Context dependent maintenance effort estimation: Case study in a small software company

Abstract

This paper presents an approach to software maintenance effort estimation based on the analysis of software maintenance requests available in issue tracking system in a very small local software company. Over 1900 requests, collected during 19 months in 2010 and 2011, were selected for analysis. The approach takes into account: the frequency of submitted user requests for particular periods of time important for organizing work in the company, the existence of maintenance service agreement for clients, the number of programmers assigned to the task associated to each request, and the number of working hours required for requests solving. The approach calculates correlation and establishes linear regression between frequencies of submitted user requests and the average time required for completing requests. This approach estimates an average number of working hours for processing user requests in periods of time that are important for internal company organization. In the paper are also discussed implications for practice and research, and provided recommendations for adapting this approach to other small software companies.