A web app that helps Kickstarter creators increase the chances of reaching their campaign's funding goal by analyzing the project description, its structure and content, using data science.
If you've recently launched a Kickstarter, add the URL to your project in the text box in the landing page. If you're still drafting your campaign, the preview URL that Kickstarter generates will work just fine. After you submit your project link, you'll receive the following insights:
- Probability that your project will reach its funding goal
- How to improve your chances by tweaking 6 structural components in your project's description
- How to improve your chances by adding specific language or topics that are highly predictive of successfully funded projects
I hope you enjoy using CampaignCritic as much as I enjoyed building it!
When you submit a link to your Kickstarter project, CampaignCritic:
- Analyzes the structure and content of your project's description using natural language processing
- Feeds the analysis into a machine learning model that has learned from having combed through over 24,000 past Kickstarters
- Reports the probability of reaching your funding goal based on the structure and content of your project's description
- Computes weighted scores reflecting how effectively your campaign utilizes 6 structural components that were found to be most predictive of funded projects
- Compares your project's scores against the top 5% of projects from the past five years
Note: CampaignCritic is only relevant for Kickstarters in the USA and doesn't take into account a project's aesthetics, feasiblility, usefulness, market demand or media buzz—just some aspects of a project's description that are relatively easy for creators to tweak. The model's performance, as measured by precision, is 0.71 +/- 0.01. For those still curious about how I built CampaignCritic, check out these slides.
- I want to help Kickstarter creators maximize their chances of reaching their funding goal. Too many innovative projects are crippled by a lackluster project description. Not everyone is a seasoned campaign designer.
- I'm a big fan of the crowdfunding world—it enables everyday folks to become entrepreneurs and even gives rise to a revolutionary new genre of products.
- I love natural language processing and want to use it to build tools that enable machines to decode and interpret the most unstructured data of all—human language. By doing so, perhaps we'll someday teach a machine to help design Kickstarter campaigns, or even write business plans for new companies.
- Jupyter notebooks, fully documented and written in Python, that illustrate every aspect of my building this project. Each notebook can be run to generate the data and content needed for the subsquent notebook.
- Jupyter notebooks, fully documented and written in Python, used to research and prototype (found in the prototyping folder) the following:
- Web app based on Bootstrap (front-end) and Flask (back-end), containing the trained model, scaler, and
$n$ -gram vectorizer (found in the application folder). - Two Python modules used by the Jupyter notebooks in this repo:
-
feature_engineering.pycontains all functions used for engineering structural (meta) features -
prediction_results.pycontains all functions used for constructing the suggestions bar graph
-
I designed CampaignCritic to help Kickstarter creators struggling with writing a project description, as well as creators who just want to maximize their chances. However, my hope is that by using CampaignCritic, creators will develop better project descriptions that are more convincing, thus enabling a richer crowdfunding scene and perhaps more frequent innovations from everyday folks.
Given more time, I'd love to incorporate the funding goal, the pledge structure, image quality and text content inside images in the feature space to improve the model's performance and provide additional knobs that creator's could realistically turn to tweak their project's description and further increase their chances.