What a recruitment agency replacement guarantee actually costs you

This article describes how US accounting standards treat two kinds of contractual promise. It is not accounting or legal advice, and which treatment applies to your agreement depends on how your clause is written. Take the questions at the end to your accountant.

Almost everyone offers a guarantee. In the only survey anyone has run on this, a 2019 poll of one US recruiting network with no published sample size, 95.9% of respondents said yes. Of those, 61.4% promise a replacement and keep the money, while 8.4% promise the fee back.

Those two clauses are not variants of one thing. On a $22,000 fee, at the same fall-off rate, the refund version costs between four and seven times what the replacement version costs, and where in that range you land depends on what you decide an hour of your own time is worth. Nothing measures how many recruiters have worked out which one they signed. What is measurable is that the two cost very different amounts, and that the difference does not show up anywhere in how the market talks about guarantees.

On a normal desk that works out to a couple of hundred dollars a placement if you replace, and around a thousand if you refund. Here is how to price yours. Sources checked August 2026; the market-share figures underneath come from a 2019 survey and are the most recent that exist.

What you actually agreed to

Four structures are in circulation. ROARK, an executive search firm that offers a one-year guarantee, sets them out clearly: replacement of the position at no cost, a full refund of the fee, a pro-rated refund based on how long the person stayed, and a pro-rated replacement priced against a new fee with a discount for time served.

ROARK also calls a guarantee "like an insurance policy for your hire," which is the right metaphor written for the buyer. That page is addressed to the client, so it describes the cover received. What nobody writes down is what the cover costs the party underwriting it, and that party is you.

ROARK lists two exclusions it treats as normal: it will not replace when the role has been restructured into a different job, and it will not place someone back into a situation involving harassment, discrimination or a hostile environment. Whether those are market-wide is not measured anywhere, so read them as one firm's terms rather than a standard. If your own agreement is silent on either, you have agreed to something broader than ROARK has.

The one survey everyone quotes, and how old it is

Ask where "90 days is standard" comes from and the trail runs short. The measurement everyone reaches for is one survey: Top Echelon polled its own recruiting network and published the results on 18 March 2019.

Read the sample before you lean on the numbers. Respondents were members of one split-fee network, agency recruiters only with corporate recruiters excluded, averaging fifteen years in the profession. No sample size was published, and the survey has not been repeated in the seven years since.

That matters for what you do next, not for whether Top Echelon did good work. If you are matching your terms to "the industry standard", the standard you are matching is one measurement of one network from 2019, and it says nothing at all about your own clients or your own screening. It is a fine place to start and a poor place to stop, which is the argument for the rest of this article.

This is a liability before it is a loss

Here is the part that reframes the clause, and it depends on which structure you picked.

If your guarantee returns money, you are holding consideration you may have to give back. The rules that specialists point to are the refund liability provisions of ASC 606 at 606-10-32-10 and 606-10-55-22 through 55-29. Accounting commentary on those provisions holds that they reach refundable services and not only returned goods, though we are relying on a secondary reading rather than on the codification text itself.

If your guarantee replaces the person and keeps the fee, no refund is in play, and the framework accountants reach for is the warranty guidance instead. ASC 606-10-55-30 splits warranties into those that assure the product meets what was agreed, and those that add "a service in addition to the assurance." The standard gives three factors for telling them apart at 606-10-55-33: whether the law requires the warranty, how long the coverage period runs, and "the nature of the tasks that the entity promises to perform." That third one is where a recruiting guarantee is actually decided, and the standard states it as a question rather than a test. A warranty in the first category is not a separate performance obligation and is accrued as a liability under ASC 460 when the work is delivered. One in the second category is a performance obligation with part of the fee allocated to it.

One caution on that second path, and it is why this section ends with a question rather than an answer. The warranty guidance is written about products, and we could not find published confirmation that it reaches a re-performed service the way the refund rules explicitly do. A promise to run the search again is not obviously the same object as a promise to repair a machine. That is exactly the kind of question your accountant answers and an article should not.

What both paths share is timing. Under either one the obligation attaches at the placement, not at the resignation. That does not mean every desk must book something today: a firm on a cash or tax basis has no ASC 606 liability at all, and even under GAAP an assurance-type warranty is accrued only when a loss is probable and can be reasonably estimated, which is hard if you have never measured your fall-off rate. The point is narrower and it survives all of that. The guarantee is a cost of the placement, and treating it as free until it fires is a decision rather than an accounting position.

Three numbers set the price, and only one is published

Expected cost equals your fall-off rate, times the cost of the remedy, times the probability the client actually invokes it.

The remedy you can calculate today, and it has two inputs you choose rather than look up. The first is how long a replacement search takes. This article uses 90 hours of delivery work end to end, which is a planning figure and not a measurement; no published benchmark exists for it either. It is the larger of the two assumptions and every dollar below is multiplied by it, so halve it if your searches are quicker and the replacement clause gets cheaper in proportion. The second is what an hour of yours is worth. Glozo's US recruitment industry salary report puts the median for the profession at $72,800. Add 25% for employer taxes and benefits and divide by a 2,080-hour year and you get about $44, which this article rounds to $45. So the remedy costs roughly $4,050 in your own time. A full refund costs the fee, and you have already spent the hours.

That $45 is a choice, not a fact, and it moves the answer. Price your hour at the unloaded median of $35 and the remedy falls to $3,150. Price it against what a billable hour actually earns you and it rises. Use your own number.

The invocation probability is a judgment about your clients and your relationship with them, and the trainer Terry Petra put the useful version of it in one line: anytime a guarantee needs to be implemented, everyone has already lost. Plenty of clients absorb a fall-off rather than open the conversation. Plenty do not.

The fall-off rate is yours and nobody has published a credible benchmark for it. There is no industry figure for how often placements collapse inside a guarantee window. What circulates is vendor blog assertions with nothing behind them, and importing one of those into your model would defeat the purpose of building it. In fairness, the pay figure this article uses is our own published data, and you should treat it the same way: a starting point to replace with your own number, not a fact.

Glozo's market data is built on 10M+ job market signals processed monthly, sourced from 30+ public and partner data signals.

How to measure your own rate

Pull three years of placements. Count how many triggered the guarantee, whether or not the client pursued it. Divide. That is your rate, and it is the only one worth using.

Three years matters because at realistic volumes a single year gives you a number made of two or three events. Count triggers rather than claims, because a client who quietly absorbed a fall-off still tells you something about your screening, and the next one may not absorb it.

Then track it forward by guarantee length. If you offer 90 days to some clients and 180 to others, they are different exposures and they deserve different rates.

What the clause is worth, at four rates

On a $22,000 fee, which is a worked example rather than a benchmark, with a replacement costing 90 hours at $45. This is the cost if the client invokes. The third factor, the chance they actually do, is deliberately left at one here, because it is the input you should argue with rather than one I can set for you. If half your clients absorb a fall-off quietly, halve every figure below.

The ratio holds at about 5.4 times across every rate, because the refund gives back the whole fee while the replacement gives back your hours. It is not a universal constant: at an unloaded $35 an hour it becomes 7.0 times, and at $55 it falls to 4.4. A pro-rated refund sits between the two structures. Twelve-month schedules are real: asked to describe their refund policy, one respondent in that survey wrote that they give a full refund within 90 days and prorate over twelve months thereafter. On a twelve-month ladder, a resignation in month two returns $18,333, and at an 8% rate that is $1,467 if invoked, or 6.7% of the fee.

Two things follow. The clause you offer matters far more than the period you offer it for, and the respondents offering full refunds are underwriting something several times more expensive than the ones offering replacement, and the survey did not ask whether any of them had priced it.

Refund, replacement, or a ladder

One caution before any of this. A signed agreement is not repriceable on your own, a contingency recruiter who refuses a term usually loses the role to the next agency, and the market-share figures above come from one 2019 survey of one network, which makes them a reason for you to price your terms rather than evidence to put in front of a client. Decide what you are willing to lose before you open the conversation.

If your fall-off rate is genuinely low and you can afford to carry a rare total loss, a full refund is a real differentiator and you should charge for it. If you do not know your rate, you cannot know whether you can afford it.

The pro-rated ladder is the structure that behaves best under uncertainty. It scales what you give back to what actually happened, so a resignation in week eleven does not cost the same as one in week two. It is also the easiest to explain to a client who is asking for more cover than you want to give.

Straight replacement is the cheapest thing you can offer and it was the most common answer in the one survey that asked. Whether those two facts are connected the survey does not say, but the first one is worth saying out loud when a client asks you to match somebody's refund terms.

When a client asks for six months

In that 2019 survey, 2.4% of respondents offered 180 days and 2.3% offered a year, with another 5.7% answering "other", so the true share beyond 90 days is somewhere above 4.7% and below 10.4%. Either way you are being asked for something uncommon, and there are three honest responses.

Price it. A longer window means more exposure, so measure your rate over that window and quote a fee that carries it. One agency-side cost analysis describes guarantee periods of 60 to 120 days as standard and says a longer guarantee makes clients more willing to pay a higher headline rate. That is a single vendor's read rather than a measured figure, but the trade it describes is real.

Trade it. A longer guarantee is worth something, so ask for exclusivity, a shorter payment term, or a volume commitment against it.

Or read it as information. A client demanding six months on a role where the market gives ninety days is telling you something about their retention, their manager, or the job itself. Ask why the guarantee matters to them before agreeing to anything. If the answer is that people keep leaving, a longer guarantee does not fix your risk, it transfers more of theirs onto you.

The fee mechanics behind all of this, including the four fee models and what they pay by role, are in our guide to recruitment agency fees. The related question of which engagement model is worth taking in the first place is in contingency versus retained recruiting. And if you are setting terms for a new desk, how to start a recruiting agency in the US covers where the guarantee sits in the wider agreement.

The four questions to take to your accountant

Ask which of the two treatments your clause falls under, and what that means for when the cost lands. Ask whether your guarantee is short enough and narrow enough to sit in the assurance category rather than being carved out of revenue. Ask what your measured fall-off rate implies for the provision. And ask whether the way you invoice matches the way the obligation is recognized, because for most desks it does not.

You will get better answers if you bring your own rate rather than an industry average, which is the practical reason to go and measure it.