Published November 8, 2026.
Early turnover is often reported as a single percentage, but that number can mix a resignation, a failed probation, and a seasonal contract ending. This review sets out a bounded protocol for measuring what share of new hires leave within the first ninety days and what the result can honestly show. It uses a start cohort rather than the month an exit was recorded. The cohort view keeps a recent onboarding problem from hiding inside an annual average.
Question and scope
This review asks what share of new hires leave within the first ninety days of their start date. The unit is one hire who begins in the defined window, followed from the start date to the end of the ninetieth day. The study separates voluntary exits, employer terminations, and contract or seasonal endings. It does not estimate a national early-turnover rate or claim that one exit caused one management decision.
Set the scope to one employer, one role family, and one hiring process. Record the definition of a new hire, the start date source, and the exit categories at the beginning. If the onboarding or the probation policy changes during the window, keep the earlier definition for the cohort already started and analyze the later cohort separately. A new probation rule can look like a retention change when it is only a different contract.
Exclude interns, temporary staff, and employees who transfer internally unless the business treats them as new hires. State the exclusion and the reason. Include rehires if the business counts them, but analyze them as a separate group because their onboarding experience is different. Test the definition on two recent start cohorts before collecting the full sample.
Data sources and methodology
Draw the start cohort from the human-resources system and the payroll record, not from a list of known departures. Record the export date, the filters, the row count, and the selection rule. Match each start to the exit record by employee identifier and date. A hire who leaves and returns within ninety days is one start with two exit events, not two hires.
Use the start date as the denominator date. A common error is to count exits by the month they were recorded, which mixes cohorts and hides a recent onboarding problem. A cohort view follows the same group for ninety days and reports the share that left. Report the cohort size beside the rate so a small group is not overinterpreted.
Collect the exit category, the last working day, the reason, the role, the location, and the manager. Add whether the exit was voluntary, involuntary, or a contract end. Where the reason is missing, mark it unknown rather than inferring it from the exit interview text. A second reviewer should be able to repeat the classification from the record.
Why the first ninety days matter
The first ninety days cover the period when a new hire learns the job, the team, and the informal rules. A departure in week two often points to a mismatch between the role and the job description, while a departure in week ten may point to a workload or management problem. The timing changes the interpretation and the fix. Early exits are also expensive because the hiring cost has been spent and the replacement has not started. The business loses the training time and the team loses confidence in the hiring process. A small early-turnover rate can still be costly if it concentrates in a hard-to-fill role. Report the role and the cost, not only the percentage.
A new hire who stays may still be at risk. The protocol measures exits, not engagement or performance. Read the result beside the probation review and the first-month check-in. A low exit rate with a high probation failure rate tells a different story from a low rate with strong reviews.
Defining an early exit
Define an early exit as a hire whose last working day falls on or before the ninetieth calendar day after the start date. A termination during probation counts. A contract that ends on the agreed date does not count as an early exit unless the business ended it sooner. A transfer to another role inside the business is not an exit and should be recorded separately.
Do not count a hire who never started. A withdrawn offer or a no-show on day one is a pre-hire event and belongs in a different measure. Do not count an exit that was reversed within the same week as a departure if the employment record shows continuous service. Each definition needs a positive and a boundary example. Report the rate with and without contract endings so the reader can see the effect of the category. A business with seasonal hiring can look like it has a crisis when the contracts simply ended as planned. The split makes the comparison honest.
Sampling and data quality
Include every start in the window rather than a sample, because the cohort is usually small enough to review in full. If the volume is too large, take a random sample within each role family and report the sampling rule. Do not sample only the roles that are easy to contact. Record the response and the missing data.
Check the start and exit dates against the payroll record. A system may record the offer date as the start date or the last pay period as the exit date. Correct the obvious errors and document the rule. Do not silently change a date to make a case fit the definition. Note the collection date separately from the cohort window. A cohort that started three months ago may still have incomplete exit data if a person left after the observation date. Censor the data at the observation date and report the censored count.
Analysis and reporting
Report the cohort size, the number of early exits, and the share by exit category. Show the counts by role family, location, and manager. Present the median tenure at exit for the voluntary group. Use the same denominator for every rate and show the number of records with unknown reasons.
Do not rank managers on a small number of exits. A manager with one hire and one exit is not comparable to a manager with twenty hires. Report the count and the interval, and describe the cases rather than computing a league table. The purpose is to find a pattern, not to assign blame.
Inspect the voluntary exits and group the reasons without naming the employee. A cluster in one role may point to a job description that oversells the work. A cluster in one location may point to a schedule or transport problem. Compare the pattern with the hiring source and the onboarding feedback.
Limitations and sources of error
The record cannot show whether the new hire would have stayed with a different manager or a different schedule. It can show only that the person left and what reason was recorded. A low early-turnover rate is not proof of good onboarding, because unhappy employees may leave later. The protocol also cannot capture a hire who mentally quit but stayed on the payroll.
Exit interviews are often completed by a small share of leavers and may soften the real reason. A contract end can be recorded as a resignation by mistake. Record those limitations and the treatment in the report instead of presenting the rate as exact. If the sample is small, describe the cases rather than computing a rate. A hire who moves to another role inside the business is not an exit, but the move may be a retention success or a failure of the original role. Record the move separately and note the reason where it is known. That data helps the business see the difference between losing a person and losing a role.
Review cadence and next action
Review the result with the hiring manager, the HR owner, and the onboarding lead. Ask them to choose one observable change: a more accurate job preview, a first-week check, a named buddy, a schedule review, or a probation checkpoint. Record the owner, the effective date, and the measure that will show whether the change worked. Repeat the protocol with the same definitions for the next cohort. If the early-exit rate falls but the ninety-day voluntary rate rises, the change may have delayed the problem rather than solved it. If the rate is stable but the reasons change, the onboarding may be improving in one area and failing in another. Close the review when the owner accepts the finding and the next cohort is scheduled.
Data Sources and Methodology
The review uses public employment guidance to bound the definitions and to keep the recommended record fields within common legal and administrative practice. The sources below are used as context, not as a claim that any one employer follows them in the same way. Each measure in this protocol is defined from the employer's own records, then compared with these public references where the reference is relevant. The citations are listed with the specific topic they support so a reader can check the boundary without searching the whole site.
- U.S. Bureau of Labor Statistics, HR specialists occupational outlook
- U.S. Bureau of Labor Statistics, employer costs for employee compensation
- U.S. Department of Labor, hour and leave standards
- U.S. Equal Employment Opportunity Commission, laws and guidance
- U.S. Small Business Administration, manage your business
- National Institute of Standards and Technology, Cybersecurity Framework
- U.S. Department of Health and Human Services, HIPAA Privacy Rule
- Internal Revenue Service, understanding employment taxes
- U.S. Census Bureau, business help site
- Gallup, workplace research
The method is deliberately descriptive. It does not pool employers or estimate a national rate. A reader who repeats the protocol should use the same definitions, the same observation window, and the same sampling rule, then report the counts beside the shares. Keep that report with the study record so a later reviewer can repeat the comparison.
Sources
- U.S. Bureau of Labor Statistics, job openings and labor turnover - accessed November 8, 2026.
- Society for Human Resource Management, onboarding resources - accessed November 8, 2026.
- U.S. Department of Labor, employment law guidance - accessed November 8, 2026.
- U.S. Equal Employment Opportunity Commission, recordkeeping - accessed November 8, 2026.