Learning from Internal Mobility : A Machine Learning Approach to Augmenting Human Capital

In companies today, work is organized in fluid teams whose composition changes as certain (new) skills become important while others become less relevant. Employees in-turn are primarily responsible for navigating their careers through positions that match their expertise and ambition. As retention and advancement of good employees is vital, many organizations provide employees the opportunity to apply for new jobs listed in the organization. Facilitating and supporting this internal mobility is a novel matching problem faced by many firms. We partnered with the IT division of a large financial services firm to develop an algorithmic model of an (internal) applicant’s probability of selection to a new position. We show that the informational content in the applicant’s current job’s description and the firm’s past selection decisions contain valuable signal to predict selection of internal applicants. The model enables the firm to identify cases where the applying employee’s chances of selection was predictably low. This allows the firm to design training programs that are tailored to improve employees’ selection prospects to their observed choices among new positions. We verify that these training opportunities are consistent with the firm’s incentives and summarize them to an interpretable set via clustering. Our machine learning approach integrates three different sources of information available to firms, to identify training opportunities directly linked to new positions and observed employee choices