Contingency recruiting marketplaces

Paraform has raised $65M and says it has paid $50M to recruiters. Those two numbers appear in the same announcement and almost nobody does the arithmetic between them. At the ~70/30 split a competitor reports the platform runs (HeroHunt — Paraform publishes no pricing page), $50M of cumulative payouts implies roughly $71M of lifetime gross placement value and about $21M of lifetime net revenue since 2023. [UNVERIFIED], derived — but it follows directly from the two sourced figures. The company has raised three times what it has ever kept.

That is not damning on its own; venture capital is meant to be spent ahead of revenue. It is damning as a sizing problem, because the pool this business fishes in is far smaller than the sector's own pitch decks say. The US staffing market is $180.2B in 2026 — but temp and contract is 89% of that, and permanent placement is only 11% (Pin, citing SIA/ASA). Perm placement is a ~$20B market, and a platform taking ~6% of salary rather than the full 20–25% fee is addressing a $1–1.2B net-revenue pool even at 100% share. Sequoia's category thesis sizes "recruitment" at $200B+. For a perm-placement marketplace that number is off by an order of magnitude, and the error is not rhetorical — it is the difference between a category that can hold a $10B company and one that cannot. See GMV is not revenue and How much money is actually in the buyer pool.

What is genuinely new is not the marketplace. It is which side of it is scarce. Hired, Vettery and Triplebyte all built marketplaces where the scarce, expensively-acquired side was candidates — a side that exits the market the moment it succeeds. Paraform's scarce side is recruiters, who are durable, repeat, and get better with use. That single swap is the strongest argument in the vertical, and it is the reason to read this page rather than dismiss it.

#Whose budget

The buyer is a venture-funded company hiring senior engineers, and the budget is real but bounded. Paraform reports an average placed-candidate cash comp of ~$260,000 ($300–400K with equity) (Paraform) — far above a generalist engineering placement, which tells you the platform is selling into the scarce tail rather than the middle of the market. At a 20–25% fee that is a $52–65K invoice per hire.

The demand side is structurally better than it looks from Big Tech headlines. SignalFire's Beacon dataset puts Big Tech hiring 25% below its 2019 baseline while early-stage startups are only 4% below, with engineering hiring at early-stage up 7% (SignalFire). Startups are the growth market for recruiting demand; the majors are not.

The healthiest single fact about the buyer: more than half of Paraform's customers already have an in-house talent team and use the platform as elastic capacity (TechCrunch). That makes it a burst-capacity purchase rather than a fight with the internal TA function — a much easier sale, and one that survives a hiring freeze better than a replacement sale does.

Against that: this is opex the buyer could substitute away from. The published cost-per-hire ladder for one $100K role runs agency contingency $20,000 → fractional recruiter ~$8,000 → in-house (SHRM average) $5,475 → AI sourcing platform ~$1,200/year (Pin). Budget scores a 3: urgent, recurring, but always priced against a salary.

#Can you get the supply

Yes, and cheaply, which is both the reason this works and the reason supply: 3 rather than a 5.

At a traditional agency the individual recruiter keeps roughly 25–35% of the fee — commission runs 15–30% of the fee depending on seniority, with a full-desk "360" recruiter at about 33% of billings, against a 60:40 base-to-commission structure (Recruiterflow). Paraform's claimed ~70%, with the standard 60-day payment hold removed (HeroHunt), is a genuine 2x on the recruiter's take.

The macro handed the category its supply: roughly 100,000 recruiters laid off in 2022–23 became independents needing demand. A useful benchmark for what they are worth: gross profit per income producer was $380,768 in FY2025 (Staffing Industry Metrics via Pin) — about $114K to the recruiter at a 30% agency split. Paraform publishes a right tail well above that: multiple recruiters at $1M+ annually, one at $300,000 in a single month (Paraform). Read those as the right tail, because that is what they are.

But supply bought with money can be bought away with money, and this is the Which side you build first problem in reverse: the hard side was cheap to assemble and is therefore cheap for a competitor to assemble. What is not cheap is the demand — 1,000+ hiring companies is the asset, and it is the thing recruiters cannot replicate alone.

speed: 4. Supply arrives as fast as you can pay for it — a doubled split needs no product to be believed — and the buyer needs no procurement, because a contingency agreement costs nothing until someone starts. What sets the clock is that the invoice is triggered by a fill, and time-to-fill has risen 37% to 59.7 days (TheHireHub) before any guarantee runs. First real money is roughly a quarter from the first req, not the weeks that get Outbound and GTM-as-a-service a 5.

#What the spread looks like

The fee stack is legible and the platform's slice is thin.

The number that makes the whole industry make sense is the fill rate. Average agency contingency fill rate is 15–25%; top performers reach 35–50% (Bullhorn GRID 2026 via Pin). At a 20% fill rate, each successful placement pays for the four that failed. That is why fees are 20–25% and not 5%.

The corollary is that the headline take rate is not the recruiter's realised economics. If several recruiters work the same req, headline earnings divide by the number who lose. See What a rake can actually be and Marketplace, staffing firm, BPO or agency — and Sales talent: AEs, SDRs and the people who close for the same ladder applied to go-to-market roles.

spread: 3. The platform keeps ~25–30% of the fee — about $15–18K on a $260K placement, or 5–7.5% of salary — mid-table, and nothing like the 70–80% Expert networks hold. What keeps it off a 2 is that the give-away is chosen rather than imposed: the 70/30 split is the wedge that bought the supply, and the fee, not the salary, is the denominator (GMV is not revenue) on a cut levied only on fills. What keeps it off a 4 is that raising the take would undo the only reason recruiters are there.

#Can you hold it

Weakly. This is the worst-scoring dimension and the honest one.

Transaction values are enormous — $30–65K per placement — so the incentive to defect on a single deal is large; the relationship is high-touch and identity-revealing, so you cannot anonymise a recruiter from the hiring manager they run a search with; and buying is episodic, so there is no habit lock-in. All three conditions are the Getting cut out problem in its most acute form.

Paraform's public Terms are unusually thin here: they bar using the platform "in a manner that violates any active contract with another existing user" but contain no explicit prohibition on an employer hiring a recruiter's candidate off-platform after introduction. The binding terms presumably live in a private Recruiting Agreement that is not public. [UNVERIFIED] — a diligence item, not a finding.

The real defences are economic rather than legal. Payment runs through the platform via Stripe, making off-platform settlement an affirmative act. And a recruiter who defects gains at most ~30 points on one deal while losing inbound reqs from 1,000+ companies — negative expected value while req supply is abundant. If demand thins, that calculus inverts.

The bigger leak is the employer, not the recruiter. Once a company has met three good recruiters through the platform, nothing but contract terms stops it putting them on a direct PSA at 18%. The counterweight is that the employer's real alternative is not "the same recruiter, cheaper" — it is "find and vet recruiters myself," which is the work they were paying to avoid.

One encouraging precedent: legacy vendor-management systems reportedly take ~25% of the recruiter's fee and hold payment 60 days, and are used by 80–85% of businesses with 1,000+ employees (Paraform, competitor-authored) [WEAK]. That enterprises have tolerated a 25% platform cut for two decades is the best available evidence the take rate is sustainable.

hold: 2. Every Getting cut out condition is present at once: $30–65K transactions, so one defection is worth a year of fees; a high-touch, identity-revealing introduction; and episodic buying, so no habit forms. What keeps it off a 1 is that the defection maths currently runs the right way — ~30 points on one deal against inbound reqs from 1,000+ companies — but that is an economic argument, not a contractual one, and it inverts the moment req supply thins.

#What AI does to it

Both directions at once, and the net is genuinely unclear.

The tailwind is large and well-measured. There were 244 applications per open role in 2025, up from 116 in 2022; 746 applications per recruiter per year, up 412%, while recruiters per organisation fell 56%; 77% of hiring teams regularly encounter AI-generated applications, up from 53% in early 2024; and time-to-fill rose 37%, from 43.6 to 59.7 days (TheHireHub, on Greenhouse data across 640M+ applications). When the inbound channel produces 244 undifferentiated candidates and time-to-fill rises anyway, inbound has stopped being a signal. That is why a 20–25% fee still clears in 2026.

The headwind is the price floor. Wellfound Autopilot is $500/month per role plus a 10% placement fee (joinnextdev) [WEAK]; Dover quotes $2,000–$8,000 per hire (HeroHunt); Juicebox sells semantic search over 800M+ profiles at $99–$179/month. Wellfound has publicly anchored a managed hiring outcome at 10% of salary — half the going rate. Wellfound's weakness is that its pool is opt-in, so it cannot reach the passive senior engineer who justifies a 25% fee; but the anchor is set regardless.

What survives is the part models cannot do: persuading a happy senior engineer to take a call, calibrating a bar with a founder who has never hired the role, and closing against three competing offers. What dies is the boolean agency. 61% of staffing firms now use AI in workflows, up from 48% in 2024, and firms using it were 3.5–4.5x more likely to be growing revenue (Pin, citing Bullhorn GRID 2026). See What better models do to each layer.

A sourcing fact with direct commercial consequence: 71% of AI/ML roles are filled by candidates whose current title is not "AI" or "ML" (Pin). Title-based search misses them. That argues simultaneously for semantic tools and for expert humans, and against everything in between.

ai: 3, a genuine net rather than a shrug. Models make sourcing cheaper — that is what Juicebox and Dover sell — while leaving the fee untouched, because the fee is levied on a hire that happened, not on the hours spent finding it. Neither side has won, which is what a 3 means. The asymmetry worth pricing: the tailwind protects the scarce senior tail while the price anchor attacks the middle, so this is a 4 if you only ever place $260K engineers and a 2 if you drift down-market.

#What would kill it

That last point is not settled. Frontlines.io frames Paraform as running "forced exclusivity" — but the page returned 403 and only the framing could be verified, not the mechanics. Dover, a direct competitor, claims the opposite: "five different recruiters are all sourcing for your senior engineer role" at once. Both cannot be right. If it is the second, realised recruiter earnings are a fraction of the headline 70% and churn follows.

#Who is already there

And then the strategic tell. The three most valuable companies that began as talent marketplaces — Mercor ($10B, in talks at $20B), Turing ($2.2B) and micro1 ($500M) — now derive the large majority of revenue from selling expert labour to AI labs, not from placing people at startups (Sacra; Sacra; Sacra). Mercor's recruiting side is largely vestigial marketing; roughly 90% of its revenue is expert data (CryptoBriefing).

#Where the record is thin