The Cost of a Bad Hire: Calculate It From Your Own Numbers, Not From a Multiplier

Why should I ignore the standard multipliers?

A multiplier is a compression of some other company's situation into a single ratio. It carries assumptions about their seniority mix, their notice periods, their revenue per employee, and their local employment costs, none of which are stated when the number gets quoted. Applying it to your company reproduces their circumstances rather than measuring yours.

The ratio also fails in the direction that matters most. A misfit in a role with high leverage, such as someone who sets technical direction or owns a customer relationship, produces consequences that scale with the decisions they made, not with what you paid them. A misfit in a well-bounded role with clear supervision produces consequences closer to the wage bill. One ratio cannot describe both, and averaging them describes neither.

There is a practical reason to build your own number rather than borrow one: a figure you calculated will survive a conversation with your co-founder or your board, and a figure you read somewhere will not. When you propose slowing a search or spending on a better assessment process, the question you will be asked is where the number came from. Having the arithmetic on a page ends that conversation.

Which line items belong in the calculation?

Start with money that left the bank. Compensation paid across the period from start date to exit, plus employer contributions and benefits, plus any signing payment, relocation, or equipment that will not be recovered. Add any severance or notice pay that applied, noting that the statutory minimums differ by jurisdiction and by contract, so use the figure your own agreement and local law produced rather than a general assumption.

Next, add the hours. The original search consumed founder, hiring manager, and interviewer time. Then the underperformance period consumed management time in check-ins, documentation, and the difficult conversations. Then the exit consumed more. Convert hours to money using the fully loaded hourly cost of the people involved, which is their total employment cost divided by working hours in the year. Managers are expensive per hour, and this line is usually larger than founders expect.

Third, add the cost of doing it again: any external fees, the tooling and advertising spend, and a second round of the same interviewer hours. Finally, hold a separate line for output that did not happen, which is the hardest item and often the biggest. Keep it separate rather than folding it in, so that anyone reading your calculation can see which part is cash and which part is estimate.

How does the arithmetic work on a worked example?

Use placeholder figures and substitute your own. Suppose the annual package cost you 120 units of currency, that the person was in seat for seven months, and that employer contributions and benefits added 20 percent on top. Compensation paid is 120 divided by 12, times 7, times 1.2, which is 84 units. Suppose equipment and a relocation payment added 6 units that you will not recover, and notice or severance under your agreement added another 10. Cash out so far is 100 units.

Now the hours. Suppose the original search consumed 60 hours across you, the hiring manager, and four interviewers, the underperformance period consumed 40 hours of management time, and the exit process consumed 15 hours including internal admin and advice. That is 115 hours. If the fully loaded cost of the people involved averages 0.05 units per hour, the time line is 5.75 units. Then re-running the search costs another 60 hours, or 3 units, plus any external fee or advertising, say 15 units. Running total is roughly 124 units.

The interesting part is what that total tells you. Against a 120-unit annual package, the visible cost of the episode is roughly a year of that package before you have counted a single unit of lost output, and every input in it came from your own records. Change the tenure from seven months to three and the number falls sharply, which is the whole argument for deciding earlier. Change the interviewer count from four to eight and watch how quickly a heavier process pays for itself only if it actually improves the decision.

How do I put a number on the output I did not get?

Do not estimate lost output as a fraction of salary, because that just reintroduces a multiplier. Estimate it against what the seat was supposed to deliver. If the seat was created to ship a feature that would unblock a segment of customers, and it shipped five months late, the lost output is five months of whatever that segment was worth, which you can approximate from the deals that stalled or the churn you can attribute.

For a revenue seat the estimate is more direct: the pipeline the person was expected to generate or close over the period, discounted by a realistic ramp, minus what they actually produced. For an internal seat with no revenue line, express it in delay rather than money: the roadmap slipped by a quarter, and say what that quarter cost in terms of a decision you had to defer.

Include the drag on other people. When a seat underperforms, colleagues absorb the difference, and the work they stopped doing to absorb it has a cost as well. This is where a small team is hit hardest and where the number becomes genuinely difficult to pin down. Give it a range rather than a point estimate and state the assumption behind each end of the range, so that the uncertainty is visible instead of hidden inside a false precision.

Why does the same mistake cost more on a small team?

Concentration. On a team of eight, one seat is an eighth of the capacity, so the shortfall cannot be absorbed by slack elsewhere, and the redistributed work lands on people who were already at capacity. On a team of eighty, the same shortfall is diluted and the organization has enough redundancy to route around it for a while.

Standard-setting is the second effect and it compounds quietly. In a small team, what one person is allowed to get away with becomes the working definition of acceptable, because there is no larger body of counterexamples. Everyone recalibrates downward, and pulling the standard back up afterward takes longer than the original episode.

The third effect is founder attention. Managing a difficult situation on a small team consumes the specific person whose time is most constrained and least substitutable. That opportunity cost rarely appears in anyone's calculation and is frequently the single largest item. Add a line for it explicitly, priced at whatever you would say your own hour is worth to the business, even if the number feels uncomfortable to write down.

How should this change how long I wait before acting?

Most of the cash cost accrues linearly with time in seat, which means the decision date is the largest variable you control. Founders routinely spend three extra months hoping for improvement, and those three months are visible in your own arithmetic as a specific number. Look at it before deciding to wait, not afterward.

That does not argue for acting on the first bad month. It argues for setting a decision date at the start of the concern, writing down what would have to be true by that date, and then honoring it. The failure mode is not waiting, it is waiting without a defined endpoint, which converts a decision into a drift.

Also weigh the reverse error. Ending an employment relationship that would have worked out costs you a person who was learning, the cost of the search again, and a signal to the team about how much room they have to struggle. Both errors are real, which is exactly why a written decision date with written criteria beats acting on how the last week felt. And in every case the process itself is governed by local employment law and your contract, so take advice in the relevant jurisdiction before acting.

What reduces the number without slowing hiring down?

Shorten the gap between doubt and evidence. Most of the cost sits in the months between the first quiet concern and the moment someone gathers actual evidence. A structured thirty, sixty, and ninety day expectation set, agreed in writing on day one, converts a vague unease into an observable comparison and pulls the decision date forward without adding a single interview.

Improve the decision quality rather than the decision volume. Adding interview rounds adds hours to every search, including the ones that were going to work out, so it taxes the many to catch the few. Replacing one conversational round with a work sample tied to the actual job usually improves discrimination at no extra elapsed time, because it tests the thing the role is.

Finally, keep the records. If you cannot reconstruct what a search cost or how long a concern ran before anyone acted, you cannot compute any of this and you will be back to borrowing someone else's multiplier next year. A short written note at the close of each search and each exit is enough, and it takes minutes.

What to take away

• ▸Build the figure from compensation paid, hours consumed, re-hire cost, and lost output, using your own records.
• ▸Keep cash items and estimated items on separate lines so a reader can see which is which.
• ▸Value lost output against what the seat was meant to deliver, never as a fraction of the salary.
• ▸Price founder attention explicitly; on a small team it is often the largest single item.
• ▸Most of the cost accrues with time in seat, so a written decision date is the biggest lever you control.
• ▸Replace a conversational interview round with a work sample before adding rounds to the loop.