Last updated: August 3, 2026
JobFunnel compares your funnel against an industry benchmark. This page is where that benchmark comes from, how each number was reached, and what it cannot tell you. Every figure below is either imported from the running product or quoted from a linked source, so you can check it yourself.
Every rate in JobFunnel is cumulative from Applied: the share of your applications that ever reached a stage, divided by the number you applied to. An application that went applied → screening → rejected counts toward screening permanently. It is not a step-to-step rate, and the two are easy to confuse.
We show a verdict only once you have at least 10 applications. Below that the numbers move too much for a comparison to mean anything, so we show your counts and say how many more are needed. We also stay quiet when your rate is close to the benchmark: a gap of a few percent at these volumes is one application, not a finding.
Most published recruiting statistics count from the employer side: every application a company receives for one opening, divided into the few that advance. That denominator is now dominated by mass-generated volume — Ashby reports over 300 candidates per opening, roughly triple its 2021 baseline. It is why you will read that only ~3% of applicants reach an interview.
JobFunnel counts from the candidate side: the applications you deliberately sent. Those are different quantities, and mixing them produces a claim that is both wrong and easy to check.
So we benchmark only against sources that share our denominator. It is the reason the numbers below are what they are, and the reason we do not use the larger, more quotable employer-side datasets.
They were not produced the same way, and we label which is which rather than presenting both as measurements.
Applied → Screening: 10% ANCHORED · MEASURED
Anchored on Huntr’s Q1 2026 job search trends report (139,927 tracked applications from 25,635 job seekers). Their most selective cohort — people who sent 11 to 20 applications — reached a human conversation on 9.25% of them. Volume-matched cohorts below that: 6.96% for 21–50 applications, 4.65% for 51–100, 2.58% above 100.
We rounded up from 9.25%, and it is worth being explicit about the two reasons rather than presenting 10% as a measurement. Their data depends on people logging their own interviews, so it under-counts by an unknown amount. And our screening bar is lower than theirs — we count any recruiter contact, they count a logged interview — so more events clear ours. Both biases point the same way. We did not apply a multiplier, because neither can be quantified without inventing precision.
For context on the spread, Huntr’s 2025 annual report (a separate dataset — 598,627 applications) puts response rates by job site at 3.1% for LinkedIn, 4.5% for Indeed and 11.3% for Google Jobs, while their job-board study (1,244,654 applications) puts the strongest boards at 3.78–7.28%. Those are different studies over different periods and we do not blend them into a single figure — they are here to show that 10% sits at the optimistic end of everything published, not in the middle of it.
Applied → Interviewing: 5% DERIVED · CALCULATED
Not measured directly — calculated. No candidate-side dataset separates “a recruiter replied” from “I got a real interview”, because trackers log one event for both. The one published measurement of that gap comes from Di Stasio and Zwysen (2021), a UK field experiment across six occupations including software developer. It reported the same applications at two thresholds: of 725 applications from majority-background applicants, 178 (24.55%) drew any positive employer response and 97 (13.4%) drew a direct interview invitation. That ratio is 0.546, and 0.546 × 10 = 5.46, which we show as 5% because the product renders whole percentages.
As a cross-check: Ashby’s 35% recruiter-screen passthrough implies about 3.5, so 5% sits at the generous end of a 3.5–5.5 band. That is an employer-side source, which section 2 rules out for levels — admissible here because we take a ratio from it and never a rate. Ratios also age far better than levels, which is why a 2016–17 study still earns a role while setting no absolute number.
JobFunnel shows your offer count and your offer rate. It does not compare them against anything, because nothing credible exists to compare them to.
Field experiments stop at the callback by construction — nobody runs a study that accepts a job. Candidate-side trackers publish no applied-to-offer rate. The only figure available is employer-side (roughly one offer per 200 applications received), which is the wrong denominator for exactly the reason in section 2.
We could have picked a plausible-looking number. We had one: adjusting the employer-side figure gives about 1%. We dropped it instead. At 1% you would need 100 applications before the comparison could even be expressed, so it would have been an unsourceable number that also did nothing.
This is the part of the page we would most like you to notice. Removing a number we could not defend is a stronger position than showing three numbers with an asterisk on one. If we ever publish an offer benchmark, it will be because we measured it.
Stated plainly, because a benchmark presented without its limits is a worse product than one presented with them.
Searching for funnel conversion benchmarks returns a large number of sites publishing precise-sounding figures with no dataset, no sample size and no method, frequently attributing numbers to sources that do not contain them. We keep an explicit blocklist and do not cite any of them.
One example of why. A widely repeated trend — “applicant-to-interview was 15.25% in 2016, 8.4% in 2023, 3% in 2024” — is attributed to a real recruiting-software company. Their actual report publishes only the 2024 figure. The other two points do not exist. That is the standard of care this page is written against.
These are labelled industry benchmarks, never peer or cohort data, and they never describe JobFunnel users. That distinction is deliberate and we hold to it: nothing you see compared against your funnel today is drawn from anyone else’s account.
It is also the honest description of a limitation. The gaps in section 5 — geography above all — are ones no published source can close, and the only population whose denominator matches ours exactly is our own. If we ever replace these numbers with benchmarks of that kind, it will be announced here, covered by the Privacy Policy first, and labelled as what it is rather than quietly swapped in behind the same word.
If that day comes, the offer stage is what returns first. It is the number nobody publishes.
If you think something here is wrong, we would rather hear it than not. Every source above is linked so you can check the derivation yourself, and support@jobfunnel.eu reaches us. A correction to this page is worth more to us than the number it changes.
Related: Terms & Conditions · Privacy Policy