get at delegation. 129 | You should aim to be super responsive to your team and the outside world, always be clear on the strategy and priorities, show up to everything important, and execute quickly (especially when it comes to making decisions others are blocked on.) You should also adopt a “do whatever it takes” attitude—there will be plenty of unpleasant schleps. If the team sees you doing these things, they will do them too. 130 | Managing your own psychology is both really hard and really important. It’s become cliché at this point, but it’s really true—the emotional highs and lows are very intense, and if you don’t figure out how to stay somewhat level through them, you’re going to struggle. Being a CEO is lonely. It’s important to have relationships with other CEOs you can call when everything is melting down (one of the important accidental discoveries of YC was a way for founders to have peers.) 131 | A successful startup takes a very long time—certainly much longer than most founders think at the outset. You cannot treat it as an all-nighter. You have to eat well, sleep well, and exercise. You have to spend time with your family and friends. You also need to work in an area you’re actually passionate about—nothing else will sustain you for ten years. 132 | Everything will feel broken all the time—the diversity and magnitude of the disasters will surprise you. Your job is to fix them with a smile on your face and reassure your team that it’ll all be ok. Usually things aren’t as bad as they seem, but sometimes they are in fact really bad. In any case, just keep going. Keep growing. 133 | The CEO doesn’t get to make excuses. Lots of bad and unfair things are going to happen. But don’t let yourself say, and certainly not to the team, “if only we had more money” or “if only we had another engineer”. Either figure out a way to make that happen, or figure out what to do without it. People who let themselves make a lot of excuses usually fail in general, and startup CEOs who do it almost always fail. Let yourself feel upset at the injustice for 1 minute, and then realize that it’s up to you to figure out a solution. Strive for people to say “X just somehow always gets things done” when talking about you. 134 | No first-time founder knows what he or she is doing. To the degree you understand that, and ask for help, you’ll be better off. It’s worth the time investment to learn to become a good leader and manager. The best way to do this is to find a mentor—reading books doesn’t seem to work as well. 135 | A surprising amount of our advice at YC is of the form “just ask them” or “just do it”. First-time founders think there must be some secret for when you need something from someone or you want to do some new thing. But again, startups are where tricks stop working. Just be direct, be willing to ask for what you want, and don’t be a jerk. 136 | It’s important that you distort reality for others but not yourself. You have to convince other people that your company is primed to be the most important startup of the decade, but you yourself should be paranoid about everything that could go wrong. 137 | Be persistent. Most founders give up too quickly or move on to the next product too quickly. If things generally aren’t going well, figure out what the root cause of the problem is and make sure you address that. A huge part of being a successful startup CEO is not giving up (although you don’t want to be obstinate beyond all reason either—this is another apparent contradiction, and a hard judgment call to make.) 138 | Be optimistic. Although it’s possible that there is a great pessimistic CEO somewhere out in the world, I haven’t met him or her yet. A belief that the future will be better, and that the company will play an important role in making the future better, is important for the CEO to have and to infect the rest of the company with. This is easy in theory and hard in the practical reality of short-term challenges. Don’t lose sight of the long-term vision, and trust that the day-to-day challenges will someday be forgotten and replaced by memories of the year-to-year progress. 139 | Among your most important jobs are defining the mission and defining the values. This can feel a little hokey, but it’s worth doing early on. Whatever you set at the beginning will usually still be in force years later, and as you grow, each new person needs to first buy in and then sell others on the mission and values of the company. So write your cultural values and mission down early. 140 | Another cliché that I think is worth repeating: Building a company is somewhat like building a religion. If people don’t connect what they’re doing day-to-day with a higher purpose they care about, they will not do a great job. I think Airbnb has done the best job at this in the YC network, and I highly recommend taking a look at their cultural values. 141 | One mistake that CEOs often make is to innovate in well-trodden areas of business instead of innovating in new products and solutions. For example, many founders think that they should spend their time discovering new ways to do HR, marketing, sales, financing, PR, etc. This is nearly always bad. Do what works in the well-established areas, and focus your creative energies on the product or service you’re building. 142 | ### Part IV: Execution - **Hiring & Managing** 143 | 144 | Hiring is one of your most important jobs and the key to building a great company (as opposed to a great product.) 145 | My first piece of advice about hiring is don’t do it. The most successful companies we’ve worked with at YC have waited a relatively long time to start hiring employees. Employees are expensive. Employees add organizational complexity and communication overhead. There are things you can say to your cofounders that you cannot say with employees in the room. Employees also add inertia—it gets exponentially harder to change direction with more people on the team. Resist the urge to derive your self-worth from your number of employees. 146 | The best people have a lot of opportunities. They want to join rocketships. If you have nothing, it’s hard to hire them. Once you’re obviously winning, they’ll want to come join you. 147 | It’s worth repeating that great people have a lot of options, and you need great people to build a great company. Be generous with equity, trust, and responsibility. Be willing to go after people you don’t think you’ll be able to get. Remember that the kind of people you want to hire can start their own companies if they want. 148 | When you are in recruiting mode (i.e., from when you get product-market fit to T-infinity), you should spend about 25% of your time on it. At least one founder, usually the CEO, needs to get great at recruiting. It’s most CEOs’ number one activity by time. Everyone says that CEOs should spend a lot of their time recruiting, but in practice, none but the best do. There’s probably something to that. 149 | Don’t compromise on the quality of people you hire. Everyone knows this, and yet everyone compromises on this at some point during a desperate need. Everyone goes on to regret it, and it sometimes almost kills the company. Good and bad people are both infectious, and if you start with mediocre people, the average does not usually trend up. Companies that start off with mediocre early employees almost never recover. Trust your gut on people. If you have doubt, then the answer is no. 150 | Do not hire chronically negative people. They do not fit what an early-stage startup needs—the rest of the world will be predicting your demise every day, and the company needs to be united internally in its belief to the contrary. 151 | Value aptitute over experience for almost all roles. Look for raw intelligence and a track record of getting things done. Look for people you like – you'll be spending a lot of time together and often in tense situations. For people you don't already know, try to work on a project together before they join full-time. 152 | Invest in becoming a good manager. This is hard for most founders, and it’s definitely counterintuitive. But it’s important to get good at this. Find mentors that can help you here. If you do not get good at this, you will lose employees quickly, and if you don’t retain employees, you can be the best recruiter in the world and it still won’t matter. Most of the principles on being a good manager are well-covered, but the one that I never see discussed is “don’t go into hero mode”. Most first-time managers fall victim to this at some point and try to do everything themselves, and become unavailable to their staff. It usually ends in a meltdown. Resist all temptation to switch into this mode, and be willing to be late on projects to have a well-functioning team. 153 | Speaking of managing, try hard to have everyone in the same office. For some reason, startups always compromise on this. But nearly all of the most successful startups started off all together. I think remote work can work well for larger companies but it has not been a recipe for massive success for startups. 154 | Finally, fire quickly. Everyone knows this in principle and no one does it. But I feel I should say it anyway. Also, fire people who are toxic to the culture no matter how good they are at what they do. Culture is defined by who you hire, fire, and promote. 155 | [I wrote a blog post with more detail.](http://blog.samaltman.com/how-to-hire) 156 | ### Part IV: Execution - **Competitors** 157 | 158 | A quick word about competitors: competitors are a startup ghost story. First-time founders think they are what kill 99% of startups. But 99% of startups die from suicide, not murder. Worry instead about all of your internal problems. If you fail, it will very likely be because you failed to make a great product and/or failed to make a great company. 159 | 99% of the time, you should ignore competitors. Especially ignore them when they raise a lot of money or make a lot of noise in the press. Do not worry about a competitor until they are beating you with a real, shipped product. Press releases are easier to write than code, which is easier still than making a great product. In the words of Henry Ford: "The competitor to be feared is one who never bothers about you at all, but goes on making his own business better all the time." 160 | Every giant company has faced worse competitive threats than what you are facing now when they were small, and they all came out ok. There is always a counter-move. 161 | 162 | ### Part IV: Execution - **Making Money** 163 | Oh yes, making money. You need to figure out how to do that. 164 | The short version of this is that you have to get people to pay you more money than it costs you to deliver your good/service. For some reason, people always forget to take into account the part about how much it costs to deliver it. 165 | If you have a free product, don’t plan to grow by buying users. That’s really hard for ad-supported businesses. You need to make something people share with their friends. 166 | If you have a paid product with less than a $500 customer lifetime value (LTV), you generally cannot afford sales. Experiment with different user acquisition methods like SEO/SEM, ads, mailings, etc., but try to repay your customer acquisition cost (CAC) in 3 months. 167 | If you have a paid product with more than a $500 LTV (net to you) you generally can afford direct sales. Try selling the product yourself first to learn what works. Hacking Sales is a useful book to read. 168 | In any case, try to get to “ramen profitability”—i.e., make enough money so that the founders can live on ramen—as quickly as you can. When you get here, you control your own destiny and are no longer at the whims of investors and financial markets. 169 | Watch your cash flow obsessively. Although it sounds unbelievable, we’ve seen founders run out of money without being aware it was happening a number of times (and [read Paul Graham’s essay](http://paulgraham.com/aord.html)). 170 | ### Part IV: Execution - **Fundraising** 171 | 172 | Most startups raise money at some point. 173 | You should raise money when you need it or when it’s available on good terms. Be careful not to lose your sense of frugality or to start solving problems by throwing money at them. Not having enough money can be bad, but having too much money is almost always bad. 174 | The secret to successfully raising money is to have a good company. All of the other stuff founders do to try to over-optimize the process probably only matters about 5% of the time. Investors are looking for companies that are going to be really successful whether or not they invest, but that can grow faster with outside capital. The “really successful” part is important—because investors’ returns are dominated by the big successes, if an investor believes you have a 100% chance of creating a $10 million company but almost no chance of building a larger company, he/she will still probably not invest even at a very low valuation. Always explain why you could be a huge success. 175 | Investors are driven by the dual fears of missing the next Google, and fear of losing money on something that in retrospect looks obviously stupid. (For the best companies, they fear both at the same time.) 176 | It is a bad idea to try to raise money when your company isn’t in good enough shape to attract capital. You will burn reputation and waste time. 177 | Don’t get demoralized if you struggle to raise money. Many of the best companies have struggled with this, because the best companies so often look bad at the beginning (and they nearly always look unfashionable.) When investors tell you no, believe the no but not the reason. And remember that anything but “yes” is a “no”—investors have a wonderful ability to say “no” in a way that sounds like “maybe yes”. 178 | It’s really important to have fundraising conversations in parallel—don’t go down a list of your favorite investors sequentially. The way to get investors to act is fear of other investors taking away their opportunity. 179 | View fundraising as a necessary evil and something to get done as quickly as possible. Some founders fall in love with fundraising; this is always bad. It’s best to have just one founder do it so the company doesn’t grind to a halt. 180 | Remember that most VCs don’t know much about most industries. Metrics are always the most convincing. 181 | It’s beginning to change, but most investors (Y Combinator being a notable exception) unfortunately still require introductions from people you both know to take you seriously. 182 | Insist on clean terms (complicated terms compound and get worse each round) but don’t over-optimize, especially on valuation. Valuation is something quantitative to compete on, and so founders love to compete for the highest valuation. But intermediate valuations don’t matter much. 183 | The first check is the hardest to get, so focus your energies on getting that, which usually means focusing your attention on whoever loves you the most. Always have multiple plans, one of which is not raising anything, and be flexible depending on interest—if you can put more money to good use, and it’s available on reasonable terms, be open to taking it. 184 | An important key to being good at pitching is to make your story as clear and easy to understand as possible. Of course, the most important key is to actually have a good company. There are lots of thoughts about what to include in a pitch, but at a minimum you need to have: mission, problem, product/service, business model, team, market and market growth rate, and financials. 185 | Remember that the bar for each round of funding is much higher. If you got away with just being a compelling presenter for your seed round, don’t be surprised when it doesn’t work for your Series A. 186 | Good investors really do add a lot of value. Bad investors detract a lot. Most investors fall in the middle and neither add nor detract. Investors that only invest a small amount usually don’t do anything for you (i.e., beware party rounds). 187 | Great board members are one of the best outside forcing functions for a company other than users, and outside forcing functions are worth more than most founders think. Be willing to accept a lower valuation to get a great board member who is willing to be very involved. 188 | I think [this essay by Paul Graham](http://paulgraham.com/fr.html) is the best thing out there on fundraising. 189 | 190 | ### **Closing Thought** 191 | Remember that at least a thousand people have every great idea. One of them actually becomes successful. The difference comes down to execution. It’s a grind, and everyone wishes there were some other way to transform “idea” into “success”, but no one has figured it out yet. 192 | So all you need is a great idea, a great team, a great product, and great execution. So easy! ;) 193 | 194 | 195 | -------------------------------------------------------------------------------- /data/vc_funding.md: -------------------------------------------------------------------------------- 1 | 2 | # Is Your Company VC Fundable? 3 | 4 | By [maxine.buchert@slush.org](https://slush.org/author/maxine-buchertslush-org/ "Posts by maxine.buchert@slush.org")October 20, 2022December 2nd, 2024[No Comments](https://slush.org/is-your-company-vc-fundable/#respond) 5 | ** _Susan Hyttinen | Elmo Pakkanen_** 6 | 7 | You might have a great idea, a functioning business model, and a stellar team – maybe even budding profits – but VCs aren’t biting. It can be difficult to get a clear-cut “no” from an investor or reasoning as to why your company isn’t for them. 8 | There are many possible explanations for a lack of investor interest, but one common reason we hear from VCs is that there is a widespread fundamental misunderstanding of what is – and what isn’t – a VC fundable startup in the first place. Lack of success with VCs might have nothing to do with the “goodness” of your company or team: you might just not fit the VC model. 9 | The truth is, VC funding isn’t for everyone. In this article, we’ll walk you through how VC works and what that means for startups, what VCs consequently look for in early-stage companies – and what to do if you don’t fit the VC prototype. 10 | #### Expect to Learn: 11 | * **VC 101: Learn the rules to play the game** 12 | * **Fund math: How VCs make money** 13 | * **Besides unicorn scale, what are VCs looking for in early-stage companies?** 14 | * **People** 15 | * **Product** 16 | * **Market** 17 | * **Who doesn’t fit the VC mold?** 18 | * **Funding for alternative ventures** 19 | 20 | 21 | ## VC 101: learn the rules to play the game 22 | _TL;DR:__VCs can’t fund your company – no matter how promising – if the size of the opportunity isn’t large enough._ 23 | VC funds range from microfunds with <1M€ of assets under management to giant crossover funds with several billions. In a typical fund, the general partners (GPs) run the fund and make the ultimate call on whether to invest in your company or not, and the lion’s share of fund capital comes from limited partners (LPs). LPs are entities that manage large amounts of capital; for example insurance companies, banks, pension funds, pooled investment funds (funds-of-funds), family offices, government and academic institutions, and high-net-worth individuals. 24 | LPs invest in venture capital for many reasons, for instance the potential for high returns, supporting [revolutionary innovation](https://slush.org/entrepreneurship-redefined/revolutionary-innovation/), and asset diversification. Compared to other asset classes, venture capital is particularly non-liquid and risky – therefore the upside needs to be _significant_ to incentivize investment. 25 | General partners try to convince limited partners that their fund will return above-market returns with their investment thesis (their playbook for choosing investments) and their access to deals that match the thesis. As a founder, your company should match the VC’s investment thesis – if it doesn’t fit their focus area, the odds of investment are significantly lower. On the other hand, as a startup you want a VC who has experience in your sector, even more so when it comes to more complex deeptech solutions. 26 | ## How VCs make money: Fund math 27 | Typically a VC fund’s lifetime is 10 years, split into: 28 | * Initial investment period (2-3 years) – this is when the fund is actively looking for and making new investments. 29 | * Portfolio development period (3-5 years) – this is when the fund is focused on portfolio development and follow-on investments. 30 | * Exit period (2-3 years) – this is when the fund will pursue to exit ownership stakes in their portfolio companies. 31 | 32 | 33 | VC firms can and often do have multiple funds at the same time, so they often raise money every 3-5 years for a new fund (if they can). 34 | Intuitively it might seem like a great win if a VC fund invests a million and later sells their stake for say, twenty million. After all, they’ve returned their initial investment twenty times over. But that’s not the case: the VC model doesn’t work by making a little bit of money on a lot of deals, but instead by a few select deals bringing in extreme outlier returns. 35 | Here’s an example of a €50M fund to explain why that’s the case. Typically a fund of this size will have: 36 | * **€20M for initial investments.** For the sake of simplicity, we can assume that the fund will invest in 20 different companies over a 2-3 year period with an average ticket size of €1M per company for 10% of ownership. 37 | 38 | 39 | * **€20M for potential follow-on investments.** As portfolio companies go through subsequent funding rounds, they will issue more stock to raise more equity. Consequently, the fund’s share in the company will be diluted, and they’ll need to do follow-on investments to defend their ownership share. In practice, VCs tend to double down on the most promising companies instead of spreading their follow-on money evenly across their portfolio. 40 | 41 | 42 | * **€10M for management fees.** VC funds typically operate under a “[2 and 20](https://www.investopedia.com/terms/t/two_and_twenty.asp)” fee arrangement. “2” refers to a yearly 2% management fee that the VC fund charges their LPs for managing their money – this roughly amounts to €10M during the lifetime of the fund. The management fee is used to cover the expenses of the fund, for example salaries, an office, travel etc. “20” refers to [carried interest](https://www.holloway.com/g/venture-capital/sections/carried-interest-and-management-fees) (or ‘carry’) – a 20% cut of money VCs get beyond returning the initial investment to their investors. It is the carry in particular that makes VC so lucrative for the individuals running the fund. For the purposes of this fund math illustration, we’re not calculating the amount of carried interest here. 43 | 44 | 45 | These investments turn into returns through _successful liquidity events_ , i.e. when a startup exits through an acquisition or goes public, whether that be through an IPO, [a direct listing](https://slush.org/article/doing-things-the-spotify-way-the-road-to-direct-listing/), or a [SPAC](https://slush.org/article/ipo-spac-or-direct-listing-picking-a-path-to-the-public-market/). Yet, building a successful startup is extremely difficult – most never reach an exit. Out of the 20 example companies above, half will go bankrupt and eight will only give modest returns. That leaves two companies to return the fund and make a profit. These two outliers have to be the so-called ‘fund returners’. For example, to make 3x returns, which some would consider moderate over a decade, the fund returners would have to exit with €1.5 billion so that the fund would return €150M for its 10% stakes. 46 | What governs this process is the _power law_ , meaning that the distribution of returns in venture capital is heavily skewed – a very small percentage of startups bring a large percentage of the returns. Fund returners make up for the losses of the other investments in the portfolio. A good example of power law in action is Sequoia’s investment in Whatsapp, which was acquired for $22B by Facebook in 2014. For a Series A investment of $60M, Sequoia Capital got a $3B return. Although, this 50x return pales in comparison with Y Combinator’s [estimated](https://news.crunchbase.com/startups/y-combinator-biggest-startups-gone-public-airbnb-doordash/#:~:text=Just%2024%20hours%20before%20Airbnb,billion%20for%20a%20YC%20company.) 112,500x return on their $20,000 investment in AirBnb – exited for $2.25B. 47 | This is why the startup ecosystem is obsessed with companies valued upwards from €1B, i.e. unicorns – VCs need them to survive. Every company VCs invest in has to have (in their eyes) the potential to become a unicorn, otherwise it makes no sense for them to take a chance on them. 48 | ## Besides unicorn scale, what are VCs looking for in early-stage companies? 49 | The power law sets clear expectations for VC funding – you need to have the potential to be a unicorn that can single-handedly return the fund the VC is investing from. Making such predictions especially in the early stages is to a large extent reliant on intuition, and while there is no way to see into the future, there are certain features VCs tend to look for. 50 | Assuming you can convince an investor that the opportunity you are pursuing is large enough, you’ll also need to convince them that your company is the one to seize that opportunity. After all, every opportunity comes with opportunity cost: _what if another company simply does a better job at this?_ According to a16z’s Managing Partner Scott Kupor, [the heuristics](https://www.wired.com/story/how-early-stage-vcs-decide-where-invest/) VCs use to evaluate investment prospects generally fall into three categories: people, product, and market. 51 | ## People 52 | #### **#1: Attitude** 53 | While believing that you can set up a billion-euro business alone won’t help you build one, your likelihood of getting there is significantly lower if that isn’t your goal. Investors look for people who are somewhat delusional with their company vision, but have the skills and conviction to make that vision a reality against overwhelming odds – a perfect mixture of “smart enough to succeed, dumb enough to try”. 54 | This general mindset of ambition and resolve has been well encapsulated in the [five traits](http://www.paulgraham.com/founders.html) to look for in a founder, outlined by Y Combinator’s Paul Graham. There are of course many other [traits](https://entrepreneurshandbook.co/13-traits-that-really-matter-in-a-startups-founding-team-454fe4d4535e) that VCs look for in founders such as knowledge, EQ and IQ, integrity, adaptability, curiosity, and leadership etc. In fact, different VCs even have their own frameworks for doing [‘founder due diligence’](https://sifted.eu/articles/founder-due-diligence-vc/), so the exact concoction of these may be VC-dependent. 55 | #### **#2: Team** 56 | There’s an often romanticized concept of startup founders having a lightbulb moment and coming up with a billion dollar idea (ehm, Social Network) and then making it big. In reality, founders often have to pivot from their original idea multiple times before finding product/market fit. In the startup canon it is furthermore often emphasized that it isn’t the ideas themselves but their execution that ultimately matters – and this hinges on the early team’s ability to work together to realize their vision. 57 | Likewise to founder traits there’s a slew of desirable team attributes to choose from. Often-cited characteristics of successful teams are things like: complementary strengths and weaknesses, having a generally diverse team, and joint commitment to the company vision. 58 | **_Past experience:_** VCs will also often look at your early team’s past experiences. This means not only work experience that would give the founders readiness to tackle their chosen problem, but also achievements that showcase outstanding abilities or a ‘[hacker mentality](https://www.ycombinator.com/howtoapply/)’. 59 | **_Founder-market fit._** While there’s a lot of talk about product-market fit, in the early stages it is particularly _founder_ -market fit (FMF) that can make the difference. FMF refers to the founders having a deep understanding of their chosen market, and a high level of personal motivation bordering on obsession to solve their chosen problem in that market. Since ideas aren’t proprietary, investors have to think if there could be someone better to execute this same idea. 60 | #### **#3: Cap table** 61 | A cap table describes the ownership structure of the company – who owns how much of which class of shares. Investors generally want to see a ‘clean’ cap table. This means having a reasonable amount of shareholders in the company, and therefore a less complicated ownership structure. As a consequence, everyone on board is a clear value add, meaning no disproportionately big shares for advisors or angel investors, and no dragging along founders who’ve left the venture. A clean cap table also makes communication and voting easier. 62 | All in all, founders should use discretion when choosing who to put on the cap table. Being on the cap table means that this person has decision-making power. If an investor proves to be a suboptimal fit it can result in a lot of problems throughout the company journey. 63 | Another important feature investors look for in a cap table is that operational founders have a large enough share of ownership so that they’re properly incentivized to build the company in the long run. It’s important that while having to scrape through the tough phases of an early-stage venture founders are motivated to stay on board for the ride which can last up to ten years or more. In practice, this usually translates to founders owning at least 50% of total equity by the time the company scales. 64 | ## Product 65 | #### **#1: Rough idea or problem space** 66 | While an idea won’t make it far without a strong team executing it, it’s really difficult to build a unicorn with a bad idea. In order for a startup idea to be successful and have ‘disruptive potential,’ it is often said that the founders should either have experienced the problem themselves or have a close connection to the problem otherwise. This way the founders will be very passionate about solving it and have the grit to endure the ebbs and flows that come with being a founder. Depending on the idea, you may need to have some domain expertise. 67 | The [organic pull](http://www.paulgraham.com/organic.html) of a problem is very attractive to VCs, and a personal backstory makes for a neat narrative also in the direction of customers. This doesn’t mean that you can’t get VC funding with another kind of problem, but you might need to demonstrate a higher level of industry knowledge and be more convincing on why the problem is important – as well as why you _specifically_ are the team to solve the problem. 68 | You might also want to have some [proof of concept](https://slush.org/article/product-glossary-soaked-by-slush/#poc) to demonstrate that your idea is worth investing in. This means that you in some way illustrate that the idea is something that can actually be built; whether that be through developing a sample or testing it on the market. However, this doesn’t mean showing that there’s a market demand for your idea, which is where traction comes in. 69 | #### **#2: Early customer traction** 70 | Even at the very earliest stage, showcasing initial customer traction in conjunction with your proof of concept is a big plus. Its purpose is to demonstrate some level of validation for the idea and that you are building something that people actually need. 71 | Traction can often be measured in revenue, daily active users, or something else depending on the sector: for example for a deeptech startup this could mean having intellectual property, since a deeptech startup likely won’t have any users at the seed stage. 72 | Since traction is essentially about validation, it can also be measured through things like focus groups, industry research, and creating a waiting list, [among other things](https://jameschurch.medium.com/how-much-traction-do-you-need-to-win-over-investors-in-a-seed-round-be37280b8c67). Some early traction examples include: 73 | * Revenue run rate 74 | * Monthly recurring revenue (MRR) 75 | * Daily active users 76 | 77 | 78 | #### **#3: Growth** 79 | Similarly to traction, growth expectations are partly dependent on the vertical and sector in which your company is. For example, for B2C companies, the focus is on engagement and retention rather than recurring revenue. For B2B companies, investors only tend to care about month-on-month (MoM) growth once you’ve achieved 1M ARR, but if they do care, the desired range tends to fall between [15 to 25%](https://www.lennysnewsletter.com/p/what-is-a-good-growth-rate?r=1gicv2&s=r&utm_campaign=post&utm_medium=web). In general, success can be showcased in a more liberal selection of ways such as the number of waitlisted users or signups, unusually low customer acquisition cost (CAC), low churn, and other factors that play into growth in the longer run. 80 | ## Market 81 | #### **Total Addressable Market (TAM)** 82 | To turn grand visions into reality, there are certain physical limitations. One of the most significant ones is the size of the total addressable market (TAM), which refers to the maximum size of your opportunity. For example, Facebook’s TAM is essentially anyone who has access to the internet, whereas a dog walking application’s TAM is people who own and don’t want to walk their dogs – significantly smaller. 83 | TAM is the proxy for the size of the opportunity for VCs. Crudely put, the bigger your TAM, the bigger the potential revenue of your company. VCs need to see that you’re targeting a large enough market and have a realistic roadmap to seizing a significant share of said market to make their investment worthwhile. TAM assessments are particularly emphasized in the early stages, where company financials are nascent or non-existent. 84 | For example, Revolut’s original estimated market size was at $3B – and this was only for its initial market, the UK. Similarly, Intercom’s original market estimate was $21B. Both companies’ early pitch decks alongside others can be found [here](https://www.cbinsights.com/research/billion-dollar-startup-pitch-decks/). 85 | Companies with smaller TAMs than Facebook can of course become unicorns, and some of the best companies have started off in markets that don’t seem to show significant promise at first but grow incredibly fast. After all, Facebook itself was initially only intended for Harvard students connecting with one another. Yet even these surprising winners at some point still have to persuade VCs to bet on them – if the market isn’t showing promise yet, founders need to have a convincing story for why it eventually will. 86 | Therefore, compared to other attributes VCs look at in the early stage, TAM is a lot more binary – you either have it or you don’t, even if the justifications for TAM assessments are debatable. [Here’s more info on TAM and how you can calculate it.](https://www.productplan.com/glossary/total-addressable-market/) 87 | _Make sure to also take a look at how these attributes may be reflected on your pre-seed pitch deck_[ _here_](https://slush.org/article/how-to-ace-your-pre-seed-pitch-deck/) _._ 88 | VCs’ expectations vary from stage to stage, and the above ones primarily pertain to very early stage startups raising their first pre-seed, and seed rounds. The difference between pre-seed and seed is that pre-seed funding is used to demonstrate a _market need_ , whereas seed funding is used to prove a _market fit_. This is why Series A is often referred to as the [P/M fit round](https://slush.org/article/how-to-raise-a-series-a-a-story-by-creandum-and-seon/), as it is the first significant round of financing after P/M fit. These are the earliest, and often the smallest, rounds a company will raise. The size of a seed round will be determined by how much money the startup needs to reach its next milestone – which in turn varies between products, geography, and how competitive the market is among other things. 89 | Overall, founders should also be mindful of the implications of accepting VC funding: once you’ve shared equity you’re no longer the only person you’re building for. You’ve entered a contract to build a company that needs to exit as a unicorn within the decade or sooner. This means that your exclusive focus should be on growth to justify a potential exit. 90 | ## Who doesn’t fit the VC mold? 91 | There are clearly many companies that don’t fit the bill. The VC model works most optimally for startups with specific traits, namely those that are massively scalable and capital efficient, can go through rapid iteration cycles, and have recurring revenues and low marginal costs – which often best describes software startups. 92 | The types of startups that are less likely to attract VC funding by these metrics are deeptech, service businesses, and hardware companies**.** While they may have scalable businesses, their timeline for scaling isn’t necessarily one that aligns with the VC cycle, their TAM may be too niche/small, or they rely heavily on labor in scaling. Where these lag, even a brilliant team and idea aren’t necessarily enough to tip the scales in favor of VCs funding your company. 93 | Another category of companies that aren’t necessarily VC fundable are those that formally meet the characteristics of a startup but don’t aim for high enough or fast enough growth. For example, a SaaS business valued at €50M may offer massive financial returns for the founders, but a negligent one for VCs looking for €1B fund returners. 94 | Naturally, there are also companies in these categories that have been VC funded, such as deeptech startups like SpaceX, Lilium, Darktrace, and DeepMind, and hardware ones like Nothing, Nest, and Beats. And of course it’s good to note that tech giants like Microsoft, Google, and Intel were also deeptech startups in their origins and/or had a hardware component like Apple. So while getting to the top may be riskier, take a longer return on investment for VCs, and require more capital, the upside is also potentially enormous. 95 | Many VC funds may indeed be industry agnostic – but due to the system overall they are less likely to invest in, for example, deeptech that can often take longer to commercialize than a software application. Even so, within the VC system there are certain funds that focus specifically on funding research-based innovations and hardware such as [Fifty Years](https://techcrunch.com/2021/10/27/fifty-years-a-deep-tech-investor-has-raised-90-million-from-tens-of-unicorn-founders/), [Lux Capital](https://luxcapital.com/), and [Obvious Ventures](https://slush.org/article/decarb-decade-is-the-obvious-ventures-path-by-andrew-beebe/). In Europe, some prominent funds in this sphere are [Pale Blue Dot](https://paleblue.vc/), [Speedinvest](https://www.speedinvest.com/), and [Norrsken](https://www.norrsken.vc/), among [others](https://sifted.eu/articles/active-deeptech-vc-investors-2021/). 96 | ## What VCs look for in deeptech and hardware startups in the early stages 97 | #### – Steven Jacobs, Venture Partner and CPO at Lakestar 98 | ##### Early indicators 99 | “When investing in deeptech companies, in addition to looking for exceptional founders, we evaluate the viability, readiness, applicability, and scale that the new technology has. This tells us if the technology can work, when it will likely be working by, what problems it can be used to solve, and how big the market is for solving those problems.” 100 | ##### Growth and progress expectations 101 | “Deeptech companies typically do not have smooth growth trajectories like with traditional enterprise SaaS. The timelines are often determined more by the technology readiness level and the nature of the the product than they are by simply being a novel deep technology. For example, timelines for a new hardware sensor may be longer than a new ML algorithm due to supply chain, tooling, and shipping limitations. What is important is that the company is able to demonstrate significant reductions in risk, both technical and commercial, in sufficient increments that they will be able to advocate for more funding to continue making progress.” 102 | ##### Assessing the scale of the opportunity 103 | “To evaluate the scale of the market opportunity, we look for where the novel deep technology intersects different markets. CRISPR for example can be used for therapeutics, fuels, materials, etc. We then look at the size of those markets and their rate of growth and determine what the TAM would be for this new technology. Obviously, any startup should be hyper-focused on one initial use case where they can be 10x better, but we also want to know that the company can scale broadly and capture significant value that is commensurate with the risk and capital requirements to develop the new deep technology. We consider the initial use case a wedge into what should be a very broad and large market.” 104 | ##### Red flags 105 | “Timing and compound risk are typically the two biggest concerns with deeptech investments. For many deep technologies the question is when and not if. If there isn’t a strong rationale as to why the timing to $100M+ revenue is within venture time horizons (5-10 years) then that is a red flag. Similarly, if the market isn’t large and obvious for a new technology, you have the compound risk of: #1 the technology working and #2 there being PMF when the technology works. Typically we want to see a very large, obvious, broad market for a new deep technology to be applied into (vis-a-vis a product) when it is ready. Venture capital is not designed to fund open ended research. Government grants are better for that. Venture capital is designed to fund risky business endeavors where the reward is at least three orders of magnitude larger than the risk. It is hard to close that math when there is significant compound risk.” 106 | ## Funding for alternative ventures 107 | Companies that aren’t VC fundable do have other forms of funding – in fact, most startups aren’t even VC funded. For example these funding sources are available: 108 | **Angel investors** 109 | Angel investors are individual investors that can provide early-stage capital to startups. Angel investors range from hobbyists investing small tickets to professionals participating in large VC rounds. The difference to VC funds is that angels are investing their own money – and are thus responsible to themselves. An angel investor might be happy with a smaller return on an investment as they’ll get 100% of the money returned. 110 | **Bootstrapping** 111 | Bootstrapping means financing the company through other means than outside investments. Often this means that the founders have other sources of income, financing through debt, sweat equity, or financing through company revenues. Some famous bootstrapped companies include MailChimp and Spanx, and for example [Infobip](https://slush.org/article/9-bootstrapping-lessons-for-early-stage-founders/) and [Celonis](https://slush.org/article/6-lessons-for-founders-from-a-european-decacorn/) in Europe. 112 | **CVCs** 113 | Corporate Venture Capital or corporate venturing refers to the venture arms of big corporations, such as [Google Ventures](https://www.gv.com/) and [Intel Capital](https://www.intelcapital.com/). The investments can either be done through a minor stake similar to a regular VC, or through a venture client model where the startup and corporation in question work on joint projects funded by the CVC. 114 | **Family offices** 115 | Family offices help wealthy families invest their money to protect their family wealth. Alongside VC and angel capital, family offices are one of the three most common funding sources in the early stages. They work much like VC firms, but without the intermediary LP-GP relationship. 116 | **Friends and family** 117 | Where founders’ own savings aren’t enough they might resort to friends and family rounds, where the investor is a founder’s personal connection. These usually take place at the same stages as angel rounds, but are on average smaller than the latter. 118 | **Government grants and subsidies** 119 | Many governments have subsidy and grant programs to support startup founders, most often in the early stages of building. 120 | **Revenue-based financing** 121 | Somewhat of a hybrid between equity financing and debt financing, revenue-based financing works through the startup in question paying off a loan they take from an investor with generated revenue. 122 | **Venture debt** 123 | Venture debt is loan financing designed to provide funding between equity rounds. Its purpose can be to act as a bridge, but can also for example provide funding for acquisitions. 124 | ## Fitting the VC mold 125 | The VC model certainly favors some types of companies over others, which means you might have to target your efforts better and highlight slightly different things to gain funding. On the other hand, if these goals, timeline, and scale seem unreasonable, VC funding’s probably not your cup of tea. If that’s the case, you should consider alternative paths for building your company. It’s important to know what VCs are looking for to not waste your time on a wild goose chase. 126 | 127 | -------------------------------------------------------------------------------- /docker/agent.py: -------------------------------------------------------------------------------- 1 | from llama_index.llms.groq import Groq 2 | from llama_index.core.agent import ReActAgent, ReActChatFormatter 3 | from llama_index.core.llms import ChatMessage 4 | from tools import vanilla_rag_tool, hyde_rag_tool, multistep_rag_tool, evaluate_context_tool, evaluate_response_tool 5 | 6 | f = open("/run/secrets/groq_key") 7 | GROQ_API_KEY = f.read() 8 | f.close() 9 | 10 | system_header = """ 11 | You are a Query Agent, whose main task is to produce reliable information in response to the prompt from the user. You should do so by retrieving the information and evaluating it, using the available tools. Your expertise is in useful startup resources, with a focus on building pitch decks. In particular, your workflow should look like this: 12 | 0. If the question from the user does not concern useful startup resources you should dismiss the user question from the beginning, telling you can't reply to that and that they should prompt you with a question about your expertise. 13 | 1. Choose a tool for contextual information retrieval based on the user's query: 14 | - If the query is simple and specific, ask for the 'query_vanilla_rag' tool 15 | - If the query is general and vague, ask for the 'query_hyde_rag' tool 16 | - If the query is complex and involves searching for nested information, ask for the 'query_multistep_rag' tool 17 | 2. Once the information retrieval tool returned you with a context, you should evaluate the relevancy of the context provided using the 'evaluate_context' tool. This tool will tell you how relevant is the context in light of the original user prompt, which you will have to pass to the tool as argument, as well as the context from the Query Engine tool. 18 | 2a. If the retrieved context is not relevant, go back to step (1), choose a different Query Engine tool and try with that. If, after trying with all Query Engine tools, the context is still not relevant, tell the user that you do not have enough information to answer the question 19 | 2b. If the retrieved context is relevant, proceed with step (3) 20 | 3. Produce a potential answer to the user prompt and evaluate it with the 'evaluate_response' tool, passing the original user's prompt, the context and your candidate answer to the tool. In this step, you MUST use the 'evaluate_response' tool. You will receive an evaluation for faithfulness and relevancy. 21 | 3a. If the response lacks faithfulness and relevancy, you should go back to step (3) and produce a new answer 22 | 3b. If the response is faithful and relevant, proceed to step (5) 23 | 4. Return the final answer to the user. 24 | """ 25 | 26 | llm = Groq(model="qwen-qwq-32b", api_key=GROQ_API_KEY) 27 | 28 | agent = ReActAgent.from_tools( 29 | tools = [vanilla_rag_tool, hyde_rag_tool, multistep_rag_tool, evaluate_context_tool, evaluate_response_tool], 30 | verbose = True, 31 | chat_history=[ChatMessage.from_str(content=system_header, role="system")], 32 | max_iterations=20, 33 | ) 34 | -------------------------------------------------------------------------------- /docker/main.py: -------------------------------------------------------------------------------- 1 | from fastapi import FastAPI 2 | from fastapi.responses import ORJSONResponse 3 | from pydantic import BaseModel 4 | from agent import agent 5 | 6 | class UserInput(BaseModel): 7 | prompt: str 8 | 9 | class ApiOutput(BaseModel): 10 | response: str 11 | 12 | app = FastAPI(default_response_class=ORJSONResponse) 13 | 14 | @app.post("/chat") 15 | async def chat(inpt: UserInput): 16 | response = await agent.achat(message=inpt.prompt) 17 | response = str(response) 18 | return ApiOutput(response=response) -------------------------------------------------------------------------------- /docker/tools.py: -------------------------------------------------------------------------------- 1 | from llama_index.core.indices.query.query_transform.base import HyDEQueryTransform, StepDecomposeQueryTransform 2 | from llama_index.core.query_engine import TransformQueryEngine, MultiStepQueryEngine 3 | from llama_index.core.tools import FunctionTool 4 | from llama_index.core import Settings 5 | from llama_index.core.evaluation import RelevancyEvaluator, FaithfulnessEvaluator 6 | from llama_index.core.llms import ChatMessage 7 | from pydantic import BaseModel, Field 8 | from llama_index.llms.groq import Groq 9 | import json 10 | from qdrant_client import QdrantClient, AsyncQdrantClient 11 | from llama_index.vector_stores.qdrant import QdrantVectorStore 12 | from llama_index.core import SimpleDirectoryReader, VectorStoreIndex, StorageContext 13 | from llama_index.core.node_parser import SemanticSplitterNodeParser 14 | from llama_index.embeddings.huggingface import HuggingFaceEmbedding 15 | 16 | 17 | f = open("/run/secrets/groq_key") 18 | GROQ_API_KEY = f.read() 19 | f.close() 20 | 21 | embed_model = HuggingFaceEmbedding(model_name="Alibaba-NLP/gte-modernbert-base") 22 | node_parser = SemanticSplitterNodeParser(embed_model=embed_model) 23 | 24 | qc = QdrantClient(host="qdrant", port=6333) 25 | aqc = AsyncQdrantClient(host="qdrant", port=6333) 26 | 27 | if not qc.collection_exists("data"): 28 | docs = SimpleDirectoryReader(input_dir="/app/data/").load_data() 29 | nodes = node_parser.get_nodes_from_documents(docs, show_progress=True) 30 | vector_store = QdrantVectorStore("data", qc, aclient=aqc, enable_hybrid=True, fastembed_sparse_model="Qdrant/bm25") 31 | storage_context = StorageContext.from_defaults(vector_store=vector_store) 32 | index = VectorStoreIndex(nodes=nodes, embed_model=embed_model, storage_context=storage_context) 33 | index1 = VectorStoreIndex.from_vector_store(vector_store=vector_store, embed_model=embed_model) 34 | index2 = VectorStoreIndex.from_vector_store(vector_store=vector_store, embed_model=embed_model) 35 | else: 36 | vector_store = QdrantVectorStore("data", client=qc, aclient=aqc, enable_hybrid=True, fastembed_sparse_model="Qdrant/bm25") 37 | index = VectorStoreIndex.from_vector_store(vector_store=vector_store, embed_model=embed_model) 38 | index1 = VectorStoreIndex.from_vector_store(vector_store=vector_store, embed_model=embed_model) 39 | index2 = VectorStoreIndex.from_vector_store(vector_store=vector_store, embed_model=embed_model) 40 | 41 | class EvaluateContext(BaseModel): 42 | context_is_ok: int = Field(description="Is the context relevant to the question? Give a score between 0 and 100") 43 | reasons: str = Field(description="Explanations for the given evaluation") 44 | 45 | llm = Groq(model="llama-3.3-70b-versatile", api_key=GROQ_API_KEY) 46 | llm_eval = llm.as_structured_llm(EvaluateContext) 47 | Settings.llm = llm 48 | faith_eval = FaithfulnessEvaluator() 49 | rel_eval = RelevancyEvaluator() 50 | 51 | query_engine = index.as_query_engine(llm=llm) 52 | query_engine1 = index1.as_query_engine(llm=llm) 53 | query_engine2 = index2.as_query_engine(llm=llm) 54 | 55 | hyde = HyDEQueryTransform(llm=llm, include_original=True) 56 | hyde_query_engine = TransformQueryEngine(query_engine=query_engine1, query_transform=hyde) 57 | 58 | step_decompose_transform = StepDecomposeQueryTransform(llm, verbose=True) 59 | multistep_query_engine = MultiStepQueryEngine(query_engine=query_engine2, query_transform=step_decompose_transform) 60 | 61 | 62 | async def vanilla_query_engine_tool(query: str): 63 | """This tool is useful for retrieving directly information from a vector database without any prior query transformation. It is mainly useful when the query is simple but specific""" 64 | response = await query_engine.aquery(query) 65 | return response.response 66 | 67 | async def hyde_query_engine_tool(query: str): 68 | """This tool is useful for retrieving information from a vector database with the transformation of a query into an hypothetical document embedding, which will be used for retrieval. It is mainly useful when the query is general or vague.""" 69 | response = await hyde_query_engine.aquery(query) 70 | return response.response 71 | 72 | async def multi_step_query_engine_tool(query: str): 73 | """This tool is useful for retrieving information from a vector database with the decomposition of the query into a series of queries that will be iteratively executed against the vector database for retrieval. It is mainly useful when the query is complex and asks for nested and multi-faceted information.""" 74 | response = await multistep_query_engine.aquery(query) 75 | return response.response 76 | 77 | 78 | async def evaluate_context(original_prompt: str = Field(description="Original prompt provided by the user"), context: str = Field(description="Contextual information, either from retrieved documents")) -> str: 79 | """ 80 | Useful for evaluating the coherence and relevance of retrieved contextual information in light of the user's prompt. 81 | 82 | This tool takes the original user prompt and contextual information as input, and evaluates the coherence of the response with the original prompt and the relevance of the contextual information. It returns a formatted string with the evaluation scores and reasons for the evaluations. 83 | 84 | Args: 85 | original_prompt (str): Original prompt provided by the user. 86 | context (str): Contextual information from retrieved documents. 87 | """ 88 | messages = [ChatMessage.from_str(content=original_prompt, role="user"), ChatMessage.from_str(content=f"Here is some context that I found that might be useful for replying to the user:\n\n{context}", role="assistant"), ChatMessage.from_str(content="Can you please evaluate the relevance of the contextual information (giving it a score between 0 and 100) in light or my original prompt? You should also tell me the reasons for your evaluations.", role="user")] 89 | response = await llm_eval.achat(messages) 90 | json_response = json.loads(response.message.blocks[0].text) 91 | final_response = f"The context provided for the user's prompt is {json_response['context_is_ok']}% relevant.\nThese are the reasons why you are given these evaluations:\n{json_response['reasons']}" 92 | return final_response 93 | 94 | async def evaluate_response(original_prompt: str = Field(description="Original prompt provided by the user"), context: str = Field(description="Contextual information, either from retrieved documents"), answer: str = Field(description="Final answer to the original prompt")) -> str: 95 | """ 96 | Useful for evaluating the faithfulness and relevance of a response to a given prompt using contextual information. 97 | 98 | This tool takes an original prompt, contextual information, and a final answer, and evaluates the coherence of the response with the original prompt and the relevance of the contextual information. It returns a formatted string with the evaluation scores. 99 | 100 | Args: 101 | original_prompt (str): Original prompt provided by the user. 102 | context (str): Contextual information, either from retrieved documents or from the web, or both. 103 | answer (str): Final answer to the original prompt. 104 | """ 105 | faithfulness = await faith_eval.aevaluate(query=original_prompt, response=answer, contexts=[context]) 106 | relevancy = await rel_eval.aevaluate(query=original_prompt, response=answer, contexts=[context]) 107 | rel_score = relevancy.score if relevancy.score is not None else 0 108 | fai_score = faithfulness.score if faithfulness.score is not None else 0 109 | return f"The relevancy of the produced answer is {rel_score*100}% and the faithfulness is {fai_score*100}%" 110 | 111 | 112 | vanilla_rag_tool = FunctionTool.from_defaults( 113 | fn=vanilla_query_engine_tool, 114 | name="query_vanilla_rag", 115 | description="""This tool is useful for retrieving directly information from a vector database without any prior query transformation. It is mainly useful when the query is simple but specific 116 | 117 | Args: 118 | query (str): Query to search the vector database""" 119 | ) 120 | 121 | hyde_rag_tool = FunctionTool.from_defaults( 122 | fn=hyde_query_engine_tool, 123 | name="query_hyde_rag", 124 | description="""This tool is useful for retrieving information from a vector database with the transformation of a query into an hypothetical document embedding, which will be used for retrieval. It is mainly useful when the query is general or vague. 125 | 126 | Args: 127 | query (str): Query to search the vector database""" 128 | ) 129 | 130 | 131 | multistep_rag_tool = FunctionTool.from_defaults( 132 | fn=multi_step_query_engine_tool, 133 | name="query_multistep_rag", 134 | description="""This tool is useful for retrieving information from a vector database with the decomposition of the query into a series of queries that will be iteratively executed against the vector database for retrieval. It is mainly useful when the query is complex and asks for nested and multi-faceted information. 135 | 136 | Args: 137 | query (str): Query to search the vector database""" 138 | ) 139 | 140 | evaluate_response_tool = FunctionTool.from_defaults( 141 | fn=evaluate_response, 142 | name="evaluate_response", 143 | description=""" 144 | Useful for evaluating the faithfulness and relevance of a response to a given prompt using contextual information. 145 | 146 | This tool takes an original prompt, contextual information, and a final answer, and evaluates the coherence of the response with the original prompt and the relevance of the contextual information. It returns a formatted string with the evaluation scores. 147 | 148 | Args: 149 | original_prompt (str): Original prompt provided by the user. 150 | context (str): Contextual information, either from retrieved documents or from the web, or both. 151 | answer (str): Final answer to the original prompt. 152 | """ 153 | ) 154 | 155 | evaluate_context_tool = FunctionTool.from_defaults( 156 | fn=evaluate_context, 157 | name="evaluate_context", 158 | description=""" 159 | Useful for evaluating the coherence and relevance of retrieved contextual information in light of the user's prompt. 160 | 161 | This tool takes the original user prompt and contextual information as input, and evaluates the coherence of the response with the original prompt and the relevance of the contextual information. It returns a formatted string with the evaluation scores and reasons for the evaluations. 162 | 163 | Args: 164 | original_prompt (str): Original prompt provided by the user. 165 | context (str): Contextual information from retrieved documents. 166 | """ 167 | ) -------------------------------------------------------------------------------- /environment.yml: -------------------------------------------------------------------------------- 1 | name: ragcoon 2 | channels: 3 | - conda-forge 4 | dependencies: 5 | - _libgcc_mutex=0.1=conda_forge 6 | - _openmp_mutex=4.5=2_gnu 7 | - bzip2=1.0.8=h4bc722e_7 8 | - ca-certificates=2025.1.31=hbcca054_0 9 | - ld_impl_linux-64=2.43=h712a8e2_4 10 | - libexpat=2.6.4=h5888daf_0 11 | - libffi=3.4.6=h2dba641_0 12 | - libgcc=14.2.0=h767d61c_2 13 | - libgcc-ng=14.2.0=h69a702a_2 14 | - libgomp=14.2.0=h767d61c_2 15 | - liblzma=5.6.4=hb9d3cd8_0 16 | - libnsl=2.0.1=hd590300_0 17 | - libsqlite=3.49.1=hee588c1_1 18 | - libuuid=2.38.1=h0b41bf4_0 19 | - libxcrypt=4.4.36=hd590300_1 20 | - libzlib=1.3.1=hb9d3cd8_2 21 | - ncurses=6.5=h2d0b736_3 22 | - openssl=3.4.1=h7b32b05_0 23 | - pip=25.0.1=pyh8b19718_0 24 | - python=3.11.11=h9e4cc4f_2_cpython 25 | - readline=8.2=h8c095d6_2 26 | - setuptools=75.8.2=pyhff2d567_0 27 | - tk=8.6.13=noxft_h4845f30_101 28 | - wheel=0.45.1=pyhd8ed1ab_1 29 | - pip: 30 | - absl-py==2.1.0 31 | - aiohappyeyeballs==2.5.0 32 | - aiohttp==3.11.13 33 | - aiosignal==1.3.2 34 | - annotated-types==0.7.0 35 | - anyio==4.8.0 36 | - attrs==25.1.0 37 | - beautifulsoup4==4.13.3 38 | - blinker==1.9.0 39 | - certifi==2025.1.31 40 | - charset-normalizer==3.4.1 41 | - click==8.1.8 42 | - coloredlogs==15.0.1 43 | - dataclasses-json==0.6.7 44 | - deepdiff==6.7.1 45 | - deprecated==1.2.18 46 | - dirtyjson==1.0.8 47 | - distro==1.9.0 48 | - dnspython==2.7.0 49 | - email-validator==2.2.0 50 | - fastapi==0.115.11 51 | - fastapi-cli==0.0.7 52 | - fastembed==0.6.0 53 | - filelock==3.17.0 54 | - filetype==1.2.0 55 | - flask==3.1.0 56 | - flatbuffers==25.2.10 57 | - frozenlist==1.5.0 58 | - fsspec==2025.3.0 59 | - greenlet==3.1.1 60 | - grpcio==1.71.0 61 | - grpcio-tools==1.71.0 62 | - gunicorn==23.0.0 63 | - h11==0.14.0 64 | - h2==4.2.0 65 | - hpack==4.1.0 66 | - httpcore==1.0.7 67 | - httptools==0.6.4 68 | - httpx==0.28.1 69 | - huggingface-hub==0.29.2 70 | - humanfriendly==10.0 71 | - hyperframe==6.1.0 72 | - idna==3.10 73 | - itsdangerous==2.2.0 74 | - jinja2==3.1.6 75 | - jiter==0.8.2 76 | - joblib==1.4.2 77 | - llama-cloud==0.1.14 78 | - llama-cloud-services==0.6.5 79 | - llama-index==0.12.23 80 | - llama-index-agent-openai==0.4.6 81 | - llama-index-cli==0.4.1 82 | - llama-index-core==0.12.23.post2 83 | - llama-index-embeddings-huggingface==0.5.2 84 | - llama-index-embeddings-openai==0.3.1 85 | - llama-index-indices-managed-llama-cloud==0.6.8 86 | - llama-index-llms-groq==0.3.1 87 | - llama-index-llms-openai==0.3.25 88 | - llama-index-llms-openai-like==0.3.4 89 | - llama-index-multi-modal-llms-openai==0.4.3 90 | - llama-index-program-openai==0.3.1 91 | - llama-index-question-gen-openai==0.3.0 92 | - llama-index-readers-file==0.4.6 93 | - llama-index-readers-llama-parse==0.4.0 94 | - llama-index-vector-stores-qdrant==0.4.3 95 | - llama-parse==0.6.4.post1 96 | - loguru==0.7.3 97 | - markdown-it-py==3.0.0 98 | - markupsafe==3.0.2 99 | - marshmallow==3.26.1 100 | - mdurl==0.1.2 101 | - mesop==0.14.1 102 | - mmh3==5.1.0 103 | - mpmath==1.3.0 104 | - msgpack==1.1.0 105 | - multidict==6.1.0 106 | - mypy-extensions==1.0.0 107 | - nest-asyncio==1.6.0 108 | - networkx==3.4.2 109 | - nltk==3.9.1 110 | - numpy==2.2.3 111 | - nvidia-cublas-cu12==12.4.5.8 112 | - nvidia-cuda-cupti-cu12==12.4.127 113 | - nvidia-cuda-nvrtc-cu12==12.4.127 114 | - nvidia-cuda-runtime-cu12==12.4.127 115 | - nvidia-cudnn-cu12==9.1.0.70 116 | - nvidia-cufft-cu12==11.2.1.3 117 | - nvidia-curand-cu12==10.3.5.147 118 | - nvidia-cusolver-cu12==11.6.1.9 119 | - nvidia-cusparse-cu12==12.3.1.170 120 | - nvidia-cusparselt-cu12==0.6.2 121 | - nvidia-nccl-cu12==2.21.5 122 | - nvidia-nvjitlink-cu12==12.4.127 123 | - nvidia-nvtx-cu12==12.4.127 124 | - onnxruntime==1.21.0 125 | - openai==1.65.5 126 | - ordered-set==4.1.0 127 | - orjson==3.10.15 128 | - packaging==24.2 129 | - pandas==2.2.3 130 | - pillow==11.1.0 131 | - portalocker==2.10.1 132 | - propcache==0.3.0 133 | - protobuf==5.29.3 134 | - py-rust-stemmers==0.1.5 135 | - pydantic==2.10.6 136 | - pydantic-core==2.27.2 137 | - pygments==2.19.1 138 | - pypdf==5.3.1 139 | - python-dateutil==2.9.0.post0 140 | - python-dotenv==1.0.1 141 | - python-multipart==0.0.20 142 | - pytz==2025.1 143 | - pyyaml==6.0.2 144 | - qdrant-client==1.13.3 145 | - regex==2024.11.6 146 | - requests==2.32.3 147 | - rich==13.9.4 148 | - rich-toolkit==0.13.2 149 | - safetensors==0.5.3 150 | - scikit-learn==1.6.1 151 | - scipy==1.15.2 152 | - sentence-transformers==3.4.1 153 | - shellingham==1.5.4 154 | - six==1.17.0 155 | - sniffio==1.3.1 156 | - soupsieve==2.6 157 | - sqlalchemy==2.0.38 158 | - starlette==0.46.1 159 | - striprtf==0.0.26 160 | - sympy==1.13.1 161 | - tenacity==9.0.0 162 | - threadpoolctl==3.5.0 163 | - tiktoken==0.9.0 164 | - tokenizers==0.21.0 165 | - torch==2.6.0 166 | - tqdm==4.67.1 167 | - transformers==4.49.0 168 | - triton==3.2.0 169 | - typer==0.15.2 170 | - typing-extensions==4.12.2 171 | - typing-inspect==0.9.0 172 | - tzdata==2025.1 173 | - urllib3==2.3.0 174 | - uv==0.6.5 175 | - uvicorn==0.34.0 176 | - uvloop==0.21.0 177 | - watchdog==6.0.0 178 | - watchfiles==1.0.4 179 | - websockets==15.0.1 180 | - werkzeug==3.1.3 181 | - wrapt==1.17.2 182 | - yarl==1.18.3 183 | -------------------------------------------------------------------------------- /frontend/Dockerfile: -------------------------------------------------------------------------------- 1 | FROM python:3.11.9-slim-bookworm 2 | 3 | WORKDIR /frontend/ 4 | COPY ./ /frontend/ 5 | 6 | RUN pip cache purge 7 | RUN pip install --no-cache-dir -r requirements.txt 8 | 9 | EXPOSE 8001 10 | 11 | CMD ["gunicorn", "--bind", "0.0.0.0:8001", "frontend:me"] -------------------------------------------------------------------------------- /frontend/frontend.py: -------------------------------------------------------------------------------- 1 | import mesop as me 2 | import mesop.labs as mel 3 | from pydantic import BaseModel 4 | import requests as rq 5 | 6 | class UserInput(BaseModel): 7 | prompt: str 8 | 9 | def on_load(e: me.LoadEvent): 10 | me.set_theme_mode("system") 11 | 12 | @me.page( 13 | security_policy=me.SecurityPolicy( 14 | allowed_iframe_parents=["https://google.github.io", "https://huggingface.co"] 15 | ), 16 | path="/", 17 | title="RAGcoon - The resources you need for your Startup", 18 | on_load=on_load, 19 | ) 20 | def page(): 21 | mel.chat(transform, title="RAGcoon - The resources you need for your Startup", bot_user="RAGcoon") 22 | 23 | 24 | def transform(input: str, history: list[mel.ChatMessage]): 25 | try: 26 | response = rq.post("http://localhost:8000/chat", json=UserInput(prompt=input).model_dump()) 27 | except Exception as e: 28 | response = rq.post("http://backend:8000/chat", json=UserInput(prompt=input).model_dump()) 29 | res = response.json()["response"] 30 | yield res -------------------------------------------------------------------------------- /frontend/requirements.txt: -------------------------------------------------------------------------------- 1 | mesop==0.14.1 2 | pydantic==2.10.6 3 | requests==2.32.3 4 | gunicorn==23.0.0 -------------------------------------------------------------------------------- /logo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/AstraBert/ragcoon/e8dee742e2153949811519daba61031db21a97af/logo.png -------------------------------------------------------------------------------- /scripts/agent.py: -------------------------------------------------------------------------------- 1 | from llama_index.llms.groq import Groq 2 | from llama_index.core.agent import ReActAgent, ReActChatFormatter 3 | from llama_index.core.llms import ChatMessage 4 | from dotenv import load_dotenv 5 | from tools import vanilla_rag_tool, hyde_rag_tool, multistep_rag_tool, evaluate_context_tool, evaluate_response_tool 6 | from os import environ as ENV 7 | 8 | load_dotenv() 9 | 10 | system_header = """ 11 | You are a Query Agent, whose main task is to produce reliable information in response to the prompt from the user. You should do so by retrieving the information and evaluating it, using the available tools. Your expertise is in useful startup resources, with a focus on building pitch decks. In particular, your workflow should look like this: 12 | 0. If the question from the user does not concern useful startup resources you should dismiss the user question from the beginning, telling you can't reply to that and that they should prompt you with a question about your expertise. 13 | 1. Choose a tool for contextual information retrieval based on the user's query: 14 | - If the query is simple and specific, ask for the 'query_vanilla_rag' tool 15 | - If the query is general and vague, ask for the 'query_hyde_rag' tool 16 | - If the query is complex and involves searching for nested information, ask for the 'query_multistep_rag' tool 17 | 2. Once the information retrieval tool returned you with a context, you should evaluate the relevancy of the context provided using the 'evaluate_context' tool. This tool will tell you how relevant is the context in light of the original user prompt, which you will have to pass to the tool as argument, as well as the context from the Query Engine tool. 18 | 2a. If the retrieved context is not relevant, go back to step (1), choose a different Query Engine tool and try with that. If, after trying with all Query Engine tools, the context is still not relevant, tell the user that you do not have enough information to answer the question 19 | 2b. If the retrieved context is relevant, proceed with step (3) 20 | 3. Produce a potential answer to the user prompt and evaluate it with the 'evaluate_response' tool, passing the original user's prompt, the context and your candidate answer to the tool. In this step, you MUST use the 'evaluate_response' tool. You will receive an evaluation for faithfulness and relevancy. 21 | 3a. If the response lacks faithfulness and relevancy, you should go back to step (3) and produce a new answer 22 | 3b. If the response is faithful and relevant, proceed to step (4) 23 | 4. Return the final answer to the user. 24 | """ 25 | 26 | llm = Groq(model="qwen-qwq-32b", api_key=ENV["GROQ_API_KEY"]) 27 | 28 | agent = ReActAgent.from_tools( 29 | tools = [vanilla_rag_tool, hyde_rag_tool, multistep_rag_tool, evaluate_context_tool, evaluate_response_tool], 30 | verbose = True, 31 | chat_history=[ChatMessage.from_str(content=system_header, role="system")], 32 | max_iterations=20, 33 | ) 34 | 35 | 36 | -------------------------------------------------------------------------------- /scripts/main.py: -------------------------------------------------------------------------------- 1 | from fastapi import FastAPI 2 | from fastapi.responses import ORJSONResponse 3 | from pydantic import BaseModel 4 | from agent import agent 5 | 6 | class UserInput(BaseModel): 7 | prompt: str 8 | 9 | class ApiOutput(BaseModel): 10 | response: str 11 | 12 | app = FastAPI(default_response_class=ORJSONResponse) 13 | 14 | @app.post("/chat") 15 | async def chat(inpt: UserInput): 16 | response = await agent.achat(message=inpt.prompt) 17 | response = str(response) 18 | return ApiOutput(response=response) -------------------------------------------------------------------------------- /scripts/tools.py: -------------------------------------------------------------------------------- 1 | from llama_index.core.indices.query.query_transform.base import HyDEQueryTransform, StepDecomposeQueryTransform 2 | from llama_index.core.query_engine import TransformQueryEngine, MultiStepQueryEngine 3 | from llama_index.core.tools import FunctionTool 4 | from llama_index.core import Settings 5 | from llama_index.core.evaluation import RelevancyEvaluator, FaithfulnessEvaluator 6 | from llama_index.core.llms import ChatMessage 7 | from pydantic import BaseModel, Field 8 | from llama_index.llms.groq import Groq 9 | from dotenv import load_dotenv 10 | from os import environ as ENV 11 | import json 12 | from qdrant_client import QdrantClient, AsyncQdrantClient 13 | from llama_index.vector_stores.qdrant import QdrantVectorStore 14 | from llama_index.core import SimpleDirectoryReader, VectorStoreIndex, StorageContext 15 | from llama_index.core.node_parser import SemanticSplitterNodeParser 16 | from llama_index.embeddings.huggingface import HuggingFaceEmbedding 17 | 18 | load_dotenv() 19 | 20 | embed_model = HuggingFaceEmbedding(model_name="Alibaba-NLP/gte-modernbert-base") 21 | node_parser = SemanticSplitterNodeParser(embed_model=embed_model) 22 | 23 | qc = QdrantClient("http://localhost:6333") 24 | aqc = AsyncQdrantClient("http://localhost:6333") 25 | 26 | if not qc.collection_exists("data"): 27 | docs = SimpleDirectoryReader(input_dir="../data/").load_data() 28 | nodes = node_parser.get_nodes_from_documents(docs, show_progress=True) 29 | vector_store = QdrantVectorStore("data", qc, aclient=aqc, enable_hybrid=True, fastembed_sparse_model="Qdrant/bm25") 30 | storage_context = StorageContext.from_defaults(vector_store=vector_store) 31 | index = VectorStoreIndex(nodes=nodes, embed_model=embed_model, storage_context=storage_context) 32 | index1 = VectorStoreIndex.from_vector_store(vector_store=vector_store, embed_model=embed_model) 33 | index2 = VectorStoreIndex.from_vector_store(vector_store=vector_store, embed_model=embed_model) 34 | else: 35 | vector_store = QdrantVectorStore("data", client=qc, aclient=aqc, enable_hybrid=True, fastembed_sparse_model="Qdrant/bm25") 36 | index = VectorStoreIndex.from_vector_store(vector_store=vector_store, embed_model=embed_model) 37 | index1 = VectorStoreIndex.from_vector_store(vector_store=vector_store, embed_model=embed_model) 38 | index2 = VectorStoreIndex.from_vector_store(vector_store=vector_store, embed_model=embed_model) 39 | 40 | class EvaluateContext(BaseModel): 41 | context_is_ok: int = Field(description="Is the context relevant to the question? Give a score between 0 and 100") 42 | reasons: str = Field(description="Explanations for the given evaluation") 43 | 44 | llm = Groq(model="llama-3.3-70b-versatile", api_key=ENV["GROQ_API_KEY"]) 45 | llm_eval = llm.as_structured_llm(EvaluateContext) 46 | Settings.llm = llm 47 | faith_eval = FaithfulnessEvaluator() 48 | rel_eval = RelevancyEvaluator() 49 | 50 | query_engine = index.as_query_engine(llm=llm) 51 | query_engine1 = index1.as_query_engine(llm=llm) 52 | query_engine2 = index2.as_query_engine(llm=llm) 53 | 54 | hyde = HyDEQueryTransform(llm=llm, include_original=True) 55 | hyde_query_engine = TransformQueryEngine(query_engine=query_engine1, query_transform=hyde) 56 | 57 | step_decompose_transform = StepDecomposeQueryTransform(llm, verbose=True) 58 | multistep_query_engine = MultiStepQueryEngine(query_engine=query_engine2, query_transform=step_decompose_transform) 59 | 60 | 61 | async def vanilla_query_engine_tool(query: str): 62 | """This tool is useful for retrieving directly information from a vector database without any prior query transformation. It is mainly useful when the query is simple but specific""" 63 | response = await query_engine.aquery(query) 64 | return response.response 65 | 66 | async def hyde_query_engine_tool(query: str): 67 | """This tool is useful for retrieving information from a vector database with the transformation of a query into an hypothetical document embedding, which will be used for retrieval. It is mainly useful when the query is general or vague.""" 68 | response = await hyde_query_engine.aquery(query) 69 | return response.response 70 | 71 | async def multi_step_query_engine_tool(query: str): 72 | """This tool is useful for retrieving information from a vector database with the decomposition of the query into a series of queries that will be iteratively executed against the vector database for retrieval. It is mainly useful when the query is complex and asks for nested and multi-faceted information.""" 73 | response = await multistep_query_engine.aquery(query) 74 | return response.response 75 | 76 | 77 | async def evaluate_context(original_prompt: str = Field(description="Original prompt provided by the user"), context: str = Field(description="Contextual information, either from retrieved documents")) -> str: 78 | """ 79 | Useful for evaluating the coherence and relevance of retrieved contextual information in light of the user's prompt. 80 | 81 | This tool takes the original user prompt and contextual information as input, and evaluates the coherence of the response with the original prompt and the relevance of the contextual information. It returns a formatted string with the evaluation scores and reasons for the evaluations. 82 | 83 | Args: 84 | original_prompt (str): Original prompt provided by the user. 85 | context (str): Contextual information from retrieved documents. 86 | """ 87 | messages = [ChatMessage.from_str(content=original_prompt, role="user"), ChatMessage.from_str(content=f"Here is some context that I found that might be useful for replying to the user:\n\n{context}", role="assistant"), ChatMessage.from_str(content="Can you please evaluate the relevance of the contextual information (giving it a score between 0 and 100) in light or my original prompt? You should also tell me the reasons for your evaluations.", role="user")] 88 | response = await llm_eval.achat(messages) 89 | json_response = json.loads(response.message.blocks[0].text) 90 | final_response = f"The context provided for the user's prompt is {json_response['context_is_ok']}% relevant.\nThese are the reasons why you are given these evaluations:\n{json_response['reasons']}" 91 | return final_response 92 | 93 | async def evaluate_response(original_prompt: str = Field(description="Original prompt provided by the user"), context: str = Field(description="Contextual information, either from retrieved documents"), answer: str = Field(description="Final answer to the original prompt")) -> str: 94 | """ 95 | Useful for evaluating the faithfulness and relevance of a response to a given prompt using contextual information. 96 | 97 | This tool takes an original prompt, contextual information, and a final answer, and evaluates the coherence of the response with the original prompt and the relevance of the contextual information. It returns a formatted string with the evaluation scores. 98 | 99 | Args: 100 | original_prompt (str): Original prompt provided by the user. 101 | context (str): Contextual information, either from retrieved documents or from the web, or both. 102 | answer (str): Final answer to the original prompt. 103 | """ 104 | faithfulness = await faith_eval.aevaluate(query=original_prompt, response=answer, contexts=[context]) 105 | relevancy = await rel_eval.aevaluate(query=original_prompt, response=answer, contexts=[context]) 106 | rel_score = relevancy.score if relevancy.score is not None else 0 107 | fai_score = faithfulness.score if faithfulness.score is not None else 0 108 | return f"The relevancy of the produced answer is {rel_score*100}% and the faithfulness is {fai_score*100}%" 109 | 110 | 111 | vanilla_rag_tool = FunctionTool.from_defaults( 112 | fn=vanilla_query_engine_tool, 113 | name="query_vanilla_rag", 114 | description="""This tool is useful for retrieving directly information from a vector database without any prior query transformation. It is mainly useful when the query is simple but specific 115 | 116 | Args: 117 | query (str): Query to search the vector database""" 118 | ) 119 | 120 | hyde_rag_tool = FunctionTool.from_defaults( 121 | fn=hyde_query_engine_tool, 122 | name="query_hyde_rag", 123 | description="""This tool is useful for retrieving information from a vector database with the transformation of a query into an hypothetical document embedding, which will be used for retrieval. It is mainly useful when the query is general or vague. 124 | 125 | Args: 126 | query (str): Query to search the vector database""" 127 | ) 128 | 129 | 130 | multistep_rag_tool = FunctionTool.from_defaults( 131 | fn=multi_step_query_engine_tool, 132 | name="query_multistep_rag", 133 | description="""This tool is useful for retrieving information from a vector database with the decomposition of the query into a series of queries that will be iteratively executed against the vector database for retrieval. It is mainly useful when the query is complex and asks for nested and multi-faceted information. 134 | 135 | Args: 136 | query (str): Query to search the vector database""" 137 | ) 138 | 139 | evaluate_response_tool = FunctionTool.from_defaults( 140 | fn=evaluate_response, 141 | name="evaluate_response", 142 | description=""" 143 | Useful for evaluating the faithfulness and relevance of a response to a given prompt using contextual information. 144 | 145 | This tool takes an original prompt, contextual information, and a final answer, and evaluates the coherence of the response with the original prompt and the relevance of the contextual information. It returns a formatted string with the evaluation scores. 146 | 147 | Args: 148 | original_prompt (str): Original prompt provided by the user. 149 | context (str): Contextual information, either from retrieved documents or from the web, or both. 150 | answer (str): Final answer to the original prompt. 151 | """ 152 | ) 153 | 154 | evaluate_context_tool = FunctionTool.from_defaults( 155 | fn=evaluate_context, 156 | name="evaluate_context", 157 | description=""" 158 | Useful for evaluating the coherence and relevance of retrieved contextual information in light of the user's prompt. 159 | 160 | This tool takes the original user prompt and contextual information as input, and evaluates the coherence of the response with the original prompt and the relevance of the contextual information. It returns a formatted string with the evaluation scores and reasons for the evaluations. 161 | 162 | Args: 163 | original_prompt (str): Original prompt provided by the user. 164 | context (str): Contextual information from retrieved documents. 165 | """ 166 | ) -------------------------------------------------------------------------------- /setup.ps1: -------------------------------------------------------------------------------- 1 | docker compose up qdrant -d 2 | conda env create -f environment.yml 3 | conda activate ragcoon -------------------------------------------------------------------------------- /setup.sh: -------------------------------------------------------------------------------- 1 | docker compose up qdrant -d 2 | conda env create -f environment.yml 3 | conda activate ragcoon -------------------------------------------------------------------------------- /shell/conda_env.sh: -------------------------------------------------------------------------------- 1 | eval "$(conda shell.bash hook)" 2 | 3 | conda env create -f /app/environment.yml -------------------------------------------------------------------------------- /shell/run.sh: -------------------------------------------------------------------------------- 1 | eval "$(conda shell.bash hook)" 2 | 3 | conda activate ragcoon 4 | cd /app/ 5 | uvicorn main:app --host 0.0.0.0 --port 8000 6 | conda deactivate 7 | -------------------------------------------------------------------------------- /start_services.ps1: -------------------------------------------------------------------------------- 1 | docker compose up qdrant -d 2 | docker compose up frontend -d 3 | docker compose up backend -d -------------------------------------------------------------------------------- /start_services.sh: -------------------------------------------------------------------------------- 1 | docker compose up qdrant -d 2 | docker compose up frontend -d 3 | docker compose up backend -d -------------------------------------------------------------------------------- /workflow.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/AstraBert/ragcoon/e8dee742e2153949811519daba61031db21a97af/workflow.png --------------------------------------------------------------------------------