Who asked: Marta Lindqvist, HR director at Saltmarsh Logistics.
Why: seven of 24 people on last year's staff list have left, and Marta needs to know where to focus retention work before the next hiring round.
The data: employees, last year's staff list: department, tenure, salary, average weekly overtime, last performance rating (two missing) and whether they have left since.
What done looks like: attrition by department and by overtime band, the rating gap between leavers and stayers, a test on tenure, a logistic model of leaving on overtime and tenure, and a labeled chart. Counts sit next to every rate, because the groups are small.
People analytics questions are small-data questions: 24 employees, 7 leavers. Every rate rests on a handful of people, so report counts next to rates, test before you claim a difference, and say what the data can't show. This project also introduces logistic regression, glm(..., family = binomial), the standard model for a yes or no outcome such as "left within a year". Each step adds to the same script.