Predictive People Analytics

Predictive people analytics applies statistical and machine learning models to workforce data to forecast outcomes before they happen: attrition risk, promotion readiness, absenteeism, hiring demand, or capacity gaps. Where descriptive analytics reports what happened last quarter, predictive analytics estimates what is likely to happen next and which factors are driving it.

A typical attrition model is trained on historical exits using supervised learning, with features such as manager change, engagement score, tenure, pay position, and promotion delay. Feature-importance methods (SHAP is the common one) show which drivers matter for a given prediction, which is what makes the output actionable rather than a black-box score. The limitation of prediction alone is that a probability still needs a person to notice it and decide what to do. The direction the field is moving is from a score to a signal that arrives with a recommended response. Darwinbox provides attrition prediction with driver-level explanations in its analytics layer and a natural-language Analytics Agent for drill-down; Cortex is designed to extend this by grouping related signals into moments with a proposed playbook.