Measure First-Year Turnover Correctly in 2026

The BLS Employee Tenure release reported that 22 percent of wage and salary workers had one year or less of tenure in January 2024. That observed national share is not an employer's first-year turnover rate, but it makes the one-year boundary analytically material. An employer claiming a 365-day outcome must give every starter the full 365-day opportunity to experience the event or use a method that explicitly handles incomplete follow-up.

Put every starter on the same clock

First-year turnover is a cohort outcome, not exits divided by average headcount. Select employees by start date and observe each for 365 days. Recent starters who have not reached that anniversary cannot be counted as retained; doing so biases the result downward.

Define rehires, acquisitions, internal transfers, seasonal assignments, leaves, and delayed starts before extraction. The start should normally represent the first day employed in the analyzed spell. Keep voluntary resignation, discharge, retirement, assignment end, death, and data correction distinct. Publish all-separation and voluntary views rather than silently redefining turnover.

Mature-cohort arithmetic

Suppose January produced 100 starters and 18 separated within 365 days. First-year all-separation turnover is 18%. A June cohort with 120 starters, ten exits, and only eight months of follow-up does not have an 8.3% first-year rate. Report its eight-month landmark outcome or use survival methods.

Kaplan to Meier estimation accounts for right-censoring but requires event dates, censor dates, and clear risk rules. Show numbers at risk beneath curves because late estimates can depend on very few people. Survival analysis improves use of partial follow-up; it does not repair bad exit codes or missing employees.

Move from rate to diagnosis

Plot hazards or exits by tenure week and month. A day-three cluster suggests different questions from a month-eleven cluster. Compare roles, sites, shifts, managers, recruiters, and start cohorts only when definitions remain comparable and groups are large enough to protect people and support inference.

Connect patterns to case evidence: schedule mismatch, pay discrepancy, access delay, supervisor conduct, workload, commute, better offer, attendance, or proficiency. Exit interviews are incomplete and retrospective, so triangulate with onboarding records, schedule changes, employee-relations cases, and manager evidence.

The BLS JOLTS handbook documents how the program estimates economy-wide quits from establishments. Its December 2024 estimate of 3.2 million quits is labor-market context, not a first-year employer benchmark. Likewise, the BLS median tenure of 3.9 years describes a broad worker population, not new-hire survival probability.

Evaluate retention action honestly

Predefine intervention eligibility, implementation date, mechanism, process measure, and mature outcome. Compare equivalent start cohorts and document compensation, manager, location, hiring standards, and labor-market changes. A lower post-program rate is association unless design supports a causal claim.

Turnover is not inherently bad. A low rate can coexist with weak performance management or limited outside opportunity. Pair retention with safety, quality, attendance, employee voice, and fair-treatment evidence. Suppress tiny manager cells and restrict identifiable exit reasons using controls such as those in the NIST Privacy Framework.

For help operating hiring and onboarding workflows, see recruiting services. To evaluate vendors, agencies, software, and internal models, visit the alternatives directory.

Applied example

January has 100 starters and 18 leave within 365 days, so mature first-year turnover is 18%. June has 120 starters but only eight months of follow-up on the report date; dividing its 10 exits by 120 and calling 8.3% “first-year turnover” is invalid. Report an eight-month cohort outcome or a censored survival estimate.

Data engineering rules for tenure cohorts

Create one employment-spell table with person key, start timestamp, separation timestamp, event type, last-observed date, organizational unit history, and source-system lineage. Validate that exits do not precede starts, concurrent spells follow a stated rule, and corrections retain their prior values. Freeze each published cohort extract so later HR-system edits can be reconciled rather than silently rewriting history.

Calendar conventions need disclosure. A 365-day definition differs around leap years from “same date next year,” and both differ from twelve payroll periods. Choose one rule and test boundary cases. For day-level hazards, decide whether an employee separating on the first anniversary experienced the event inside or after the window.

Managers need counts as well as rates. Eighteen exits among one hundred starters convey more information than 18% alone. Include missing exit classifications and employees lost during system migration. When comparing cells, show uncertainty and avoid league tables that encourage managers to discourage legitimate transfers or necessary separations.

A diagnostic cohort can attach the attributes known at start, but later variables require time-aware treatment. Assigning an employee to the manager they had at exit can wrongly credit or blame that manager for the entire spell. Preserve effective-dated manager, shift, pay, and location histories when the analysis asks about changing exposure.

Reporting package for decision makers

A useful quarterly package contains mature 365-day outcomes, 30/90/180-day landmarks for newer cohorts, a survival curve with numbers at risk, exit-type composition, and tenure timing. Attach definition version, extraction date, cohort counts, exclusions, unknown classifications, and known system changes. Separate organization-wide context from manager-level action.

Set investigation thresholds before reviewing names. A threshold can combine minimum cohort size, material rate difference, repeated periods, and case evidence. This prevents attention from chasing random fluctuation. Do not wait for statistical significance to address a credible safety or harassment concern; the EEOC harassment resource describes reporting and employer response responsibilities, and individual cases follow their own response process.

When leaders request projected savings, label assumptions. Multiply only an observed, suitably supported change by a locally defined avoidable cost, and show a range. External replacement-cost ratios should not be inserted as if they were the employer’s ledger.

Evidence coverage

The formal evidence ledger also supports the article’s definitions, safeguards, and boundary conditions through Talent Management Reference Materials, Form I-9 Acceptable Documents, Training Requirements in OSHA Standards, Workplace Posters, Recordkeeping Requirements. These materials are used for the claims and limitations stated above; they are not presented as proof of effects beyond their stated populations.

Data sources and methodology

This article synthesizes the ten primary and professional sources recorded in frontmatter and linked below. Regulatory materials define compliance context; federal statistical publications define their own populations; operating examples are explicitly illustrative. Sources were checked July 22, 2026. No vendor headline was treated as a universal benchmark, and associations are not described as causal effects.

Frequently asked questions?

Can current-year starters be included in a first-year rate?

Only those with complete 365-day follow-up belong in a simple mature-cohort rate. Use landmark or survival reporting for the others.

Should internal transfers count as turnover?

Usually report them separately from organizational separations. The correct treatment depends on whether the question concerns company, site, manager, or role retention.

Sources reviewed

The article uses the sources below for definitions, context, governance, or safeguards. Inline links identify the relevant source at the point of a material claim. Publication dates in the record use n.d. when a reliable page date was not available; all links were reviewed with an access date of July 22, 2026. For measure first-year turnover correctly, apply this rule to the defined population and decision above.

• Talent Management Reference Materials : U.S. Office of Personnel Management, n.d.
• Form I-9 Acceptable Documents : U.S. Citizenship and Immigration Services, n.d.
• Training Requirements in OSHA Standards : Occupational Safety and Health Administration, 2015-01-01.
• Workplace Posters : U.S. Department of Labor, n.d.
• Employee Tenure in 2024 : U.S. Bureau of Labor Statistics, 2024-09-26.
• Job Openings and Labor Turnover, December 2024 : U.S. Bureau of Labor Statistics, 2025-02-04.
• Job Openings and Labor Turnover Survey Handbook of Methods : U.S. Bureau of Labor Statistics, n.d.
• Harassment : U.S. Equal Employment Opportunity Commission, n.d.
• NIST Privacy Framework : National Institute of Standards and Technology, n.d.
• Recordkeeping Requirements : U.S. Equal Employment Opportunity Commission, n.d.