Everything that goes into cost per hire, including what nobody invoices you for.

The number everyone quotes, and the number next to it

Start with the figure every article on this subject repeats. In 2022 the Society for Human Resource Management reported that "the average cost per hire was nearly $4,700," from benchmarking data collected across 2021. The precise figure in the underlying Talent Access report is $4,683. Fewer articles quote the number one column over: the median in the same table was $1,244. That gap is not noise. Cost data gets dragged upward by a handful of expensive hires, which is why the median is the more useful of the two numbers when they diverge this far.

Executive hiring stretches the spread further still, an average of $28,329 against a median of $8,750 in the same 2021 data.

SHRM's 2025 Recruiting Executives Benchmarking data brief (SHRM, 2025) reports median cost per hire at $1,200 for nonexecutive roles and $10,625 for executives, and notes the executive figure has climbed 113 percent since 2017. For 2025, this article uses those medians rather than an average.

A benchmark is for calibration, not for targeting. Survey respondents do not all count the same things, so two firms with identical hiring can report numbers an order of magnitude apart. A team that counts only invoices lands near the median. A team that loads in recruiter salaries lands far above it. Neither one is lying. They are answering different questions through the same field on a form.

So the published figures run from a $1,200 median (SHRM, 2025) to a $4,683 average on 2021 data, depending on which statistic and which survey you pick, and that is before a single fact about your company enters the math. A spread that wide should prompt one question: what were the survey respondents counting?

The answer is where this article actually starts, because the definition has a gap in it, and the gap is the expensive part.

What the standard counts, and who it forgets

Cost per hire has an official formula. The ANSI/SHRM cost-per-hire standard, approved in 2012, defines it as total external costs plus total internal costs, divided by total number of hires. The two buckets do more work than the formula. External costs leave the building, so they carry invoices and land in a budget someone owns. Internal costs are mostly time, and time arrives pre-paid through payroll. No invoice, no line item, no owner.

A recruiting budget holder sees the job-board renewal to the dollar and never sees the forty hours of engineering time the same role consumed.

SHRM's benchmarking glossary spells out the components it counts: "the sum of third-party agency fees, advertising agency fees, job fair costs, online job board fees, employee referral costs, travel costs of applicants and staff, relocation costs, recruiter pay and benefits, and talent acquisition system costs divided by the number of hires." Read that list twice. Exactly one internal cost is named, recruiter pay and benefits. The hiring manager who writes the brief and sits in every final round appears nowhere.

Neither do the four engineers who interview each candidate, nor the coordinator who rebuilds the schedule when one of them travels. The official checklist stops at the recruiting team. The undercount lives exactly there.

The referral bonus is the instructive edge case. It is internal money, paid through payroll to one of your own employees, yet SHRM's list counts it, because a bonus has a number attached and lands in somebody's budget. That is the pattern in miniature. Components get counted when they generate paperwork, not when they cost the most. Interviewer time generates none.

Here is the component list with the missing rows restored.

The invoiced slice, priced from the vendors' pages

Build the external bucket from invoices, written quotes and dated vendor prices. The ATS figures below were checked October 18, 2026, and the LinkedIn pricing references October 21. Other vendor figures retain their stated August checks.

Applicant tracking runs from limited free plans to custom quotes. At October 18, 2026, Workable Standard at 1-20 employees shows $360 monthly or $299/month equivalent annually. Ashby Foundations shows $300 monthly at 1-10 employees or $270/month equivalent yearly, with higher workforce bands priced separately. BambooHR's cards and FAQ disagree on some per-employee rates; obtain the applicable scope in writing. Use our current ATS price table and blank-preserving quote workbook for all eight vendors, service commitments and unresolved costs rather than carrying an entry headline into a full cost-per-hire budget.

The category, tier by tier, is priced in our ATS pricing comparison.

Sourcing sits on top of the tracker. For LinkedIn Recruiter Lite, record the recurring monthly or yearly amount shown at checkout for your account, along with the country, currency and paid licenses. The English US pages checked October 21, 2026 do not establish a universal ongoing dollar price. Full Recruiter needs a sales quote; its official pricing page lists team size, region and contract terms among the factors. Our August 2026 checks recorded SeekOut Core at $149 a month paid annually, Juicebox Starter at $99 per seat per month billed yearly, and hireEZ at $494 a month for a solo recruiter. Those dated figures are historical checks, not new quotes.

The wider sweep lives in AI recruiting tools pricing in 2026, and LinkedIn's tiers get a LinkedIn Recruiter cost breakdown of their own.

Two budgeting notes on this slice. Quote-only pricing means the number in your plan came from a sales call, and it will not match your neighbor's, so treat any third-party figure for those tiers as a rumor with a date on it. And published prices move without announcement, which is why every figure in this section carries one. A price without a date is a screenshot of the past.

To budget the stack, record the ATS commitment and sourcing subscription separately. Use the full annual invoice for an annual agreement; otherwise multiply the recurring monthly amount by the number of paid months. Leave a missing Lite price blank until checkout confirms it.

Divide the known software subtotal by hires in that same period. If a subscription is still unpriced, label the result partial. Compare it with internal labor and other fees only after checking that they cover the same period and hiring scope.

The external line that actually moves totals is the agency fee. The commonly quoted market range is 20 to 30 percent of first-year salary, and our pricing page names 30 percent only as the top of that convention. The citable band is narrower, and it is the one our pricing page leads with: SHRM's guidance on working with search firms puts contingency fees at 20 to 25 percent of the role's first-year cash compensation, and standard retained-search fees at about a third of first-year pay (SHRM, 2016). On a $150,000 salary, and treat that salary as an assumption for arithmetic rather than a statistic, 25 percent is $37,500 for one placement.

One placement fee outweighs years of software. In between the two models, some vendors now blend them: Wellfound's Autopilot service runs $500 a month per open role plus a 10 percent placement fee on a successful hire (checked August 2026). When the agency trade makes sense, and when it is software money set on fire, is worked through in agency fee vs software cost.

The biggest line, and nobody invoices it

Now the internal bucket, which is where cost per hire actually lives for most teams. The method fits in one sentence: hours spent on the hire, multiplied by the loaded hourly rate of the people spending them.

Loaded means salary plus benefits, employer taxes, and overhead, which is why a loaded hour costs meaningfully more than salary divided by 2,080 working hours. The hours scale multiplicatively, because every extra candidate passes through every stage, and every extra stage touches every candidate, which is why the shape of the funnel sets the internal bill far more than the tooling does.

The loaded rate deserves a beat, because it is the input people lowball. A $150,000 salary is about $72 an hour across 2,080 working hours. Assume benefits, employer taxes, equipment, and overhead add a quarter to a half on top, an assumption rather than a published figure, and $72 becomes $90 or more before anyone books a room. Recruiter time follows the same math when the recruiter sits on your payroll.

SHRM's component list does count that line, and it is the one internal cost some teams already track.

Here is the method applied to one interview loop. Every input that follows is an assumption chosen to be round, not a statistic, and the point of the exercise is the shape, so swap in your own numbers. Say a role brings six candidates through a panel of four one-hour interviews. That alone is 24 interviewer hours. Say twelve candidates got a 30-minute screen upstream: six hours more. Debriefs at half an hour with four people per candidate add twelve. Give the hiring manager eight hours for the brief, resume reviews, and syncs, and a coordinator five hours of scheduling. The loop totals 55 hours before anyone drafts an offer. Price those hours at a blended loaded rate of $90, roughly where salaries around $150,000 land once loaded, and this one hire carries about $4,950 in time cost. Run cleanly.

No reopened search, no declined offer.

In this worked example, internal time is a substantial part of the budget. The assumptions are illustrative, not observations of a typical hire. Substitute your team's actual hours and loaded rates, and keep this estimate separate from SHRM's population benchmark.

The sprint that slipped while four engineers ran a panel does not appear in any hiring dashboard.

Two things inflate the invisible line fastest. The first is unqualified candidates. A candidate who reaches the panel and misses still consumes the full four interview hours plus debrief time, call it six senior hours to produce a no, near $540 at the same assumed rate. Screening is exactly the work that protects those hours, and it is also the work most teams do least well, because it happens in the evenings around everything else. The second is reruns. A declined offer or a mis-scoped role restarts most of the loop at full price, and the second pass books the same calendars all over again. If you want to check any of this against your own records, you probably cannot: the stage-by-stage time breakdown is one most teams never measure.

That is not a criticism, since there has never been a line item to hang the measurement on.

Time-to-fill multiplies everything above

Internal cost has a second dimension: how long the role stays open, which SHRM's 2025 data brief puts at a median of 44 days for nonexecutive roles and 45 for executives, counted in calendar days from requisition to accepted offer.

That is the clock the internal bill runs on, and how we try to move it is the subject of the interview-scheduling page. What that clock measures, and how to cut waiting time without deleting evidence, is in how to reduce time to hire.

A 44-day median is six weeks of open role, and six weeks of open role is six weeks of the meter running. Pipelines need warming, candidates on hold need chasing before they take other offers, panels need reconvening after memories of the last debrief fade, and the work the role exists to do either waits or lands on the same people doing the interviewing. There is also the cost of the empty seat itself, the output it fails to produce. We will not put a number on that one, because every published vacancy-cost figure depends on assumptions its publisher rarely shows. It is enough to say the direction: every added week multiplies the hours in the section above and adds drag nobody logs.

Scheduling deserves its own mention, because it is where calendars go to die.

Every interview in the loop above had to find a slot that four busy people and one candidate could all make, and every reschedule multiplies the coordination hours while the candidate cools toward some other offer. We publish one stage timing we can stand behind, and nothing finer than that, because the stage-by-stage record is exactly the thing this industry has never kept.

Compression is the point of our product, so read our figure knowing that. In 95 percent of roles, the first interview is scheduled within 24 hours of defining the role, where scheduled means a confirmed slot on your calendar, not a same-day meeting. The figure is company-reported and not audited. Separately, the five anonymized production accounts on our customers page show the funnel only up to a confirmed interview, so this article sets no time-to-fill figure beside SHRM's medians.

What actually cuts cost per hire

Rank the components by size and the order of operations writes itself. Cancel a tool and you save a few hundred dollars a month. Negotiate an agency down five points and you save real money on that one placement. The line that moves the total is internal time, and only two things shrink it: fewer unqualified candidates entering the loop, and fewer days on the calendar.

Every other move works on lines that were never large enough to explain the total, which is why cancelling subscriptions so rarely shows up in the number.

Pre-screening attacks the first. Every candidate filtered out before the panel returns those six assumed hours to the team, and a loop fed only with people worth meeting spends its hours producing decisions instead of producing polite rejections.

The catch is who does the filtering. Hand a team a better search tool and the screening hours do not vanish; they move to whoever operates the tool, which is usually the same hiring manager whose time was the problem. Scheduling velocity attacks the second, and it is the least glamorous work in hiring, which is why it compresses so well when software runs it.

This is the reasoning behind our meter. Metix AI sells software subscriptions rather than candidates: plans run $99 to $499 a month and include 5,000 to 27,000 monthly usage Credits plus 1 to 5 AI Agents. Work inside an active Agent's workflow uses 0 Credits, so Credits are not a per-candidate charge, and how Metix AI Credits are metered outside that workflow is laid out on its own page. Sourcing across 900M+ profiles, outreach you approve before it sends, and the first screen all happen inside that subscription, so your team's hours start at the interview instead of at the search. A staffed team that already operates a search seat should price its recruiters' hours beside the seat. Metix AI's case is the interviewer hours and calendar days its AI Agents remove, which the method above lets you price for your own team instead of taking our word.

How credits compare with seats is worked through in per-seat vs usage-based pricing, and the plans themselves are on the pricing page.

In practice the models mix. Teams with a staffed sourcing function keep their seats for pipeline work and point Metix AI's AI Agents at the roles with a deadline, because a quarter-end engineering hire needs different throughput from an evergreen funnel.

The arithmetic stays the same either way. Whichever model spends fewer of your interviewers' hours per accepted offer is the cheaper one, whatever the sticker says, because the sticker only ever prices the part of the hire that arrives with an invoice attached.

None of this requires believing our marketing. The component list is SHRM's, extended by three rows of payroll. The method is multiplication. Run both on your last hire and the biggest line will not be a subscription.