Track 15 · Applied Statistics for Professionals · Lesson 5 of 12
Employee turnover and retention
Calculate turnover and churn rates with the right denominator, and analyse tenure data honestly when many people have not yet left (censoring).
An HR director asks: “Are we losing more of last year's graduate intake than the year before?” The newer intake has simply had less time to leave, so a raw count of leavers favours it. The same trap appears in customer churn, and the same tools — rates per unit of exposure and retention at fixed tenures — fix it.
Employee turnover and customer churn are the same statistical problem: people enter (are hired, sign up), stay for a while, and some leave (resign, cancel). In this lesson we use the synthetic subscriptions dataset — 2 000 customers of a fictional streaming service who joined during 2023 and were followed to 31 December 2024 — as a stand-in for an HR tenure file. Every method transfers directly.
Included with a Formula Quest membership
It has 8 graded exercises with feedback, hints and worked solutions. After it you will be able to:
- Calculate annual turnover with the average headcount and state the denominator
- Compute leaving rates per person-month of exposure
- Recognise censored records and explain why ignoring them biases tenure and cohort comparisons
- Compare cohorts at a fixed tenure, or with a life-table retention estimate
- Describe differences in turnover between groups as associations unless they come from an experiment