Social Network Topology as a Predictor of Startup Success
Sean Wise, Michael Mihalicz, Marc SmithStartup Founders use social networks to make up for resource scarcity, that is, they have access to the resources of others through their networks. Startup accelerators help nascent startups grow by providing funding and resources. Hochberg (2016) defined a startup accelerator as a fixed-term, cohort-based boot camp for startups offering educational and mentorship programs for startup founders, exposing them to a wide variety of mentors. In this same paper, Hochberg explains that startup accelerators add value to startups by providing seed capital, coworking space, mentorship and by creating a community with strong peer effects. It is on this last benefit that we focus. As accelerators grow and proliferate, the hunt for tangible cohort peer effects has been gaining pace. We use social network theory to explore these peer effects and argue that an accelerator cohort’s social network topology can be useful in predicting startup success.
MIT’s Peter Gloor and his co-authors released a series of papers (2010-2014) detailing the impact of an entrepreneur’s social networking behavior. In 2010, an article entitled “The Power of Alumni Networks” found the success of startup companies to correlate with the online social network structure of its founders. A few years later, Gloor and a new set of colleagues found that centrality in the network predicts entrepreneurial success — the more central actors are in the different types of networks, the more successful their firms. Based on these findings, we expect the highest performing startups in an accelerator cohort to be central in their cohort’s social network and the network itself dense in intra-cohort ties.
Using a sample of 923 startups across 155 cohorts between 2008 and 2018, we map the intra-cohort social network activity over the course of the Techstars accelerator program using NodeXL and the corpus of cohort activity on Twitter. Each cohort’s social network topology is recorded with nodes representing startups in a cohort and vertices representing intra-cohort Twitter activity. Cohort performance is measured as the likelihood of each cohort to generate a Top 50 Startup — the 50 highest performing startups based on valuation. Based on previous work showing social network topology to be a predictor of success across teams with the same goals (Wise, 2014), we hypothesize that cohorts successfully generating a Top 50 Startup share common topological elements that visually differentiate these cohorts from cohorts failing to generate successful startups. In addition, we hypothesize that the Top 50 Startups will occupy more central positions in the network (i.e. high in-betweenness centrality) and that the top performing cohorts will be denser than their underperforming cousins.
We use social network theory, peer effect theory and prior research to explain our findings and describe how accelerator participants generate peer effects by supporting each other over Twitter through likes and retweets. This research has implications on the approach that accelerators and their management teams adopt to better ensure startup success and can also help us better understand and measure the elusive peer effects driving startup success at accelerators.