The cost of a bad hire usually gets quoted as one number, normally a multiple of salary. No published multiple was measured on your firm, and a figure you cannot trace to a sample, a date and a definition of "bad" is not something to plan around.
You can build the real number instead. Two federal datasets cover the payroll side, your time system holds the expensive side, and the exit has costs that arrive whether you budget for them or not.
The Bureau of Labor Statistics puts wages at 68.5 percent, on average, of what a professional employee costs an employer (BLS, Employer Costs for Employee Compensation, private industry by occupational group), so the salary line understates the loss from the first week onward.
What the Cost of a Bad Hire Is Actually Made Of
A hire that does not work out costs a firm in five places, and they do not arrive at the same time or in the same ledger.
The five lines below are not ordered by size. The second is usually the largest one you can actually evidence.
- Payroll spent on unusable output. Every month the seat ran, loaded with benefits and employer taxes, not just base pay. This is the line that decides how much a slow decision costs you.
- Review and rework. Senior hours spent finding and fixing what the hire produced, over and above what the same work normally takes. Your time system already holds it.
- Hiring spend, counted twice. Sourcing, screening and onboarding for the person who left, plus the same block again for the replacement.
- Separation costs. The state unemployment insurance consequence, the benefit-continuation notices, and the administrative hours, all landing in the weeks you are trying to refill the seat.
- Capacity you turned away. Work declined, deadlines extended, clients lost. The least verifiable line, sometimes the largest, which is exactly why it needs named evidence rather than an assumption.
Price the first four from records. Price the fifth only where you can name the work.
Why a Published Multiple Will Not Do the Job
Ask three questions of any bad-hire figure before you repeat it. Who measured it, on which employers, and what did they count as a bad hire? A number that cannot answer all three is marketing copy, and one published by a business that sells recruiting or screening has an obvious direction of interest.
There is a second problem, specific to accounting. In a firm, a weak preparer does not produce less work. They produce work that takes a reviewer longer, which moves the cost onto your most expensive people during the weeks you can least spare them. A figure averaged across all employers cannot carry that effect, because review depth is not the same across industries.
So the number you want is not "the" cost of a bad hire. It is yours, for this role, in this season.
What Is Considered a Bad Hire?
Here is a workable definition. A bad hire is a placement where, after a fair ramp on your own procedures, the output still costs more senior time than the role was meant to free up, and the trend is not moving.
Three parts of that definition do real work. "After a fair ramp" rules out judging someone in week three. "Costs more senior time than the role was meant to free up" ties the judgment to the reason you opened the seat. "The trend is not moving" separates a slow starter from a wrong fit.
Line 1: The Payroll That Bought Work You Could Not Use
Start from what the seat actually cost per month, not from base pay.
If you need a reference point for a role rather than your own band, the mean annual wage for accountants and auditors was $94,750 in May 2025, across 1,449,500 people in the occupation (BLS, Occupational Employment and Wage Statistics, national table).
Then gross the wage up to a compensation cost. For management, professional and related occupations in private industry, employer compensation costs averaged $78.10 per hour worked in March 2026, of which wages and salaries were $53.50, or 68.5 percent, benefit costs were $24.61, or 31.5 percent, and paid leave alone was $7.34, or 9.4 percent (BLS, Employer Costs for Employee Compensation, private industry by occupational group).
Read that as the shape of the cost, not as your number. One step matters when you use it on a salary. A salary already pays for leave hours, while the ECEC counts that pay as a benefit rather than as a wage, so add the paid leave share back to the wage share before you divide, or you charge the same leave twice. Grossed up that way, total compensation lands a bit over a quarter above the salary line, before you add software seats, workspace or supervision. If you have your own benefit and payroll tax detail, use it. The federal share is a stand-in for firms that do not.
The decision this line drives is about time rather than money. Every month between "this is not working" and the exit date costs a full loaded month, so what you lose here is set by how long the decision takes, not by the hiring judgment that started it.
Line 2: Review and Rework, the Line Only Your Records Can Price
This is the line that makes a bad hire expensive in a firm, and it is the one no external benchmark can give you.
Pull it this way. Take the jobs the hire touched, total the reviewer and manager hours charged against them, and compare that to the same job codes in the prior year or to the same work prepared by someone else. Price the difference at the reviewer's cost per hour, not their billing rate, unless the hours genuinely displaced billable work. Do the comparison at the job level, because a firm-wide realization number will hide it.
That gap is real money, and past a point it stops being a coaching problem. The practice rules put the duty to keep adequate firm procedures on whoever holds principal authority over the firm's tax practice, and the trigger for discipline is the part worth reading closely. It runs through three routes, and every one of them needs two things at once, willfulness, recklessness or gross incompetence by the person holding that authority, and a pattern or practice of noncompliance by people at the firm (eCFR, section 10.36(b), Procedures to ensure compliance). One weak preparer caught by review is not that. A preparer whose work keeps failing while nothing in the firm changes is closer to it than it feels from inside the season.
The attest side asks the same question in a different register. Resources is one of the 8 components of the system of quality management that firms performing audit, attest and review engagements have had to run since the December 15, 2025 effective date (Journal of Accountancy, AICPA unveils new QM resources to help firms meet Dec. 15 deadline). Whether the firm has the people to do the work at the required standard is a documented judgment now rather than a private worry. That changes what a bad hire is. It stops being bad luck and becomes evidence about a system.
Line 3: The Hiring Spend You Are About to Repeat
Everything you spent to fill the seat is sunk, and then you spend it again.
Count the partner, manager and administrative hours that went into sourcing, screening and interviewing, and price them at internal cost. Add any fee paid to an agency or a job board. Then add onboarding, which means the software license, the setup, and the hours other staff spent teaching rather than producing.
The reason to total this block separately is that it is the only line you can compress by design. A slower, better-evidenced selection process raises the number once. A weak one raises it every time the seat turns over. If you are refilling the same chair for the second time, this line is where the pattern shows up first.
One warning on cost per hire. There is no honest external benchmark to compare yourself against here, because a published figure that does not say whether internal time was counted, or at what rate, is not comparable to yours. Build the number consistently inside your own firm and compare it to your own prior year. That comparison is worth something. A cross-industry average is not.
Line 4: The Separation Costs Nobody Budgets For
A departure has costs that never touch the payroll register, and two of them are set by law rather than by you.
The first is your state unemployment insurance rate. State unemployment tax is experience rated and works like an insurance program, so an employer starts at an initial rate that then varies with the amount of benefits charged to its account, and more charges mean a higher rate while fewer claims mean a lower one (DOL Employment and Training Administration, Conformity Requirements for State UC Laws, Experience Rating Overview).
The mechanics vary by state, and so does whether a particular separation gets charged at all. The most common formula is the reserve ratio, the second most common is the benefit ratio, Delaware and Oklahoma use a benefit-wage ratio, and Alaska rates employers on variations in payroll (DOL Employment and Training Administration, Conformity Requirements for State UC Laws, Experience Rating Overview). Check your own state's rules before you assume a discharge is free.
The federal piece, by contrast, is capped and small. Federal unemployment tax is 6.0% of taxable wages on the first $7,000 paid to each employee during a calendar year, with an offset credit of up to 5.4% for employers who pay their state unemployment taxes on time, which puts the federal rate at 0.6% for employers in states not subject to a credit reduction, a maximum of $42.00 per employee per year (DOL Employment and Training Administration, Unemployment Insurance Tax Topic).
The second legal cost is the benefit-continuation notice, and it comes with a clock. Where a group health plan is subject to COBRA, the administrator must furnish the election notice to each qualified beneficiary not later than 14 days after receiving notice of the qualifying event, and where the employer is also the plan administrator, not later than 44 days after the loss of coverage if the plan runs the period from that date, or after the qualifying event in all other cases (eCFR, section 2590.606-4(b)(1) and (b)(2), Notice requirements for plan administrators).
Smaller practices may sit outside that regime, and the threshold is worth knowing before you rely on it. A plan is a small-employer plan when the employer normally employed fewer than 20 employees during the preceding calendar year, which means fewer than 20 on at least 50 percent of its typical business days (eCFR, section 54.4980B-2, Q&A-5, COBRA continuation coverage). Many state laws extend similar rights to smaller employers, so treat the federal threshold as the floor rather than the answer.
None of these lines is large on its own. Together they are certain, they are administrative, and they land in the same weeks you are trying to interview a replacement.
Line 5: The Capacity You Turned Away
This line can be the biggest of the five, and it is by far the easiest to inflate, so hold it to a higher standard than the rest.
Price it only from things you can name. A client engagement you declined because the team was full. An extension you filed that you would not otherwise have filed. A client who left and told you why. Each of those has a fee attached to it and a date you can point at.
Do not put a number on morale, reputation or "team disruption". They are real, and they are the reason partners act, but a figure you cannot defend weakens every other line on the page when someone asks where it came from. Describe them and leave them unpriced.
A Worked Example on Stated Inputs
The illustration below uses the federal figures already cited and three numbers that have to come from your own records, so every step stays visible.
``` Illustration, stated inputs only
Role modeled one staff accountant Annual wage $94,750 (BLS mean, accountants and auditors, May 2025) Wage plus paid leave 77.9% (BLS ECEC, management, professional and related share of total cost occupations, March 2026: wages 68.5% plus paid leave 9.4%, standing in for your own benefit and payroll tax lines) Months on payroll 5 (your number: offer date to exit date) Extra review hours 100 (your number: the job-level gap above) Reviewer cost per hour $90 (your number, not a benchmark)
Loaded cost per year 94,750 / 0.779 = $121,630 Loaded cost per month 121,630 / 12 = $10,136 Payroll for the 5 months 10,136 x 5 = $50,680 Review and rework 100 x 90 = $9,000
Subtotal, before hiring spend, separation and turned-away work = $59,680
Notes on the inputs - Paid leave is added to the wage share before dividing. An annual salary already pays for leave hours, while the ECEC counts that pay as a benefit, so dividing by wages alone charges the same leave twice. - The recruiting and onboarding block sits on top and runs twice, because the seat still has to be filled. - The BLS benefit share is measured per hour worked and across all private employers, so the loaded figure is an estimate, not your payroll register. - Swap in your own wage and your own benefit lines and the shape of the answer does not change. Only the size does. ```
Two things are worth noticing about that subtotal. The payroll line is the one a partner can shorten by deciding sooner. The review line is the one that quietly grows while the decision is pending.
When the Firm Is the Defect
Run that definition on the firm before you run it on the person. Were the procedures written down or did they live in a reviewer's head? Was the standard for acceptable workpapers ever stated? Did the job description quietly contain three jobs? Was anyone assigned to answer that person's questions during their first month?
If the honest answer to any of those is no, some of the cost you just totaled belongs to the firm, and hiring a different person will reproduce it. That is the practical value of treating resources as a quality-management question rather than an HR one. It forces the firm to write down what "capable of doing this work here" means, which is also the only standard a new hire can be fairly measured against.
How to Make Being Wrong Cheaper Next Time
You will hire wrong again. The goal is not to eliminate the miss. It is to shrink what a miss costs.
Four changes do most of the work, and each one attacks a specific line above.
- Set the decision date before the person starts. Write down what has to be true by when, and who decides. This attacks the payroll line, which is pure elapsed time.
- Grade real work early, against a written standard. Give a fixed block of representative work, review it the way you would review a client file, and compare it to what a competent preparer would have produced. This attacks the review line by finding the gap in weeks rather than in April.
- Write the standard down before you recruit. The procedures, the workpaper expectations, the review chain. This attacks the hiring line, because a role you can describe precisely is a role you can select for.
- Prefer arrangements where a replacement does not restart the whole cycle. Whether that is a bench, a pool or a provider obligation, the question is the same. If this person is wrong, how long until someone else is doing the work?
That last one is worth pressing on when you talk to anyone supplying people. A firm carries the full cost of a bad hire when it hires directly. Under some arrangements, part of that cost sits with the supplier instead, and the difference only matters if it is written into the agreement rather than promised in a meeting.
Questions Firms Ask
What Does an Employee Cost Above Their Pay?
More than the pay rate, by a wide and measurable margin. Across management, professional and related occupations in private industry, benefits were 31.5 percent of total employer compensation costs in March 2026, with wages and salaries making up the other 68.5 percent (BLS, Employer Costs for Employee Compensation, private industry by occupational group). One caution if you apply that share to a salary rather than to an hourly rate. Paid leave sits inside the benefit share at 9.4 percent, and a salary has already paid for those hours, so add paid leave back to the wage share before you gross up or you will count the same leave twice.
How Do I Price the Rework if My Time Records Are Thin?
Approximate it, and say so in the file. Ask the reviewer for a considered estimate of extra hours per job, apply it to the number of jobs the hire touched, and mark the figure as an estimate. An honest estimate you label is more useful than a precise number you invented. Then fix the time capture, because the next version of this question comes around every year.
Is a Short Tenure Always a Bad Hire?
No. Some separations are the system working as intended, including the ones where the person is capable and the role was wrong. The cost is still real and still worth totaling, but the corrective action is different. Treat every short tenure as a hiring failure and you go hunting for better interview questions when the actual defect was the job.
When Should the Firm Stop Trying to Save It?
When the review gap has stopped closing over consecutive review cycles and the extra senior hours have become the plan rather than the exception. Coaching is worth real time when the trend is moving. Once the trend is flat, every additional month is a loaded month of payroll plus a reviewer's time, and both of those are on the list above.
Price the Last One Before You Open the Next Role
Go back through the last hire that did not work out and total the five lines on real records. If the payroll and review lines together dwarf the recruiting fee you were focused on, that finding should change what you fix, shifting the effort from the interview to the ramp, the written standard and the decision date.
Then take the number into your next capacity decision, because it belongs on both sides of it. The cost of hiring wrong is part of the cost of hiring, and any comparison that leaves it out is flattering the option you already prefer.
If the honest answer is that your firm cannot absorb another miss this season, test the work before you commit to a person. Accountably has made 30+ placements inside 20+ US firms since 2022, and the ramp on a firm's own software and procedures runs about 3 to 4 weeks.
The Free 40-Hour Proof Pilot is built for exactly this problem. A fixed 40-hour block of your own representative work, prepared on your procedures and put through the full review chain, so your reviewer grades real output before anything is riding on it. If a placement is not the right fit in the first 30 days, we replace them free. Don't trust us. Test us.