The counter-offer statistic everyone quotes is wrong

Probably not, and above all nobody knows. The figure the whole industry repeats, “80 to 90% of candidates who accept a counter-offer leave again within six to twelve months”, traces back to no study anybody can open. The two measurements that do have a source give 32% and roughly 50%, which is not a nuance but a different job: at 85% you dissuade; at 32% you prepare.

This article is one instance of a wider problem, and that is what makes it worth writing. A number an industry passes around without ever opening it ends up organising commercial practices, sales arguments and training courses, and nobody goes back to check because everybody heard it from somebody else.

Where does the number everybody quotes come from?

From nowhere you can verify, and the attempt to trace it is more instructive than the number itself.

One recruiter spent several years looking for the original study. The trails lead to an untraceable business-press article several decades old, to an American professional association whose site no longer exists under that name, and to a recruitment software vendor publishing the figure with no methodology, no sample size and no observation period. None of those three leads ends in data.

The clearest sign, though, is not the missing source but the instability. The same fact circulates under at least six different values depending on where you read it, 48%, 66%, 80%, 89%, 90%, 93%, and over two distinct horizons, six months or twelve. A measurement does not behave that way. A measurement has a value, a method and a margin of error, and it does not round upward every time somebody retells it.

I do not believe in organised dishonesty here. I believe in something more ordinary: the number suits everyone along the chain, it is useful at exactly the moment a recruiter needs it, and nothing in the job forces anybody to open it before using it.

What do the figures with a source actually say?

They say the counter-offer works better than the story claims, and that it is far more common than people think.

The Robert Half survey published in 2026 gives the fullest picture. 85% of employers made at least one counter-offer during the year to an employee holding an external offer. Of the employees concerned, only 7% turned it down. And among those who accepted, 32% left within the following twelve months. On the employer side, 39% see the tool as valuable and 20% as a short-term patch.

The other identifiable measurement is older and higher: CEB data, cited in the trade press in 2016, puts the share leaving within a year of accepting at around half. Its method is not fully public, and I quote it with that reservation rather than picking whichever number suits me.

Both agree on what matters. An accepted counter-offer is not a time bomb, it is a bet the employer wins most of the hands. Telling it otherwise means preparing to lose while believing you are going to win.

Why this comfortable number costs you money

Because it replaces a method with an argument, and because it no longer survives being checked.

A recruiter who believes in 90% departures treats the counter-offer as a mistake by the candidate, and their reflex is to reason them out of it: “statistically, you will be leaving in six months”. A recruiter who knows the figure is 32% understands they are up against a rational decision, and the work changes nature. It is no longer about convincing somebody they are wrong, it is about having built, across the whole process, a reason to leave that is not the salary.

The second cost is more recent and growing fast. A candidate who hears that figure checks it, and these days checks it by asking a conversational assistant, which will happily answer that the statistic circulates with no identifiable source. You have then lost two things: the argument, and the presumption of sincerity over everything else you said in the interview. This is the individual version of what we describe about content read by machines before it is read by humans, and the conclusion is the same. An unsourced number eventually gets turned against whoever used it.

There is a third cost, invisible on any dashboard. When a counter-offer is experienced as the candidate’s betrayal rather than as a predictable stage, the process never gets corrected. You start again identically on the next assignment, and lose the same candidate for the same reason six weeks later.

A counter-offer is prepared three weeks in advance

The only moment you can act is during the process, when the candidate is engaged but not yet decided, and that window is narrow.

What works fits into one conversation, held early enough not to look like a manoeuvre. You ask the candidate what they will do if their current employer offers more, and you let them answer for real, including if the answer is that they will think about it. Then you get them to name the non-salary reason for leaving, the one that existed before any offer arrived. A sentence the candidate formulated themselves in week three holds far better, on the day of the resignation, than the best argument supplied by a third party under pressure.

The rest is a question of calendar rather than persuasion, and this is the part the industry most underestimates. Counter-offer risk is proportional to the time between the verbal agreement and the signature: every week added to that gap is a week handed to the current employer. We have shown elsewhere that placement delay is played out in the waiting, not in the tasks, and the counter-offer is where that delay is paid in cash.

And if accepting were sometimes the right call?

It happens, and refusing to consider it is what makes the recruiter’s speech so easy to dismiss.

An employee whose only reason for leaving was pay, at a company they like, in a role they have mastered, makes a reasonable decision by accepting. The 68% still there after twelve months in the 2026 measurement are not all people who were wrong and have not yet admitted it; some simply got what they asked for without changing environment. A recruiter who concedes that gains something no sales argument will ever provide, and it pays off on the next placement.

What remains true, and can be said without an invented number: a counter-offer settles a question of pay and settles nothing else. If the real reason was the content of the role, the manager or the absence of a path, it will be intact in six months at a higher salary, and that sentence is what makes a candidate think, not a percentage.

What changes in your process, concretely

Three modifications, and none of them needs a tool.

The third-week conversation becomes a stage of the process, written into the tracking like a technical interview, with its date. The gap between verbal agreement and signature becomes a tracked indicator, because that is what carries the risk. And the number disappears from your sales arguments, replaced by what you have actually observed across your own placements, which is both more accurate and harder to dispute.

That last point deserves to be taken seriously rather than treated as moral posturing. Your data is worth more than a conference statistic: you know how many of your candidates received a counter-offer last year, how many accepted, and how many are still in post today. Nobody else can quote those three numbers, and they convince a candidate in a way no generic percentage ever will.

Those three numbers come out of your software in one query, and nobody pulls them, because nobody has time to do it at the moment the conversation happens. That is exactly the kind of thing we have Balt do: not an opinion on counter-offers, the figure you already own and never look at, said in the conversation where the question comes up.

That leaves the question that decides the rest, and that arises well before any counter-offer: how many of your candidates are waiting for an answer right now, and for how many days. It is the same mechanism at the other end of the process, and silence there costs more than bad news.