A resignation letter has a date on it, and that date is the input most turnover formulas ignore. The cost of employee turnover in an accounting firm is set by what the empty seat was going to do in the weeks it stands empty, and the deadlines arrive whether the chair is filled or not. A multiple of salary cannot know that, because it does not know your calendar. One voluntary departure of a competent person can be priced on the lines only a departure creates, from records you already keep.
What the Cost of Employee Turnover in an Accounting Firm Is Made Of
Start with what makes a departure different from a hiring mistake. A bad hire's seat was occupied and producing something, however thin. A turnover seat produces nothing while the same work stays due.
So the departure creates its own lines. A vacancy with a length, knowledge that walked out with a person, a notice period that either buys something or buys nothing, and a team that absorbs whatever is left. The payroll gross-up and the statutory separation costs behave the same way in both cases, and both are already priced in the cost of a bad hire, along with the reason a published multiple will not do this job.
One line runs the other way, and leaving it out is what gets the total argued away. A vacant seat is not on payroll, so the honest figure is what the vacancy cost minus the compensation that was never handed over.
Start the Clock at the Resignation, Not the Last Day
The vacancy is not the gap between a last day and a start date. It runs from the day the resignation lands to the day the replacement is doing the work without supervision, and it holds three intervals: the notice period, the search, and the ramp.
Two of those usually get measured. The ramp rarely does, and it is the interval that consumes reviewer hours rather than freeing them. Your own records hold every date you need for it. What your exit records answer reads those files for reasons, and pricing a vacancy reads the same files for elapsed weeks.
Write the three intervals down for the last person who left. The total is usually longer than the number partners quote each other.
Price the Vacancy Against the Deadline It Sits In Front Of
A vacancy week in February and a vacancy week in July are not the same cost, and the difference sits on the tax calendar rather than in the payroll register.
For calendar-year filers, the 2026 calendar puts partnership and S corporation returns on March 16 and individual returns on April 15, with the extended partnership and S corporation returns falling on September 15 and extended individual returns on October 15 (IRS Publication 509 (2026), Tax Calendars). A seat that empties in February stands in front of the first two dates. A seat that empties in May stands in front of the other two.
That is also why the work does not vanish when the seat does. Whatever the vacancy pushed out of March arrives again in September and October, on top of what was already scheduled there, and that is the same window a replacement hired in the spring is still ramping through. Extending is a real lever when it is chosen rather than defaulted into, which is how capacity planning treats the decision.
Where the Work Went: Absorbed, Delayed, or Declined
Every hour the empty seat was going to work went to one of three places, and each carries a different price.
Absorbed. Somebody still at the firm did it. The cost is what that hour actually cost less what it was going to cost, which is usually a reviewer doing preparer work, plus whatever that reviewer did not get to.
Delayed. The job moved into the extension window. The cost is the fee arriving later than planned, the second touch when the file is reopened months later, and the client's patience for both.
Declined. The firm did not take the work. The cost is the fee, and it is the easiest of the three to state without an argument, as long as you can name the engagement and the date you turned it down.
Sort the vacancy's committed hours into those buckets before pricing anything. A firm that absorbed everything has a capacity cost and no revenue cost. A firm that declined work has both.
```text One vacancy, sorted before it is priced (replace with your own inputs)
Vacancy window resignation Feb 6 to replacement productive May 4 Committed hours in it 420 hrs (your number: that seat's scheduled work)
Where those hours went Absorbed 260 hrs (your number: reassignment records) Delayed 120 hrs (your number: jobs extended) Declined 40 hrs (your number: work turned away)
How each one prices Absorbed 260 x the cost per hour of whoever actually did it, plus the work that reviewer did not do Delayed the fee received later, plus the hours to reopen and re-familiarize on each extended job Declined the fee, named client by client, with a date Less payroll not paid across the vacancy window
Absorbed hours carry their full cost above, because the single Less line does all of the netting. Do not net them twice.
The subtraction at the end is not optional. A total that skips it overstates the departure, and it gets dismissed in the room. ```
What Left With the Person
Some of what a departure costs never shows up in an hours report, because it was never written down anywhere.
Three things tend to leave in the same week. The client contacts, meaning who at the client actually answers and what they were promised in November. The file conventions, meaning how this client's workpapers are built and which prior-year treatments are deliberate. The queue, meaning what was waiting on what.
Documentation is one part of that you can hold in advance, and SOPs, checklists and treatment notes that live in your systems survive turnover while the same knowledge in one preparer's head does not. The practical version is narrower than a documentation program. Can a competent preparer who has never seen this client finish the current job from the file alone? Run that test on the handful of clients one person owns, before anyone resigns, and the risk is priced rather than discovered.
Another part you can hold in advance is a second preparer who has already worked those files, which is what cross-training is for.
A client who follows that person out adds work on the way through the door, because your firm still owes the prompt return, on request, of the client records that are necessary for that client to comply with their Federal tax obligations (eCFR, section 10.28, Return of client's records). Which categories have to go back, and which do not, is worked through where a whole book of files moves at once, in switching providers. The work lands on whoever is left, in the same weeks they are covering the seat.
What the Notice Period Actually Buys
The notice period is a resource, and most firms spend it on the wrong thing.
The default is that the leaver spends it finishing their own jobs. That protects this month and buys the vacancy nothing. The alternative is to spend part of it writing down what only they know and introducing clients to whoever is picking them up, which shortens the vacancy at the other end.
Decide which of those you are buying on the day the resignation lands, not in the final week. Name the receiver for every client, list the files whose conventions have to be written down, and book the calls while the person is still on payroll.
An overlap between the person leaving and the person arriving is the strongest version of this, and in a direct hire it is usually impossible, because the replacement does not exist yet. Where an arrangement can promise one, it is a contract term rather than a courtesy, which is what the rolloff rule is for in a staffed team.
The People Who Stayed
The vacancy gets absorbed by people who did not resign, and what it costs them is the part most often left out of the total.
Two questions make it concrete. Who took the reassigned work, and what did they stop doing in order to take it? Whether that allocation was decided or simply accreted shows up in the names against your extended files, which is a culture reading of the same records.
Whether the extra hours cost cash at all depends on how those seats are classified, and the classification test for accounting roles runs through hiring a staff accountant. For staff who are exempt from overtime the hours are free in the payroll register and expensive in every other ledger.
Then check the interval in your own records. Take the last several departures and measure the weeks between each resignation and the next one. If a second resignation regularly follows the first inside the same season, treat that as a pattern to look into rather than a proven cause, and start with the workload absorbed in between. What burnout research actually measures is a separate question with its own evidence.
What the Total Is For
A turnover cost that lands in a dashboard changes nothing. Two decisions can use it.
The first is retention against replacement. Put the departure total next to what keeping the person would have cost, remembering that a raise is annual and permanent while the departure cost lands once per departure and comes back at whatever rate your firm turns over. Your own staff turnover ratio sets that rate, and the published medians by firm size say what a firm your size recorded. One limit is worth stating plainly. No survey establishes that more money would have kept the person who left you.
The second is capacity. If the same seat has turned over twice, the question stops being who to hire and becomes whether the work is shaped as a job somebody will stay in. That decision has different options, and one of them is not filling the seat.
Questions Firms Ask
What Is the Typical Cost of Employee Turnover?
Published answers arrive as a percentage of salary, and the spread between them is wide enough to support any plan you already wanted. Ask who measured it, on which employers, and what they counted, which is the test the cost of a bad hire applies to that whole family of figures. For a firm, the useful number is built from your vacancy weeks and your committed work.
What Is the Average Turnover Rate for Accountants?
The published rates are firm-level rather than accountant-level, and they are medians rather than averages. The median firm turnover ratio was 0.00% in the four size bands below $1.5 million in net client fees and 12.00% above $10 million (2025 National MAP Survey), and why accountants are quitting reads every band.
Use a rate to annualize a departure cost you built yourself, rather than as evidence about your own firm.
Only One Person Left. Is This Worth Doing?
Yes, and one is the easy case. Every input is a record rather than a rate: one resignation date, one set of committed hours, one list of what left with the person. A firm with a single departure can build a more accurate number than a large one, because nothing has to be averaged.
When Is Refilling the Seat the Wrong Answer?
When the seat was not the constraint. If the departure exposed that one person held the only route past a bottleneck, refilling it rebuilds the same single point of failure. Fix the route first, then decide how many hands it needs.
Price the Last Departure Before the Next One
Take the last person who resigned and write down four things: the date they resigned, the date their replacement was genuinely productive, the committed hours in between, and where those hours went. That is what the departure cost, and it is defensible in a way no multiple of salary will ever be.
Then take it into the decision it belongs in. If the total is mostly absorbed hours, you have a capacity problem wearing a retention costume. If it is mostly declined work, the seat was paying for itself and the search deserves more than it is getting.
If the honest answer is that the next resignation would cost you a season, test the work before you commit to a person. Accountably places trained offshore accountants and tax preparers inside US CPA and EA firms, ramped on your software and SOPs in about 3 to 4 weeks, and since 2022 that has meant 30+ placements across 20+ US firms. The signature, the opinion and the final judgment stay with your firm. The entry point is a Free 40-Hour Proof Pilot on a fixed block of your own representative work, put through multi-layer review, so your reviewer grades real output before your name is on the line. If a placement is not the right fit in the first 30 days, we replace them free. Don't trust us. Test us.