Learn › Recruiting unit economics
In-house TA vs agency — when one internal recruiter outperforms three agency seats
The build-or-buy question for recruiting in Japan tends to resolve through anecdote rather than arithmetic — someone heard about a company that built a great in-house team, someone else heard about one that built a team and wished they hadn’t. The actual answer is determined by hiring volume, role mix stability, and the unit-economic math comparing the loaded cost of an in-house TA hire against the placement fees the team would otherwise pay an agency. This guide derives the volume threshold honestly from production data.
The volume threshold where in-house TA beats agency in Japan mid-career hiring is roughly 25–35 placements per year per in-house TA hire. Below that, agency placement fees are cheaper than the loaded cost of an internal recruiter who isn’t running enough placements to amortize their salary. Above that, an internal recruiter beats agency on a per-placement basis. The threshold drops to ~12–18 placements per year if you add AI sourcing — the productivity lift compresses the loaded-cost-per-placement enough that in-house wins at lower volumes. The volume math is the question; everything else is framing.
What you're actually comparing
The agency-spend side is easy to compute: total agency placement fees over the trailing twelve months. Pull the invoices. Add them up. Don’t include retained search fees if those buy a different product (executive-level retained mandates are usually not the work an in-house TA hire would do). The result is the agency baseline.
The in-house TA side is harder. The right number is the fully-loaded annual cost of one TA hire — base salary, bonus, social insurance contributions, equipment, software licenses, allocated office cost, manager time, internal coordination overhead. In the Japan mid-market, this lands around ¥10–14M for a mid-level TA recruiter and ¥14–20M for a senior, fully loaded. The number that gets used in most build-or-buy conversations is the base salary, which understates the real cost by 30–50%.
Then divide the loaded annual cost by the placements per year that recruiter would actually deliver. This is where the math gets honest. A mid-level in-house TA recruiter without AI sourcing typically delivers 20–30 placements per year in Japan mid-career bands; with AI sourcing, 30–50. A senior delivers more on the strength of established relationships and faster qualification; a junior delivers fewer for the inverse reason.
The volume threshold, computed
Take a Japan mid-career placement as having an agency fee of roughly ¥4M (consistent with our cohort’s ¥4,266,675 average; your number depends on your role mix). Take a fully-loaded mid-level in-house TA recruiter at ¥12M. The breakeven volume is the number of placements where in-house cost-per-placement equals agency cost-per-placement.
= = 3.0 placements/year breakeven; with 25-placement productivity, in-house wins above year-3
That math is misleadingly simple. The breakeven on cost-per-placement is 3 placements per year — at 25 placements per year, the in-house TA delivers placements at ¥480K each vs the agency’s ¥4M, an 8.3× cost-per-placement advantage. That’s the headline number that makes build-or-buy look obvious. It’s not.
What it’s not capturing: the in-house TA is a fixed cost regardless of placement count. If your hiring volume drops to 8 placements in a year, the in-house TA cost is still ¥12M and the cost-per-placement jumps to ¥1.5M. Agency spend at 8 placements is ¥32M — still higher — but the agency spend was fully variable. If your volume drops to 3 placements, agency spend is ¥12M and in-house spend is ¥12M, breakeven. Below that, agency wins on cost.
Why volume stability matters more than peak volume
The hidden cost in the in-house TA model isn’t the loaded annual cost — it’s the cost of having that loaded annual cost during a hiring slowdown. Japan mid-market hiring is cyclical. A company hiring 30 placements in a strong year may hire 8 in a slowdown year. The in-house TA hire amortizes the loaded cost across both. Across a five-year window with two strong years and three average ones, the math works. Across three years with a slowdown in the middle, the math frequently doesn’t.
The thing to model isn’t "will we hit 25 placements next year." It’s "what’s the trailing-five-year average and the variance across years." If the average is at or above 25 placements and the worst year is at least 12, in-house wins on a long-run basis. If the average is 25 but the variance is wide enough that some years drop below 8, in-house has years where it costs more than agency would have.
What AI sourcing does to the threshold
AI sourcing changes both sides of the math. On the in-house TA side, it raises the placements-per-year that single recruiter can deliver — from 20–30 to 30–50 in our calibration data. The same loaded cost amortizes across more placements, dropping cost-per-placement. On the agency side, AI sourcing doesn’t directly change agency placement fees, but the agencies that adopt AI sourcing themselves typically pass some of the productivity gains through as smaller fee structures or as expanded scope at the same fee.
= = ¥375K cost-per-placement; in-house wins above ~12 placements/year breakeven
The threshold drops by half. A company that wouldn’t have justified an in-house TA hire at 18 placements per year — because pre-AI breakeven was at 25–35 — can justify the hire at the same 18 placements with AI sourcing. The hiring volume threshold for build-vs-buy is meaningfully lower in 2026 than it was in 2022.
What this means for hiring planning
Three rules of thumb that survive against production data. First, if your trailing five-year average is below 12 placements per year and your role mix is volatile, stay agency-only. The fixed cost of an in-house hire isn’t justified by the volume. Second, if your average is 12–25 with AI sourcing or 25–35 without, in-house wins on cost; the question becomes whether your role mix is stable enough to specialize the in-house hire (a recruiter who has to cover finance and software in the same week is less productive than one who specializes). Third, above 35 placements per year per recruiter, in-house wins on cost regardless and the question becomes how many in-house recruiters to hire — not whether to.
The thing not to do is build the in-house team because it feels like the right move strategically. The math determines whether it’s the right move. If the math doesn’t support it, the strategic argument is usually a rationalization for a hiring decision that wasn’t going to amortize.
Frequently asked
What about the value of in-house TA's company-specific knowledge?
It’s real, but it’s not free — it gets built over the first 6–12 months on the job, during which the in-house recruiter’s productivity is below average. The company-specific knowledge then compounds and improves placement quality and offer-acceptance rates relative to agency. Net, this is a premium that flows through as higher placement quality rather than as direct cost savings; the build-or-buy math captures the cost side, and a company should layer in a quality premium of 10–25% on the in-house side when modeling. The threshold doesn’t move dramatically once you do.
Should we hire one in-house TA or two part-time agency contracts?
If your hiring is steady at 15–25 placements per year, the AI-sourcing-augmented in-house hire usually wins on three-year cost. If it’s lumpy (8 placements one year, 28 the next), agency contracts win on optionality — you can spin spend up and down without the fixed-cost overhang. The third option, which often beats both at higher volumes, is one in-house TA recruiter plus AgentRPO or equivalent for the spike capacity above what the in-house recruiter can absorb.
What if our hiring volume is genuinely uncertain over the next three years?
Run the model with three scenarios — low (your conservative case), base (your expected case), and high (your aggressive case). Compute the in-house cost-per-placement at each. If the high case is materially better than agency and the low case is materially worse, the right answer is usually agency-now-with-an-in-house-trigger — keep agency relationships, monitor monthly hiring volume, and bring an in-house hire in once your trailing-six-months volume is consistently in the breakeven range. AI sourcing makes this trigger more responsive because the productivity-per-recruiter is higher.
Does this analysis change for senior or executive hires?
Yes. Above ~¥15M base salary candidates, the placement fee math shifts (fees scale with salary, so a senior placement is often ¥6–9M in agency fee). The breakeven volume drops because each placement is more valuable. Senior in-house TA can also outperform agency on candidate-side judgment because the company-specific knowledge premium is higher at senior levels. The threshold for an executive-focused in-house TA hire often sits at 8–15 placements per year, well below the mid-career equivalent.
What's the simplest way to start the math myself?
Three numbers. Pull last year’s agency placement-fee invoices. Estimate the loaded cost of a mid-level in-house TA hire in your firm’s compensation structure (base × 1.4–1.6 for full loading is a defensible rule of thumb). Estimate the placements-per-year an in-house TA hire could deliver at your typical role mix — 25–30 without AI sourcing, 35–45 with. Run the cost-per-placement on both sides. The answer doesn’t require a spreadsheet; the spreadsheet adds confidence rather than the answer.
Sources
Loaded-cost figures (¥10–14M mid-level, ¥14–20M senior) are based on Japan TA recruiter compensation surveys (Robert Walters Salary Survey, Hays Salary Guide), augmented by ESAI Agency K.K. internal benchmarks. Placement-per-recruiter productivity figures are averages drawn from the published 25-month sample documented on our methodology page (a representative slice of internal data; the firm’s complete placement record is not disclosed) and the 2026 production cohort. Agency placement fee average ¥4,266,675 from the meeting unit-economics cornerstone. AI sourcing productivity lift figures from the 16-week 2026 cohort and the calendar-audit data documented in the calendar audit briefing.
Run the build-or-buy math on your team
If you’re at 15–25 placements per year and currently agency-only, AI sourcing could move you across the in-house breakeven threshold. Talk to sales for help running the cost model on your specific volume and role mix.