Startup Success Prediction in the Dutch Startup Ecosystem

Abstract

What makes a startup successful? How to define success? Existent models on startup success prediction often left aside significant predictors which may not be available in typical business databases such as crunchbase. com. In this research, focused on a population of five thousand organisations from the Dutch startup ecosystem, we go beyond previous approaches and deliver predictive models that include novel and distinctive variables. To achieve this goal, we depart from an extensive selection of variables drawn from the literature review. The initial selection is discussed, refined and enriched by carrying out interviews with knowledgeable actors in the ecosystem. At the end of the study, a total of eight significant predictors are used to construct three predictive models on startup success. The first model predicts a startup having total funding of one million euros or above, the second model predicts a startup having ten or more employees, and the third model predicts a startup having an average annualized return of at least 20% in the past three years. After testing the models, accuracies of 71%, 71% and 76% respectively are obtained. The results of this research are meant to be used by the organisation techleap.nl. By enriching the data, employing more sophisticated MLmodels and conducting this research at different points of time, techleap.nl will be capable of monitoring and predicting the performance of the ecosystemboth accurately and dynamically.