Short answer
Investigate manufacturing turnover as a bounded plant question. Define which departures count, calculate rates for the same population and period, then examine role, shift, tenure, production changes, staffing, training, and supervision. Add accounts from current and departing employees, preserve disagreements, and choose one controllable work condition for a named owner to test. The result explains where to look; it does not predict an individual's departure.
Fix the turnover definition before explaining it
“We lost 18 people” can describe voluntary quits, dismissals, layoffs, retirements, transfers, seasonal endings, or a mixture. Those events need different decisions.
The US Bureau of Labor Statistics JOLTS definitions separate quits, layoffs and discharges, and other separations. JOLTS also excludes transfers within the same establishment from separations while treating transfers from other locations as hires. Your internal rule may differ, but it must be stable and visible.
Write down:
- the plant, legal entity, roles, and employment arrangements in scope;
- the start and end date;
- what counts as a voluntary quit;
- treatment of retirements, dismissals, redundancies, seasonal completions, agency workers, and internal transfers;
- the opening and closing headcount for the same population; and
- whether corrected or late HR records will restate the result.
For a defined group, a simple period rate is:
qualifying separations during the period / average headcount during the period × 100
The employee turnover rate guide covers calculation choices in detail. This guide uses the result as the starting point for a plant investigation.
Build the manufacturing investigation board
Use one board for one plant question. Do not pool sites until definitions and contexts match.
| Layer | Evidence to place on the board | Discipline |
|---|---|---|
| Definition | Leaver rule, period, population, headcount source | Keep numerator and denominator in the same scope |
| Pattern | Count and rate by role, shift, tenure band, and plant | Show counts beside rates; protect small groups |
| Production context | Product or volume change, overtime, staffing, line or shift change, downtime | Record timing without declaring a cause |
| Learning and support | Training access, approved instruction, buddy or trainer coverage, supervisor span | Distinguish attendance from demonstrated use in work |
| Employee accounts | Current-worker, leaver, supervisor, trainer, and representative perspectives | Preserve source, context, disagreement, and missing groups |
| Formal records | HR, safety, quality, grievance, absence, and maintenance records as authorised | Keep each system's definitions and access controls |
| Candidate action | One condition the plant can change | Name owner, approvals, measures, review date, and stop condition |
One person's account can reveal a question worth checking. It does not establish prevalence. A repeated theme can show recurrence in the material reviewed. It still does not prove that the theme caused departures.
Analyse four cuts without manufacturing a story
Role
Separate jobs with different labour markets, training paths, physical demands, or certification requirements. A plant-wide rate can hide a local maintenance, quality, operator, or supervisor pattern.
Shift
Use the average headcount for each shift, not the whole plant denominator. Add relevant context: supervision, training availability, overtime, transport, handovers, break cover, and recent production changes. Do not treat “night shift” as the cause merely because its rate is higher.
Tenure
Choose stable bands linked to a decision, such as the planned onboarding period and established-role period. Verify hire dates and internal moves. An early-tenure pattern may point to job preview, access, learning, support, or selection questions; it does not identify one mechanism by itself.
Plant
Before comparing sites, check role mix, use of agency labour, production cycle, reporting rules, and workforce changes. A percentage from a small specialist site is not directly comparable with a large mixed plant.
Worked example: calculate, then investigate
This example is fictional. A plant reviews voluntary operator quits for one quarter.
Average operator headcount is 180, and nine operators voluntarily left:
9 / 180 × 100 = 5.0%
The shift view uses separate denominators:
| Shift | Voluntary quits | Average operator headcount | Rate |
|---|---|---|---|
| Day | 2 | 80 | 2.5% |
| Evening | 5 | 60 | 8.3% |
| Night | 2 | 40 | 5.0% |
The tenure view shows six quits among an average 30 operators in their first six months, or 20.0%. Three quits among an average 150 longer-tenure operators equal 2.0%. These are overlapping views of the same nine departures, so the rates must not be added.
A cross-tab then confirms that four of the five evening-shift quits were among operators in their first six months. The separate shift and tenure totals alone could not establish that intersection. The checked pattern supports a closer look at early-tenure evening-shift work, but does not show why people left. HR then places dated facts and accounts on the board:
- a new product run began four weeks before three of the departures;
- evening overtime rose during two of those weeks;
- two departing employees mention difficulty finding the approved escalation route;
- current employees disagree about whether trainer coverage is the problem; and
- a supervisor notes that the same route is clear on day shift but not demonstrated during evening induction.
The evidence supports a question about access to approved guidance and trainer coverage. It does not support blaming one supervisor, changing a safety procedure through the turnover review, or scoring the remaining employees for departure risk.
Run one owned operational test
The fictional plant assigns its training owner, operations owner, and competent safety or HSE reviewer to assess a bounded test for the next evening-shift onboarding group:
- name the person responsible for the approved escalation walkthrough;
- schedule the walkthrough within paid induction time;
- ask new operators to locate the current approved route using a fictional scenario;
- record access gaps and questions without rating personal potential;
- route any safety concern through the plant's formal system; and
- review implementation, employee accounts, and turnover only after the agreed period.
The first result is whether the walkthrough happened and people could use the approved information. Later turnover is a separate, lagging result with many possible influences. The test can be useful even if it does not change the turnover rate; it may confirm or reject the proposed mechanism.
Where an action changes work, equipment, shifts, procedures, or health and safety, use the appropriate competent review and consultation route. HSE explains what UK employers must consult workers about, including new procedures, equipment, ways of working, shift patterns, risk information, and training. OSHA's worker-participation guidance similarly covers hazard reporting, job analysis, procedures, training, and evaluation in US safety programmes.
Design participation around production reality
A day-shift meeting or email link does not automatically reach night workers, temporary staff, or people with language or literacy needs. HSE's guidance on how to involve employees recommends choosing methods for the circumstances and combining them when one route is insufficient.
Record who was eligible, invited, able to access the method, participated, and remained missing. Offer the established representative and formal reporting routes. Keep raw accounts restricted to authorised reviewers, and do not publish narrow breakdowns that expose a small group.
Where Lontra can fit
Lontra can support a focused employee conversation about a defined work condition and prepare a manager brief for human review. Managers receive the brief, not employees' raw responses. The conversation should supplement HR records, current-worker and leaver evidence, and formal consultation.
It does not predict individual departure risk, integrate production facts automatically, or replace safety, grievance, union, HR, or HSE procedures. Human owners verify themes against the plant context and decide whether an operational test is justified.
Explore Lontra's employee-conversation workflow, then scope one plant question, eligible population, authorised audience, action owner, and review date.
Frequently asked questions
How should a manufacturer calculate turnover?
Define the leaver event, period, population and denominator first. Report voluntary quits separately from employer-initiated departures, retirements and internal transfers, then use average headcount for the same role, shift or plant scope.
What should a manufacturing turnover investigation compare?
Compare rates and counts by plant, role, shift and tenure alongside production changes, staffing, overtime, training access, supervisor coverage and employee accounts. Keep facts, interpretations and counterexamples separate.
Can employee conversations predict who will leave?
No. Conversations can add context about work conditions within a defined population. They should not produce individual departure-risk scores or replace human review, safety reporting or workforce records.