Cost per hire is the total spent to fill a position divided by the number of hires: agency fees, job ads, tools, and the fully loaded time of recruiters and hiring managers. In 2026 three models compete, success-fee agency, in-house team and subscription-based AI agent, and the same role does not cost the same depending on the model.
A Head of Talent Acquisition receives an agency invoice: 20% of a €65,000 salary, €13,000 for one hire. In the same quarter, the in-house team filled twelve roles and nobody can say what each one cost. Finance is asking. This grid answers, line by line.
How do you calculate cost per hire?
Cost per hire = (external costs + internal costs) / number of hires over the period. The hard part is not the division, it is the numerator.
External costs are visible: agency or headhunter fees, sponsored ads, job boards and resume databases, sourcing tool licences, assessment fees. They sit in the accounts.
Internal costs are the ones that get forgotten: recruiter time, hiring manager time in interviews, HR time on admin. To value them, start from the fully loaded salary and bring it back to an hourly rate. In the Umamy benchmark, a recruiter paid €42,500 gross per year in the Paris region (Michael Page average, France) represents roughly €57,000 fully loaded with French employer contributions, a little over €35 per hour on 1,600 working hours. That hourly rate feeds every line of the grid.
One rule: count per successful hire, not per open role. An abandoned search or a failed probation period weighs on the next hires.
Model 1: the success-fee agency
The agency charges a percentage of the hired candidate's annual gross salary, typically around 20%, more for executive search; 20% is the central assumption of the cost per hire benchmark.
The grid must add what the percentage hides: the internal time that remains (briefing, shortlist reviews, final interviews, often 10 to 15 hours per role), the limits of the guarantee (the replacement is free, the delay of the new search is not) and the absence of capitalisation: the profiles approached stay with the agency, the next role starts from zero. The model is rational for a rare, urgent role. It gets expensive as soon as the same type of role comes back several times a year.
Model 2: the in-house team at fully loaded cost
In-house hiring looks free because salaries are already budgeted. The grid still has to count every hour.
For a standard role, the typical lines are: writing and posting the ad (3 to 5 hours), sourcing and outreach (15 to 25 hours), screening and phone qualification (8 to 12 hours), manager interviews (6 to 10 hours), offer and follow-up (2 to 3 hours). These ranges are orders of magnitude to replace with your own data: your ATS has the dates, time three recent hires.
At €35 per loaded hour, 40 recruiter hours are €1,400. Manager time is often the most expensive line: a manager on €80,000 gross is close to €70 per loaded hour, and ten hours of interviews cost €700. Then come the fixed costs to allocate: ATS, LinkedIn Recruiter licence (a declared median of €10,153 per year for a Corporate seat in France, according to the 2026 tool pricing survey by Les talents narratifs, 22 October 2026; see our LinkedIn Recruiter pricing breakdown), job boards, divided by the year's number of hires.
The hidden cost is capacity: a recruiter carrying fifteen open roles stretches every timeline, and each week of vacancy has a cost of its own.
📄 Compare the three models on your own roles without rebuilding the spreadsheet. The cost per hire benchmark uses the assumptions in this guide (20% agency fee, loaded recruiter cost, AI agent subscription) and shows how they move with your yearly volume. Full benchmark, 15-minute read.
The agency, in-house, AI agent benchmark
Model 3: the AI agent on subscription
The third model is recent: an AI sourcing agent that finds profiles, ranks them against a scorecard and contacts candidates on behalf of the hiring manager, for a fixed monthly subscription. At Umamy, pricing starts from €249 per month, excluding VAT, with no placement fee.
In the grid: monthly subscription multiplied by the length of the search, plus the remaining internal time. That time does not drop to zero: writing the brief and scorecard (about an hour), reviewing proposed profiles (a few minutes a day), running interviews, which stay identical across all three models. Two effects deserve their own line: the licence covers a role type, not a hire (three technicians hired on the same search divide the subscription by three), and context accumulates: criteria, reasoned rejections and candidate replies stay in your system for the next role. The limit mirrors the agency's: for a very rare executive role, a headhunter still makes sense.
The full grid, role by role
For each role or role family, the grid has the same columns whatever the model. A worked example, presented as an illustration and not a statistic, for a €45,000 gross role filled in eight weeks:
| Cost line | Success-fee agency | In-house team | AI agent on subscription |
|---|---|---|---|
| Fee or subscription | €9,000 (20%) | €0 | €498 (2 months at €249) |
| Ads and job boards | €0 | €300 | €0 |
| Sourcing tools (share) | €0 | €200 | included |
| Recruiter time (€35/h loaded) | €350 (10 h) | €1,400 (40 h) | €350 (10 h) |
| Manager time (€70/h loaded) | €700 (10 h) | €700 (10 h) | €700 (10 h) |
| Indicative total | €10,050 | €2,600 | €1,548 |
The amounts depend entirely on your assumptions: agency rate, time actually spent, number of roles covered by one licence. The point of the table is to force every line to exist: an agency does not erase manager time, an in-house team is not free, a subscription does not remove interviews. Two lines are often missing: the cost of vacancy and the cost of a failed hire, detailed in the real cost of a bad hire. Both weigh on whichever model is slowest.
How many hires a year tip the decision?
Volume decides. A 20% agency costs the same on the first and the tenth role. An in-house team has a fixed cost that amortises with volume, until it saturates and timelines stretch. Saturation is more common than it used to be: according to Greenhouse, cited by the French newsletter Sourceurs non peut-être on 8 November 2026, recruiting teams averaged 4.6 recruiters in 2025 against 10.4 in 2022, for an average of 60 days per hire. A subscription costs the same whether it serves one or five hires of the same type in the year.
A practical method: sort the year's roles into three families. Rare roles, where an agency can be justified; recurring roles, where in-house and AI agent compete on time recovered; volume roles, where the subject is screening inbound applications. The grid fills in family by family.
What we've seen
In October 2026 we met the technical leadership of a business services group with around 400 branches across France and Europe. Their hardest roles are workshop managers and mechanics: turnover of about 15% means close to 160 departures a year on those two jobs alone, for searches that take three months on average and sometimes six to eight. Regions occasionally use agencies on a success fee of 15 to 20% of annual salary, paid from regional rather than central budgets. When we asked what a vacant workshop manager position cost, the answer was clear: nobody tracked it. The first useful deliverable was not a tool, it was the grid.
Frequently asked questions
What is the difference between cost per hire and cost of a vacancy?
Cost per hire is an average over a period: total spend divided by the number of hires. Cost of a vacancy is what an unfilled role costs the business each week it stays open: lost output, overtime, temp staff. The first serves the budget, the second the role-by-role decision.
Should hiring manager time be included in cost per hire?
Yes. It is often the highest hourly line and the most forgotten one. Ten hours of interviews for a manager on €80,000 gross represent roughly €700 fully loaded, whatever hiring model you use.
Does an AI agent replace the recruiter in the calculation?
No. It takes the sourcing, first outreach and screening line; the recruiter keeps qualification, manager coordination and the decision. In the grid, recruiter time goes down, it does not disappear.
Where Umamy sits in this grid
Umamy, based in Paris, is an AI agent for sourcing and application review. It starts from a scorecard written with the hiring manager, ranks profiles with a Dynamic Fit Score justified criterion by criterion, and contacts candidates on behalf of the founder or manager, never an agency; nothing is sent without approval. The system self-calibrates from your team's accept and reject decisions (context-enriched scoring). Across 11,010 people contacted, one in two replies; first candidates arrive ten minutes after the brief; the median time to hire is five weeks. Fixed pricing from €249 per month, excluding VAT, with no placement fee, on our pricing page. Umamy is currently filing for AI Act and GDPR compliance.
📄 Before you defend your budget, price the three models on your real volume. The cost per hire benchmark gives the assumptions, the formulas and the sensitivity to the number of roles per year.
The agency, in-house, AI agent benchmark
Sources
- Michael Page, France salary survey, average recruiter salary in the Paris region (€42,500 gross), used in the Umamy cost per hire benchmark, October 2026.
- Les talents narratifs, “Le prix des outils 2026”, Laurent Brouat, 22 October 2026 (166 French respondents).
- Sourceurs non peut-être, issue 295, Nicolas Darcis, 8 November 2026, citing Greenhouse.
- Umamy, customer proof points, October 2026.



