Prof. Gaia Rubera (Bocconi University)

Leider musste der Termin im Juni 2019 kurzfristig geändert werden. Wir bitten um Verständnis.

Programm

Datum: 28. Juni 2019

Zeit: 10.15 - 11.45 Uhr

Ort: Professor-Huber-Platz 2, V005

Titel: Blessed from Birth? Using Twitter Data to Predict Startup Success

Abstract

Startups are vital for the economic growth of a nation. They complement large firms in innovation, give birth to new industries, and create new employment opportunities. Albeit relevant, startups have a high failure rate: approximately, only 35% of startups make a profit and most startups fail. Given the importance of picking up the right startups, venture capitalists and governmental associations have developed sophisticated business practices to try and predict the success of a startup. However, predicting startup success remains such a daunting task that Bill Maris, managing partner of Google Ventures, compared venture investing to “buying lottery tickets”. In this research, we put forth a novel prediction: “Startups whose founders have a higher mindset similarity with the potential customers are more successful than other startups.” The above prediction has never made explicit or tested, even though the innovation, marketing, and entrepreneurship literatures all pinpoint to it. While theoretically grounded in many streams of research, two obstacles have prevented the use of mindset similarity between founders and potential customers as a key predictor of startup success. First, it is hard to identify the potential customers of a startup in its early stage. Second, even if we could identify potential customers, it is very difficult to measure the mindset of hundred of thousands of potential customers with traditional techniques. We solve these two challenges by taking advantage of Twitter data and latest advancements in the Natural Language Processing field. We test our hypothesis in a large dataset of 743 startups and find a positive correlation between our measures of mindset similarity and startup performance 3 years after its birth. Thus, we propose a novel algorithm to predict the success of a startup, while still in its early stage. The findings of this research could be highly beneficial for a variety of stakeholders, including governments, policy makers, banks, venture capitalists, non-governmental organizations, and society at large. Better predictions would lead to a more efficient allocation of funds to those startups that can really grow and drive the economy.