Andrea Contigiani, Management, The Wharton School
Abstract: By allowing ventures to obtain market feedback during the product development process, experimentation confers a learning advantage. However, we know little about its costs. In particular, experimenting with an early-stage idea requires some degree of disclosure, leading to imitation risk. This paper examines the tension between learning and imitation in the experimentation process of entrepreneurial ventures, seeking to understand the conditions under which early experimentation is optimal. I propose a theoretical framework to explain the relationship between uncertainty, appropriability, and experimentation. Learning is effective when the primary source of uncertainty is commercial rather than technological. The appropriability regime has an ambiguous effect: it raises the value of experimentation if it minimizes imitation risk or lowers it if it increases infringement risk. I test this framework using a novel dataset of US-based software ventures. To measure experimentation, I exploit the practice of beta testing, a process common among in software and generally mentioned in the news. I analyze a large body of online news to identify the timing of experimentation, combine it with data on market, team, product, IP, and financing, and build a dataset of 1700 ventures. Using a vector-based distance measure across product features to proxy market uncertainty and a 2014 US Supreme Court decision as exogenous shock to appropriability, I implement a triple-difference specification to estimate the impact of market uncertainty and appropriability on experimentation behavior. Preliminary results suggest that ventures under high market uncertainty and low appropriability tend to experiment more, providing evidence for the arguments of learning effectiveness and risk of infringement.