Recruiting Statistics 2026: 200 Applications per Hire, 39 Days to Fill, and Only 3% of Roles Filled from Inside

Most recruiting statistics pages are a pile of numbers with a vendor's name attached and no reference period. This one is built from four bodies of data that publish their methods: the BLS Job Openings and Labor Turnover Survey and Employment Situation for the market as a whole, SHRM's 2026 Recruiting Executives Benchmarking for what employers report, Gem's 2026 Recruiting Benchmarks for what actually happens inside applicant tracking systems, and the NFIB Jobs Report for the small end of the market that national aggregates flatten out.

Every figure below carries its source and reference period. Where we have computed something ourselves — and we have, six times, because some of the most useful recruiting benchmarks are not published by anyone — it is labelled as derived and the arithmetic is shown. Section 13 lists the widely quoted recruiting statistics we could not verify, so you know which ones to keep out of a board deck.

1. The market everything else sits on: 7.3 million openings, 5.1 million hires

Recruiting benchmarks only mean something against the market they were measured in. In July 2026 the United States had 7.3 million open jobs and made 5.1 million hires. Separations were also 5.1 million, so the labour market is running flat: employers are refilling, not expanding.

Source: BLS Job Openings and Labor Turnover Survey, July 2026 reference month, released 1 September 2026, seasonally adjusted. The annualised column is ours — a twelve-month projection of the monthly level, not a BLS figure.

The supply side, from the Employment Situation for August 2026 (released 4 September 2026): unemployment 4.1%, 7.0 million people unemployed, labour force participation 61.6%, and payrolls up 162,000 in the month against a prior twelve-month average of just 31,000. Long-term unemployment is the uncomfortable number — 27.0% of unemployed people have been out of work 27 weeks or more, some 1.9 million people.

That last line is the one worth keeping. Recruiting volume is not primarily a growth activity. At the national hire rate a 1,000-person company runs roughly 374 hiring processes a year, and 227 of them exist because someone resigned. Every retention decision is therefore a recruiting budget decision, which is why this page leans on our employee turnover statistics and employee tenure statistics throughout.

2. Time to fill: 39 days, 43.6 days, or 63.5 days?

Three credible sources give three different answers, and the gap between the smallest and the largest is 63%. This is the most misquoted metric in recruiting, so it is worth being exact about what each one measures.

Two conclusions follow. First, 43.6 sits far closer to SHRM's 39 than to Employ's 63.5, so the employer-reported benchmark is the better-calibrated of the two for a general role. Second, applicant-tracking numbers run systematically longer because a requisition is opened in the system before sourcing starts and closed only after an offer is accepted, whereas survey "time to fill" is whatever the respondent's own report happens to define. Quote the number with its definition attached or it means nothing.

Benchmark a 55-day time to fill against SHRM's 39 and you are failing. Against Employ's 63.5 you are winning. Same company, same month, opposite conclusion.

3. Cost per hire: $5,475, and what it buys

SHRM's cost-per-hire definition is broad, which is why its figure is higher than most internal ones: it includes agency fees, advertising, job fairs and job boards, referral payments, candidate and staff travel, relocation, recruiter pay and benefits, and applicant-tracking system costs. If your internal number is $1,200, check whether it contains recruiter salaries. Usually it does not.

The $4,700 figure is the one you will see most often online, and it is several benchmarking cycles old. The 2026 cycle publishes the direction of executive cost per hire but not the level, so the most recent quotable levels remain the 2025 pair above.

4. The funnel: 200 applications to produce one hire

Gem's 2026 benchmarks are the most useful funnel dataset published this year because they are observed rather than reported: 165 million applications from 15 million candidates, resulting in 1.2 million hires, covering June 2021 to May 2025. The shape is brutal.

Company size changes the shape more than most people expect. At small companies 25% of applicants reach the pre-onsite stage but only 0.3% are hired; at large companies fewer than 10% reach pre-onsite yet 0.7% are hired. Small companies interview widely and hire narrowly; large ones screen hard up front and convert better afterwards. If you are borrowing a funnel benchmark, borrow one from your own size class.

5. The application flood: 2.4× more applications per hire than 2021

The defining recruiting fact of 2026 is not that hiring is slow. It is that the volume of applications has decoupled from the volume of hiring. Gem's panel, indexed to 2021, gives three numbers that let us quantify it: hiring volume is 30% below 2021 (though up 8.3% year-on-year), recruiting headcount is down 14%, and applications handled per recruiter are up 93%.

This single ratio explains most of the rest of the year's recruiting complaints. Candidates report sending more applications and hearing less back; recruiters report drowning while their requisition count barely moves. Both are describing the same denominator problem. It also means any conversion-rate benchmark older than about 2022 is describing a different funnel — a 2019 pass-through rate is not a target you can hold a team to in 2026.

6. Recruiter workload: 13.4 requisitions each, and 67% more at the top end

Two independent sources agree that load per recruiter rose sharply, which makes this one of the more reliable findings on the page.

The interviews figure deserves a second look, because it cuts against the hiring slowdown. Interviews per hire are up a third while hiring volume is down 30%, so total interview volume is running at 1.33 × 0.70 = 0.93 of the 2021 level — within 7% of peak-hiring interview load, on 30% fewer hires. Interviewing is the one activity that did not scale down when hiring did. If your hiring managers say they are in interviews constantly despite a hiring freeze, the data agrees with them.

7. Channel yield: referrals convert at 11×, internal moves at 32×

Gem's most quotable finding is the spread in conversion between sourcing channels. Measured per application, against inbound applicants as the baseline:

Sourcing's leverage shows up in the totals too: direct sourcing produced 11% of hires from just 2.6% of applications, a roughly fourfold yield advantage on volume share, and Gem reports sourced candidates as nearly eight times likelier to be hired. Candidate rediscovery is now the bulk of that work — 46% of sourced hires came from candidates already in the database, up from 26% in 2021. The most productive sourcing channel in 2026 is the pile of people who applied to you in 2023.

Referrals at 11× are the channel most employers can act on immediately, and the economics are unusually clean: the cost is a referral payment rather than an agency fee, and the conversion advantage is per application rather than per pound spent. Our employee referral program statistics covers the participation and payout data, and employee referral bonuses covers how to structure the reward so it is claimed rather than ignored.

8. The contradiction at the centre of 2026 recruiting

Put SHRM's fill-source data next to Gem's channel yield and you get the most interesting finding on this page. SHRM reports that the median share of non-executive positions filled externally rose from 93% in 2025 to 97% in 2026. That means internal fills fell from 7% to 3% — a 57% reduction in one year. Meanwhile internal candidates convert at 32 times the rate of inbound applicants.

Eight internal applicants, or 244 external ones. In 2026 employers chose the 244 for 97% of their non-executive roles.

9. Small employers: 35% cannot fill, 18% see no qualified applicants at all

National aggregates hide the small end of the market, where most hiring processes actually happen. NFIB surveys its members monthly; these are the August 2026 readings.

The 47% splits into 29% reporting few qualified applicants and 18% reporting none at all. Among owners actually hiring or trying to hire, that 47% is 82% of them. The unfilled-openings reading of 35% was down a point on July but remains 11 points above its historical average, and 56% hiring or trying to hire was down five points on the month — softening demand, persistent difficulty.

10. Offer acceptance: 82%, and what the other 18% costs

An 82% acceptance rate means the median employer runs 1.22 offer cycles per hire. It is the cheapest metric on this page to improve, because everything upstream of the offer has already been paid for by the time a candidate declines.

Per 100 hires, an 82% acceptance rate requires 122 offers. At 87% it requires 114.9 — so every five points of acceptance removes about seven full offer cycles per 100 hires, each of which carries the full cost of sourcing, screening and interviewing a finalist. At the 13.4 requisitions per recruiter reported above, those seven cycles are most of a recruiter's week.

11. AI in hiring: what is actually published

This is the cluster where recruiting statistics are least trustworthy in 2026, so it is worth separating what is verifiable from what is circulating.

Verifiable. SHRM's State of AI in HR 2026, published 31 March 2026, draws on 1,908 HR professionals and reports which HR functions AI has reshaped most. That sample size and date are on SHRM's own page. The report's function-by-function adoption percentages are behind membership.

The mechanism matters more than the percentage anyway, and section 5 measures it without needing an adoption survey: applications per hire are up 2.4× since 2021. Whatever share of employers has adopted AI, candidates clearly have — and the load landed on the screen stage, where pass-through stayed at 8%. For the adjacent data on HR systems and automation, see our HR technology statistics and AI in HR and rewards automation guides.

12. Benchmark cheat sheet

Everything on this page that is safe to quote, with its source and reference period. Derived rows are ours and their arithmetic is in the section named.

13. Recruiting statistics we could not verify

Every number in this section is widely republished and we could not trace any of them to a primary source with a stated method. They are listed so you can recognise them, not so you can quote them.

14. Methodology and sources

Where sources disagree we say so rather than picking the friendliest number. Reference periods are given because recruiting data ages quickly: a benchmark from the 2021–22 hiring peak describes a market that no longer exists.

Citation. If you use these figures, please cite the underlying source rather than this page and link here as the compilation. The six derived metrics — openings per unemployed person and the annualised hire and quit rates, the 43.6-day JOLTS-implied vacancy duration, the $2.05 million recruiting cost per 1,000 employees, the 244 applications per accepted hire, the 2.4× application inflation since 2021, and the eight-internal-applicants-versus-244 comparison — are original to this page and should be attributed to Rewordin together with their stated inputs and assumptions.