Attrition risk

Attrition risk is a forward-looking, individual-level score estimating the probability that a specific employee or customer will leave within a defined window, typically built with predictive analytics techniques trained on historical departures and the signals that preceded them.

For employees, common inputs include tenure, time since last promotion or raise, engagement survey scores, manager changes and compensation relative to peers; for customers, inputs mirror those used in a customer health score, such as usage decline and support escalations. This differs from employee turnover rate or churn rate, which describe what already happened in aggregate, whereas attrition risk scores individuals before they leave, so action can be taken while there is still time to intervene.

Attrition risk models are used to prioritize retention conversations, whether that means a manager check-in for a flagged employee or a customer-success outreach for a flagged account, and their value depends entirely on whether the underlying model is well calibrated to the organization's own history. A common pitfall is treating a risk score as a verdict rather than a probability, or intervening in a way, like a surprise retention bonus, that itself signals something is wrong and accelerates the departure it was meant to prevent, which can also depress employee engagement among peers who hear about it.