Data-Driven Strategies for Project Pricing and Resource Allocation in Software Development

B13.ai are a software development company specialising in developing bespoke solutions to clients’ needs. Key to estimating the cost of a software development project is forecasting the time required to deliver the project and address defects/improvements. We perform a data-driven analysis of this task, investigating the accuracy of B13.ai’s current (typically, linear) rubrics. We: • Perform predictive modelling to investigate hours worked by developers vs. other roles. We show that a linear model is sometimes adequate but often inadequate. • Use PCA to understand the roles which most influence total hours, finding that these are the Dev, Senior Dev, and QC roles. • Perform a life-cycle analysis to elucidate phases of projects and the involvements for each role in each phase. • Study the time spent on defects/improvements vs. development, finding that the former has a different skew and kurtosis to the latter, so the times cannot be linearly translated.