First-year attrition rate: how to calculate it on a cohort and what to do with it

First-year attrition rate is the share of new hires who leave within 12 months of starting: hires from a start cohort who left within 12 months of their own start date ÷ hires in that cohort × 100. Unlike an annual turnover rate, it follows a defined group of new hires, so it tells you something specific about how well those people were hired and brought in. It is also the most objective input to quality of hire.

This page covers the cohort method, a worked example split by exit type and role family, how to handle cohorts that are not yet 12 months old, and a report format that hiring managers can act on.

The formula, done on a cohort

First-year attrition = leavers within 12 months of start ÷ hires who started in the cohort × 100

Cohort:   everyone who started in a defined period (a quarter or a year)
Complete: every member has either left or passed their 12-month anniversary

The alternative that often appears in HR reports is a period method: people with less than a year of tenure who left this year, divided by people hired this year. The two groups do not match. Someone hired last November who leaves in March counts as a leaver this year but was hired last year, so the fraction mixes cohorts and can even exceed 100% in a year when hiring slows. The cohort method avoids that and answers the actual question: of the people we hired, how many stayed a year?

Worked example: a full-year cohort

The 90-day figure deserves its own line. People who leave in the first three months are usually telling you the job was not what they expected, which points at the job description, the interview or onboarding. People who leave between months 6 and 12 more often point at the manager, pay progression or workload.

Split it by role family and source

Customer support accounts for 7 of the 11 leavers and 3 of the 4 early leavers. That is where to look first. Operations' 0% is ten people; one leaver would make it 10%. Show counts beside every percentage, and read groups smaller than ten or so case by case rather than as rates.

The same split by source of hire is often the most useful one for recruiting. If referral hires in support stayed and job-board hires did not, that changes where the next support requisition is advertised. Reliable source data is its own discipline; see source of hire tracking.

Handling cohorts that are not yet 12 months old

The cohort method has one cost: you cannot report the first-year rate for people hired eight months ago, because the year is not over. Three ways to report recent hires without waiting:

• Report the 90-day rate early. A cohort that started in Q1 has a complete 90-day figure by early Q3. Use it as the leading indicator.
• Report survival at each checkpoint. For each start quarter, show the share still employed at 3, 6, 9 and 12 months, leaving cells blank until they are complete.
• Never divide by the whole cohort early. Eight months in, "2 of 20 left" is not a 10% first-year rate; it is a 10% eight-month rate, and the final figure can only go up.

The four quarters add up to the 60 hires and 11 leavers above: 3 + 4 + 2 + 2 = 11 left by 12 months, and 1 + 1 + 1 + 1 = 4 within 90 days. Reported this way in, say, April 2026, the Q4 row would show only its 3-month column, and nobody would mistake a partial number for a final one.

Is there a benchmark?

Be wary of first-year attrition percentages quoted without a source, a definition and a date. The most widely used official US data, the Bureau of Labor Statistics' Job Openings and Labor Turnover Survey (JOLTS), publishes monthly hires, quits and separations rates across all employees by industry. It does not break separations down by how long the person had been employed, so it cannot be compared with a new-hire cohort figure. Your own previous cohorts, by role family, are the benchmark that matters.

What to count, and what to leave out

• Internal transfers and promotions: not leavers. The person is still employed.
• Fixed-term, seasonal and temporary hires: leave them out, or report them separately. Someone whose planned contract ended did not leave early.
• Rehires: treat each employment spell as a separate start, or exclude rehires; pick one.
• Deaths and long-term leave: exclude deaths; count people on leave as still employed.
• Layoffs and restructuring: count them, but in their own category. They are not a hiring outcome and should not be charged against the recruiting process.

Record a reason for every early exit. For voluntary leavers, the exit interview is the best source; see exit interview questions and how to analyse exit interview data. For people still in their first year, a structured check-in can catch problems before they become resignations; the new hire 90-day review template and stay interview questions both help.

Turning the number into a hiring change

First-year attrition is only useful to recruiting if it leads back to a decision in the hiring process. Take the early leavers in the worst-performing group and, for each one, answer three questions from the records you already have:

1. What reason did they give for leaving, or what was the performance issue?
2. Was that issue tested in the interview? Look at the scorecard for the competency that failed.
3. If it was tested, what score did it get, and what did the candidate actually say?

If three of the seven support leavers cited the volume and pace of tickets, and the interview never asked about sustained high-volume work, the fix is a question and a realistic job preview, not a new sourcing channel. If the interview did ask and every candidate scored well, the question is not separating people, and the scoring anchors need work. That review only works if interview scores were recorded per competency with the evidence behind them; notes that say "good energy" cannot be checked a year later.

How to report it

FIRST-YEAR ATTRITION, 2025 START COHORT (complete as of Dec 31, 2026)
Method: left within 12 months of own start date ÷ starters.
Excludes internal moves, fixed-term hires, layoffs (reported separately: 0).

Starters: 60   Left < 12 months: 11 (18.3%)   Left < 90 days: 4 (6.7%)
Voluntary: 7 (11.7%)   Involuntary: 4 (6.7%)
Previous cohort (2024 starters): 16.0%, on 50 starters

By role family: Support 7/24 · Sales 3/14 · Engineering 1/12 · Ops 0/10
Top reason, support leavers: pace and ticket volume (3 of 7)
Action: add high-volume work sample to support final round from Feb 1.

Common mistakes