What’s one of the main causes of startup failure? – The wrong team.
Explainable AI study
There are two main components that drive startups: the product and the people creating it. In general, it’s easier to estimate the startup’s success depending on the objective facts of the product (the price, target audience, etc.). What is so much harder is to assess how the characteristics of the people behind the product influence the chances of survival. Hence, that’s exactly what we are trying to tackle!
Our AI-based model aims to predict a startup’s success based on the psychological profiles of the founders and employees. We gained thorough knowledge in this field while closely working with startups on their HR-related issues and conducted the Startup Anatomy study to find out what are the success factors for founders. This knowledge combined with our explainable AI model is going to be one of the biggest milestones for the following years, and we cannot wait for you to see the full results!
The journey has just started, but we already earned an award with our research paper explaining the methodology – Best Service Innovation Paper Award – ISM 2020!
Do you want to nerd it out and discover more about the possible correlation between AI and founder profiling?
Download and read the paper!
And if you want to make sure you are amongst the first ones being notified when the whole AI model is developed, then subscribe to our newsletter and follow us on our socials!
More about the study
The problem
It’s not just proven in the literature, but our Startup Anatomy study also revealed significant correlations between certain personality traits and the startup’s financial success and the chance of survival. Entrepreneurs could fail despite having financial means, good ideas or outstanding qualifications, if they don’t possess the necessary personality traits. The problem is that this phenomenon is too dependent on personality and we cannot thoroughly examine its predictive effect with the tools available to us today. That’s why we think it’s crucial to develop such a tool!
The solution
To address the prediction problem, we are using a hybrid intelligence model able to process multidimensional data. Errors are captured and explained so that a second model can predict the error. The new model is not interpretable: this is called deep learning – even the model creator doesn’t know how the model comes to a decision, but the result is a prediction.
The goal is to train the model on different data sets, to constantly improve it. The approach is called BAPC (referring to ‘Before and After prediction Parameter Comparison’, developed at SCCH), since the AI correction (meant as the difference between the previous result and the current one) is the main player in the formula.
Applying the model to predict the success of startups requires precise indicators: as in the Newsvendor Model, one of the first can be norm orientation. Success factor, as in the Startup Anatomy study, is measured as headcount growth. The discussion about the indicators is still open, and we are working to refine those in order to deliver the best!