We used Kickstarter data to predict crowdfunding success based on project parameters. The tool is now live here – and below is a short description of what you can do with it. Further, this is just one example of what will follow when you combine machine learning and database into a predictive database.

One of the challenges when introducing the predictive database as a new database category, is understanding the use cases for it. Which problem can a predictive database solve, why should I use it in my company? Well, all companies do projects and – as we will soon demonstrate – any project’s success can be predicted with a simple query IF you have enough data points (parameters).

When preparing our materials to be presented at Slush, the world’s leading startup & tech event, we wanted to prove our statement with live value, with real data. And optimally, to show something that would be a topical use case for the primary audience: early-stage startups looking for funding.

Example: using Kickstarter data to predict crowdfunding success

We dug into Kickstarter projects, to see if the Aito predictive database can be used to predict and understand what contributes to the success of a fundraiser campaign. After all, how cool would it be to predict if your lifelong dream of building a bike made of recycled materials would get the funding it needs?!

Step 1: simple as Google search

You’ll find the demo via this link. You start by searching for an existing Kickstarter project which relates to yours. Just type in some keywords, as you would do for any Google search.

Assuming you really want to move forward with the bike-idea, let’s go for that. Searching “recycled bike” gives you an overview of projects with a relevancy score that’s based on the name, description, category and locations. The list can be sorted any way you like, using the column names. Clicking the “show Aito search query” will show you what Aito query was used to get the results.

Searching a relevant project from past Kickstarter projects

Step 2: deep dive into the project

Select the most relevant project to see more details (click the green button SELECT, clicking the name of the project will bring you to the original Kickstarter funding page).

On the left, you’ll get an overview of the project's parameters that are used as a basis for the prediction. The parameters are for example description, category, goal funding, launch date and so on.

On the right, you’ll instantly see the prediction of the probability of this particular project reaching its funding goal. You can see returned results from the predictive database when you click the “Show Aito Returned Result”.

Overview of the project's parameters that are used as a basis for prediction

The predictive database is also able to give you an explanation of the returned results.

Scrolling further down, you’ll find a visualization of the features that are most impactful (biggest crosses) and even the ones that have the biggest negative impact. Hovering over the crosses will give you more info on how the probability has been measured. Clicking “Show Aito feature explanation query” will show you the query used to get this explanation.

Visualization of the features that are most impactful to the prediction result

Step 3: Starting the fun part

You can start changing the parameters on the right and see how those will influence the probability rate. Every time you change a parameter, a predict query is sent instantly to the predictive database. You’ll get an instant prediction on how likely your crowdfunding campaign will reach its goal by the set deadline - and an explanation of why.

You can also scroll down to see how similar projects perform and click on them to analyze why Aito thinks they succeed or not.

Overview of how similar projects perform

And, to highlight, in our first sub-heading “using Kickstarter data to predict crowdfunding success” you can replace the “Kickstarter” with “your” and “crowdfunding success” with “project success”. You can use your data to predict project success – you just need to use our predictive database to do the math.

Watch the 2 minute video on Youtube where we go through the demo step-by-step as well.