A Case Study of Software Project Replacement: A Time Series Analysis

Abstract

Enterprise software requires constant updates to keep it usable. These updates originate in correcting errors and mainly in new organizational demands. Over time, these demands generate a significant workload that becomes increasingly complex than the first requirements. For this reason, the organization providing the software may choose to continue updating the old product or make it obsolete and replace it with a new one. Identifying the ideal moment to carry out this migration involves, in addition to the costs of keeping the product obsolete for a while, the effort to develop a new one. This work addresses a case study that comprises fifteen years with two migrations of the software project. Due to the availability of the collection of activities performed by the development and support team, performed sequentially over time, the applicability of time series was possible. Furthermore, the historical base of the activities performed made it possible to use the time series decomposition to obtain its trend, seasonality and noise. Time series decomposition indicated many random events in the first migration, while in the second, the team self-regulated, but there were tension points. This study identified a preliminary model whose purpose is to determine when to develop a new software version.