Short answer
Employee turnover rarely has one proven cause. Start by defining which separations you are investigating, locating the pattern, combining operational records with employee accounts, and testing competing explanations. A resignation reason is evidence, but it is not a causal verdict. Choose a limited action and compare what changes.
Stop treating every separation as the same event
A root-cause exercise fails quickly when its numerator mixes voluntary resignations, dismissals, redundancies, retirements and internal transfers. These events have different owners and possible explanations.
The US Bureau of Labor Statistics separates quits, layoffs and discharges, and other separations in its Job Openings and Labor Turnover Survey. It defines quits as voluntary separations and excludes retirements and transfers to other locations. That is a useful reference, but an employer still needs a written internal definition that matches its HR system. UK teams should apply the same discipline before using a turnover formula or an external comparison.
Write a scope statement before asking why people left. For example:
We are investigating voluntary resignations by permanent warehouse operatives in their first 180 days at two UK sites during April to June. Internal transfers, dismissals and fixed-term contract completions are excluded.
This statement prevents an apparent retention problem from being driven by a change in contract mix or a redundancy programme.
Use cause categories as hypotheses
Lists of turnover causes are prompts for investigation, not league tables. A practical hypothesis library can include:
- Job expectations and role design: the work, autonomy or physical demands differ from recruitment messages.
- Schedule and access: shift notice, commuting options, overtime or flexibility make attendance hard.
- Workload and resources: staffing, equipment, handovers or priorities prevent sustainable work.
- Pay and recognition: total reward, pay progression, fairness or appreciation do not meet expectations.
- Manager and team practices: support, feedback, conflict handling or allocation of work varies locally.
- Development and mobility: people cannot see or access a credible next step.
- Fairness, inclusion and safety: treatment, opportunity, psychological safety or physical safety needs review.
- Personal and external factors: caring responsibilities, relocation, education or another labour-market opportunity.
- Organisational change: a merger, restructuring, policy change or leadership transition altered the work.
One departure can involve several categories. Do not force a single primary reason just to simplify reporting.
Build an evidence register before choosing an intervention
For each plausible explanation, record the evidence that would support it, evidence that would weaken it, and what is still missing.
| Hypothesis | Evidence to inspect | Alternative explanation | Decision test |
|---|---|---|---|
| Late schedule changes contribute to resignations | Rota publication dates, swap requests, repeated employee accounts, resignation timing | A new competitor raised local pay at the same time | Publish rotas earlier for one eligible group and track the agreed outcomes |
| New hires leave because the role differs from the advert | Job advert, realistic job preview, onboarding questions, early exit accounts | One trainer or site had an onboarding gap | Revise the preview and compare early-tenure outcomes with a similar intake |
| Progression is unclear | Internal applications, promotion criteria, development conversations, accounts from stay and exit interviews | Vacancies fell because the organisation paused expansion | Publish criteria and hold role-path sessions, then test awareness and applications |
| Manager practice is part of a local spike | Team-level timing, manager changes, workload allocation, repeated accounts | The team received the hardest customer or production mix | Review both work allocation and manager routines before assigning responsibility |
Use at least three evidence families where available:
- Workforce records: separation type, tenure, role, site, contract and manager history.
- Operational records: schedules, workload, vacancies, absence, incidents, pay changes and process delays.
- Employee accounts: exit interviews, stay conversations, onboarding feedback and employee voice.
- External context: local transport, labour demand, regulation or a site-specific event.
Agreement across sources strengthens a hypothesis. It still does not prove causation. Disagreement is useful because it shows where more specific questions are needed.
Ask questions that preserve context
Acas says an exit interview can help an employer understand why someone is leaving and improve future recruitment or retention. Treat that as a learning opportunity, not a forensic verdict.
Useful questions move from event to mechanism:
- What first made leaving a realistic option?
- What changed in the weeks or months before that point?
- Which part of the job made staying harder?
- What, if anything, had you already tried to resolve?
- Which practice should the organisation keep?
- What would have needed to be different, and by when, for you to reconsider?
Record the employee's own account separately from HR's interpretation. Avoid asking a manager to validate criticism about their own team. Explain who will see the response, how it will be used and any confidentiality limits before the conversation begins.
Worked example: diagnose before fixing
This fictional example illustrates the method. It is not a client result or proof of causation.
A US distribution business sees 18 voluntary quits among 140 permanent evening-shift employees over one quarter, compared with 7 among 145 in the previous quarter. Day-shift exits are broadly unchanged.
The first explanation offered is pay. The evidence register finds:
- pay rates did not change between quarters;
- a local employer began recruiting nearby, so an outside pull remains plausible;
- nine leavers and six current employees describe rota changes with less notice;
- the scheduling log confirms a system change reduced average notice;
- workload rose during the same period, so schedule notice may not be the only mechanism.
The team does not label everyone on the evening shift a flight risk. It restores the previous notice standard for eight weeks, keeps paid hours and workload rules unchanged, and records exceptions. The owner tracks schedule notice, swap requests, voluntary quits and short employee accounts for evening and day shifts.
If the evidence improves, the result supports continuing the change. It still may not isolate one cause because seasonal demand and outside hiring can move at the same time. The honest conclusion is narrower: the scheduling hypothesis was credible enough to test, and the test produced evidence for the next decision.
A six-step turnover diagnosis
- Define the population and event. State who, where, when and which separation types count.
- Locate the concentration. Compare role, site, shift, tenure and time without exposing small groups.
- List competing explanations. Include operational and external factors, not only employee sentiment.
- Triangulate evidence. Compare records, employee accounts and timing.
- Choose a reversible action. Assign an owner, guardrails, outcome and review date.
- Update the register. Mark what the test supported, contradicted or left unresolved.
For rate definitions and denominators, use the retail turnover calculation guide. For a conversation method, use the stay interview guide while people are still employed and the exit interview guide after notice.
Where a conversation platform fits
The method can run with interviews, spreadsheets and existing HR systems. A conversation platform becomes relevant when HR needs to collect comparable open-text accounts across groups and return to the source context during review.
Lontra's product approach can support employee conversations and group-level synthesis for human review. It is not an attrition model, an HRIS or a system that proves why someone left. Managers receive a brief rather than raw employee conversations, and small-group reporting still requires disclosure-risk judgment. The related retention signals guide explains how to review patterns without scoring individuals.
Frequently asked questions
What are the main causes of employee turnover?
Pay, scheduling, workload, manager practices, role design, development, fairness, safety, personal constraints and outside opportunities are useful hypotheses. Their importance varies by workforce, location and period, so employers should diagnose rather than assume a universal ranking.
Can an exit interview prove why someone left?
No. It records that person's account and may identify a useful lead, but causal conclusions need corroboration from repeated accounts, operational evidence, timing and competing explanations.
How should HR investigate a turnover spike?
Define the separation population, locate the pattern by role, site, tenure and time, build competing hypotheses, gather evidence from several channels, then test one bounded action with a named owner and follow-up date.
Should HR score individuals for resignation risk?
No. Use group patterns to decide where a human review is warranted. Do not turn comments, engagement responses or workforce records into individual departure predictions or employment decisions.