Using a Predictive Rating System for Computer Programmers to Optimise Recruitment: Using Ratings to Optimise Programmer Recruitment

Paul J. Bracewell (DOT loves data, Wellington, New Zealand), Ankit K. Patel (DOT loves data, Wellington, New Zealand), Evan J. Blackie (DOT loves data, Wellington, New Zealand), and Chris Boys (Umano, Sydney, Australia)
Copyright: © 2017 |Pages: 14
EISBN13: 9781522553472|DOI: 10.4018/JCIT.2017070101
OnDemand PDF Download:
OnDemand PDF Download
Download link provided immediately after order completion


Using a quantitative assessment system, the number of resumes reviewed to identify a suitable developer was reduced to 3.5% with a successful recruitment decision made in 10 working days of posting the job advertisement. This paper summarises the methodology for developing that rating system. The depth and quality of an available talent pool is a function of demand, which is demonstrated by comparing globally-scaled individual performance metrics. Public code repositories are accessed and the code quality assessed algorithmically. The performance score combines accuracy, timeliness and difficulty from a series of challenges. These three attributes form a meaningful predictive measure of performance by using a non-linear optimisation routine. Bootstrapping is used to validate the approach. This process randomly omitted a scored performance observation per coder in order to calculate the performance score from the retained scores. There was a strong relationship (r = 0.70) between the predicted 1-omitted-performance score with the actual omitted score highlighting the predictive power.
InfoSci-OnDemand Powered Search