Marketplace, staffing firm, BPO or agency

There is no canonical industry taxonomy here, which is convenient for anyone raising money. A company can call itself a marketplace, operate as a staffing firm, book revenue like a BPO and be priced like software, and no rule stops it.

But the distinctions people treat as marketing are actually accounting distinctions with valuation consequences, and only two variables generate all of them:

A. How much of the value chain you own — matching only → matching plus workflow → matching plus workflow plus delivery risk → you are the supplier. B. Who the counterparty of the labour contract is — nobody, the buyer, or you.

Answer those two and everything else follows.

#The six forms

The boundaries are sharper than they look.

Listings versus managed is the difference between selling a lead and selling a transaction. Anand Iyer's framing: the managed operator takes on onboarding and vetting of both sides, software-powered scheduling, payments and payouts, QA, trust and safety, dispute resolution, pricing control and demand forecasting — "Provide More Value Through Software → Capture More Business."

Managed marketplace versus staffing firm is one variable only: who signs the labour contract. That single question moves gross margin by about 40 points and moves revenue recognition from net to gross. Version One's three tests for a genuine managed marketplace are value-chain involvement beyond matching, assumption of considerable risk, and a higher take rate that offsets the added operating cost — it works "when value unlocked substantially exceeds added complexity and risk."

Staffing versus BPO is whether you sell bodies or a function. Staffing sells hours at a markup and the client manages the worker; BPO sells an SLA'd process and manages the worker itself. The difference shows up in tenure: Concentrix's largest-client relationships average 16 years, while Robert Half's own filing describes staffing contracts as "generally terminable on short notice and without penalty."

#Nearly everything here is sold as one thing and run as another

Apply the test across the atlas and the costume comes off quickly.

Mercor is described as a marketplace and books gross billings, paying contractors 60–70% of the flow — a 27–33% gross margin, which is the margin band of a staffing firm, not of Etsy at 71% or Upwork at 77%. Turing's model is described in its own coverage as "a staffing spread — Turing manages the engagement end-to-end, pays developers on fixed monthly or hourly terms, and earns margin on the difference" (Sacra). Appen employs or contracts its crowd and books the work: it is a BPO, and the market prices it as one at ~0.9x revenue.

The instructive counter-examples run the other way. Expert networks genuinely are net-booking managed marketplaces with a compliance layer, and they hold ~70%. Prolific charges a 42.8% platform fee on top of participant rewards, so its take is separable by construction. Whop and ShopMy are honest listings-and-payments infrastructure — 5.5% and 2.9–3.9% blended respectively — and their thin takes are exactly what the form predicts. In Paid creators, clipping and UGC ad ops the marketplace layer keeps 9–18% and the money sits in the managed agency layer, which is why the holding companies bought agencies (Publicis/Captiv8, Accenture Song/Whalar) rather than platforms.

#What employing your supply does to you

It works. That is the problem.

Managed by Q is the cleanest case in the record. It went the opposite way from every "Uber for X" of its era, making cleaners W-2 employees with stock options, a 401(k) and health insurance, and adding a marketplace layer in 2017. Quality and retention improved. WeWork acquired it in April 2019 — and then sold it to Eden in March 2020 for $25M, reportedly about 11% of the purchase price, eleven months later.

The failure modes cluster around the same move. Munchery vertically integrated into its own kitchens and delivery, turning a marketplace into a capital-intensive food manufacturer, and its bankruptcy showed $1–10M of assets against $28.5M of secured plus $6M of unsecured debt. Shyp employed couriers as W-2 staff and priced a $5 pickup fee against a genuinely expensive physical operation. Handy survived by selling to ANGI — into a lead-generation listings business, which is down the stack, not up.

Andela is the pivot case worth studying: it began as a four-year fellowship training junior African engineers — a training business with a labour-supply subsidy — and in 2019 widened its criteria and became a global contractor marketplace across 135+ countries. The direction is the point: from manufacturing supply to matching supply, because manufacturing supply is a training-cost sink whose output walks out of the door. See Andela.

And the strongest counter-evidence to the whole venture model here: Toptal raised $1.4M in 2012, nothing since, and is still independent fourteen years later. Every VC-funded competitor from its cohort — Hired, Vettery, Triplebyte, Andela, Braintrust, A.Team — is dead, absorbed or silent.

#The decision table

Apply this to any company in the atlas, or to your own.

The last two rows are where the durable positions come from. The only anti-leakage mechanisms that survive a sophisticated buyer are liability transfer and workflow embedding — the buyer cannot legally disintermediate an expert network without recreating the chaperone, and cannot route around ADP without recreating payroll-tax liability. Ratings, escrow and non-circumvention clauses are speed bumps.

A labour middleman with a large but finite, interchangeable supply pool is almost always an asymptotic marketplace in NfX's taxonomy: past the point where the buyer can always get a qualified person inside their tolerance window, additional supply adds roughly zero buyer value while adding cost. That is the most under-appreciated structural fact in this model, and it is the reason "we have 1M experts" is not a moat — GLG has more than a million and watched its market share fall from 51% to 24%.

#Where the record is thin