Learning from Failure: Big Data Analysis for Detecting the Patterns of Failure in Innovative Startups

Abstract

This article aims at identifying appropriate models for analyzing large datasets to serve a twofold goal: first, to better understand the dynamics impacting innovative startups' performance and their managerial practice and, second, to detect their patterns of failure. Therefore, we investigate the interaction of economic–financial, context, and governance dimensions of 4185 Italian innovative startups created from 2012 to 2015. Once startups have been grouped, we focus only on those that are unsuccessful. Then, failure patterns have been uncovered, integrating the use of factor and cluster analysis, where factor scores for each firm are used to identify a set of homogeneous groups based on clustering methods. The integrated use of those large-dimensional data techniques permits to classify items in rigorous ways and to unfold structures of the data, which are not apparent in the beginning. The analysis suggests that each pattern of failure is a multidimensional construct and, as a consequence can generate different managerial implications. Therefore, an effective handling of failure requires management to use appropriate interventions targeted at the challenges faced by that particular pattern of failure in the age of different firms.