Start by defining the population and the window precisely, then split leavers into voluntary and involuntary, and voluntary further into regretted and not. Cut by tenure band, team, manager, location and hiring source. Read cohorts rather than a monthly headline rate, and check exit-interview reasons against what the same people said months earlier.
Nothing else is meaningful until this is written down. Decide who counts: permanent staff only, or contractors and interns as well; whether an internal transfer is a leaver from the old team; whether someone who resigned in one period and left in another is counted at resignation or at exit. Decide the denominator too, since opening headcount, closing headcount and an average across the window produce very different figures in a growing organisation. Then fix the window and keep it fixed. Arguments about an attrition figure are usually arguments about definitions, and they are settled once, in writing, so the same query can be re-run later without a fresh debate.
They carry opposite meanings, and mixing them produces a number that cannot be acted on. A performance exit and a resignation both reduce headcount, but one is the organisation making a decision and the other is an employee making one. Redundancies, the end of fixed-term contracts, retirement and dismissal belong on the involuntary side; resignation and non-return from leave belong on the voluntary side. Capture the reason code close to the event, recorded by someone who knows the facts, rather than reconstructed months later from memory. Where the split is missing, a restructuring shows up as a retention crisis, and a genuine retention problem gets excused as restructuring.
Regretted attrition is the departure of somebody you wanted to keep, and it is the number that should worry you. Deciding it needs a rule set agreed before the fact, not a judgement made after a resignation letter arrives, because hindsight makes almost everyone look either irreplaceable or expendable. A workable rule draws on the last performance outcome, whether the role is being backfilled, and the manager's view captured at the time, with the tie-break sitting outside the manager's hands. Record the classification alongside the leaver so the split can be recomputed later. Without it, the headline rate mixes people you fought to keep with people you were relieved to see go.
A monthly rate answers how many people left recently; a cohort answers whether people who joined together are staying, which is the question worth asking. Group leavers by joining month or joining quarter and track how much of each group remains as time passes. Patterns emerge that a rate cannot show: a group hired during a rushed expansion thinning out early, a group from one hiring source holding steady, a cliff at a particular point in tenure that lines up with a vesting date, a probation boundary or the end of a bond. Cohort survival curves are also the fairest view for a growing organisation, which always looks retentive when measured against swelling headcount, and pulling them together is what an [HR analytics layer](/hr-analytics-software) is genuinely good at.
Tenure band first, because early departures point at hiring, onboarding and how the job was described, while late ones point at progression and pay. Then manager, team, location and role family. Then hiring source, which closes the loop back to recruitment: if one channel or one agency produces people who leave inside the first year, that is a sourcing decision rather than a retention one. Cut by shift or roster pattern where it applies, since departures in operations often track the roster more closely than anything else. Every cut needs the leaver record and the employee record joined, which is straightforward when both live in one [employee database](/employee-database-software) and painful when they do not.
Carefully, and often by not publishing it. A team of a handful swings wildly on a single departure, and a manager confronted with a percentage built from two leavers will reasonably dispute it. Report counts rather than rates when the group is small, pool small groups into a sensible parent, or widen the window until the numbers steady. There is also a privacy line: a cut fine enough to identify who left, presented next to a reason code, is a disclosure rather than an analysis. Set a minimum group size for anything published and apply it to every report, including the ones leadership asks for informally.
Less than they are usually given credit for. People leaving soften the account, protect a reference, and name a proximate cause such as pay rather than the manager relationship that made pay start to matter. The correction is corroboration: read the stated reason against what the same person wrote in an earlier [engagement survey](/employee-engagement-software), against their leave and absence pattern, against whether they applied internally and were passed over, and against what others who left the same team said. Where the sources agree, act. Where they conflict, the exit interview is the weakest of them. Aggregate patterns also travel better than any single conversation, so resist rewriting policy from one articulate leaver.
Pitch N Hire is an applicant tracking system built for recruiters and hiring teams. If this answer described something you want to run properly, the ATS is where it lives.
Free for 1 user · No credit card · Talk to a real hiring expert
Get a personalized walkthrough of Pitch N Hire on your own roles and workflow. No slides, no obligation.
Prefer to talk? Book a demo · Talk to sales · View pricing
Free 1-user plan · No credit card · Talk to a real hiring expert
See your true cost-per-hire and how much Pitch N Hire could save you — our free Recruitment ROI Calculator gives you the numbers in under a minute. No signup required.
Open the free ROI calculatorPrefer a tailored walkthrough on your real roles? Drop your work email:
★ Free 1-user plan · No spam · Talk to a real hiring expert
