Predicting Startup Survival from Digital Traces: Towards a Procedure for Early Stage Investors

We investigate whether digital traces can be used to predict early stage startup survival. Based on common survival factors from the entrepreneurship literature, we mined the digital footprints of 542 entrepreneurs and their ventures. Using a context-specific text mining approach, we performed a bootstrapping simulation in which we predict 5-year survival for different survival rates that range from 50% to 10%. Our results indicate that we can predict 5-year survival with an accuracy of up to 91%. With this study, we will provide an evidence-based taxonomy of digital traces for predicting early stage startup survival, identify the most important digital traces for doing so and benchmark our predictive approach against the actual investments of 339 business angels.