We introduce a novel approach to measure the founding strategic differentiation of startups and its relationship to follow-on performance. We use natural language processing and historical websites to estimate the similarity between the founding website of an individual startup, the historical website of public firms at the startup’s founding year, and the founding website of other startups founded in the same year. We propose that distance in the value proposition stated in these websites represents differentiation in the market. Startup differentiation is estimated as the average text-based distance from the five closest incumbents (public firms). We implement this approach using a large sample of startups from Crunchbase. Our measure predicts a meaningful increase in early-stage financing and equity outcomes, unconditionally and controlling for cohort and industry fixed effects. The positive benefits of equity outcomes only evidence themselves after year 6 of age, suggesting more differentiated firms may take longer to prove themselves. Using out-of-sample tests, we also demonstrate that our measure is economically important, predicting 30% of the total variation in the receipt of early-stage financing and 20% of variation in equity outcomes. Public datasets of our differentiation score and scraped website data are provided, together with open-source code to replicate our approach in other settings.
This paper was accepted by Joshua Gans, business strategy.
Supplemental Material:
The data files and online appendices are available at
https://doi.org/10.1287/mnsc.2022.4369
....

Measuring Founding Strategy

10.1287/mnsc.2022.4369