A decline at final stage is the most expensive outcome in hiring. You have spent the sourcing, the screening and the interview panels, you have taken your engineers away from delivery to sit on those panels, and you finish with nothing. Worse, you usually finish without knowing why, because the reason a candidate gives on the phone is rarely the reason they decided.
This is what the current data actually shows, including two numbers the recruitment industry repeats constantly that nobody can source.
The baseline: most offers are accepted
Gem's 2026 Recruiting Benchmarks Report, drawn from more than 165 million applications and 1.2 million hires, puts offer acceptance at 82 per cent across all roles and industries, the highest level since 2021. Engineering and technical roles in the same dataset average 62 days from open to hire.
So roughly one offer in five is declined, and that is normal. A hiring manager whose acceptance rate is 80 per cent does not have a problem. A hiring manager at 60 per cent has a process problem that will not be fixed by raising the offer.
The Australian picture is tighter than the global one. Robert Half's 2026 Australian research reports that 88 per cent of employers lost at least one candidate over the past year. The reasons employers gave for candidates declining were uncompetitive compensation at 31 per cent, poor job alignment at 29 per cent, and inflexible working models at 29 per cent. Looking forward, the barriers those employers expect in 2026 are a lack of skilled applicants at 61 per cent, salary expectation mismatches at 58 per cent and competing offers at 58 per cent.
Read those three decline reasons together and the shape of the problem appears. Only the first is about money. The other two are about whether the role was described accurately and whether the working arrangement was stated honestly, and both of those were settled weeks before the offer was drafted.
Two numbers to stop quoting
If you have been given a reason to move fast or to fear a counter-offer, you have probably been given one of these.
"Eighty per cent of people who accept a counter-offer leave within six months." This appears across recruitment blogs, agency pitch decks and LinkedIn posts, usually with no citation, occasionally attributed to a body that has no such study. Some versions say six months, some say twelve, some cite ninety per cent. It has been challenged directly by people inside the industry who went looking for the original research and could not find it. We do not use it, and you should treat any supplier who leads with it as quoting folklore.
"Top talent is off the market in ten days." Same problem. It circulates through statistics round-ups citing each other, with the trail going cold at aggregator sites rather than at a study. It also sits awkwardly beside the same reports' median time to hire of 41 to 44 days, which would mean nearly every hire made is somebody who was, by the ten-day rule, no longer top talent.
There is a real point buried under both, which is that good engineers in a live search do have competing conversations and slow processes do lose them. That point does not need an invented number, and using one is how a supplier loses credibility with the first person in the room who checks.
The reason a candidate gives on the phone is rarely the reason they decided. It is the most polite true thing available.
Money is the reason given, not usually the reason
Compensation is the cleanest thing for a candidate to say. It is impersonal, it does not insult anyone, and it ends the conversation. That is exactly why it appears at the top of every survey of declines, and why it should be read carefully rather than at face value.
When compensation genuinely is the cause, it is nearly always a failure of sequencing rather than budget. Nobody discovers at offer stage that they cannot afford a candidate they screened six weeks ago, unless the range was never put in writing. Robert Half found 99 per cent of Australian employers were open to negotiating on salary and 97 per cent were willing to offer alternative benefits, with more flexible working the most common at 56 per cent, a higher performance bonus at 55 per cent and a one-off signing bonus at 28 per cent. An employer that flexible does not lose people on money. It loses them on finding out too late.
The market is also moving under this. Research on pay transparency consistently finds that advertised ranges increase both application volume and the quality of match, and a 2026 study in the Journal of Applied Psychology analysing close to ten million postings found that very wide ranges create their own problem: they leave candidates uncertain what they will actually be paid, and women disproportionately avoid roles advertised with wide bands. A range is only useful if it is narrow enough to be a commitment.
Process speed is a decline reason that never gets recorded as one
Nobody declines an offer saying the process was slow. They decline because they took the other one, and they took the other one because it arrived first.
Talent Board's CandE benchmark research, which has surveyed hundreds of thousands of candidates across more than a decade, finds that poor communication alone is enough to make 47 per cent of candidates withdraw, and that only about one candidate in four is satisfied with the recruitment process overall. Industry survey data for 2026 puts candidate drop-out ahead of sourcing as the single biggest recruiting challenge for 27 per cent of employers, with the heaviest losses at the late interview stages and in the gap between acceptance and start date.
For engineering specifically, the assessment stage is where processes bleed. Long take-home exercises were already a poor trade for candidates with options, and they have become harder to justify now that AI assistance has degraded their signal: a 2026 Karat survey of 400 engineering leaders found 71 per cent believe technical skills have become harder to assess, with take-home signal rated as degrading fastest. An eight-hour take-home now costs you strong candidates and tells you less than it used to.
Flexibility is a hard constraint, not a perk
For senior engineers, the working model is frequently a threshold rather than a preference, and it is one of the few decline reasons that cannot be negotiated at offer stage because it is a policy rather than a number.
Research on engineering populations has found around a fifth would leave outright rather than return to a full-time office, with roughly half saying they would begin looking. Remote-eligible roles attract a disproportionate share of the strongest candidates, and multiple studies of firms imposing strict return-to-office mandates have measured materially higher attrition afterwards, concentrated among senior people and high performers, which are the two groups you are usually hiring.
None of this is an argument for remote work. It is an argument for stating the arrangement in the advert, in the first call, and in writing before the panel, because a candidate who learns your policy at offer stage is a candidate you spent six weeks on for nothing.
What we see in our own searches
Our Q2 2026 Tech Hiring Pulse is built on our own live briefs rather than survey aggregation, and it is a small sample by design: one quarter of one firm's searches. Two things in it are worth reporting anyway, because they point the same way as the wider data.
The median from intake call to signed offer was 21 days across senior individual contributor placements, excluding executive searches. The spread matters more than the median. Briefs with the interview loop agreed and the band calibrated at intake closed in the low teens. Briefs that re-opened the level or the range partway through ran past 35 days, which is the pattern in every one of the causes above.
On counter-offers, roughly one in four of our Q2 candidates received one at resignation, and none accepted. That is a small number of people and we are not presenting it as a rate. What it suggests is narrower and more useful than the folklore statistic: a counter-offer is an attempt to fix a reason for leaving, and where the reasons were surfaced honestly before the resignation, there was nothing left for it to fix.
Where the decision is actually made
Almost everything above resolves to the same thing. The final-stage decline is the visible end of a decision made much earlier, usually in the first conversation and the second interview.
Four things move acceptance more than anything you can do once the offer is drafted:
- Put the band in writing before the first technical interview. Not a gesture at a range. The number you would actually pay this person at this level.
- State the working model in the advert. Days in office, which office, and whether it is under review. If it is under review, say that too.
- Book the whole loop up front. Interviewers with time already blocked, and an approval chain agreed before the role opens, not discovered after the panel.
- Introduce a peer, not just the panel. The questions a candidate will not ask a hiring manager are the ones that turn into a decline, and they will ask a future teammate.
If you have a live offer about to go out, the Offer Acceptance Check scores it against the reasons above and names the one costing you most. It runs in your browser and nothing is sent to us.
And if your acceptance rate has moved and you cannot see why, that is usually diagnosable from the shape of the process rather than the offers themselves. It is worth a conversation.