Ask Google what a solo recruiting desk earns and the answer box tells you net profit margins run "often between 60% and 80%." Ask it the same question a different way, about whether a recruiting business really clears 60 to 70 percent net, and it says that number is not realistic, that standard net margins land between 10% and 35%, and that the 60 to 80 band is a gross margin or a delivery markup wearing a net-margin label.
Both answers came from the same search engine on the same day. Neither is lying.
On a desk making ten placements a year at a $22,000 average fee, one placement costs about $11,000 to produce, fully loaded. At six placements a year the same cost base puts it at $18,333, which is more than the fee. The 60 to 80 figure is real for a desk billing $450,000 and roughly minus two percent for a desk billing $108,000, and the whole difference sits in one line that a solo P&L almost never carries: the cost of your own time.
This is the cost side of your desk, line by line, with a model you can run on your own numbers. Data as of August 2026.
Why the 60% number keeps circulating
A solo desk has an unusually short list of cash expenses. Tools, an accountant, insurance, an entity, a phone. Take a year's billings, subtract the money that actually left your bank account, divide, and you get a number in the high eighties. Call that your margin and you have manufactured the 60 to 80 percent figure yourself.
The line you skipped is the largest one. If you stopped billing tomorrow and hired someone to run your desk, you would pay them. That salary is a cost whether or not it appears on a bank statement, and every published agency benchmark already includes it. That is why agency numbers look so much worse: Execue's 2026 cost analysis puts most agencies at 4 to 10 percent net, with the top five percent reaching 20 to 30 percent. Those are net margins, so by definition they are struck after recruiter salaries have been paid.
So an agency's 8 percent and a solo desk's 80 percent are not the same measurement. Comparing them tells you nothing. Price your own hour, and the two become comparable for the first time.
What your hour is worth, and why every vendor model gets it wrong
Here is a pattern worth noticing. Execue prices a recruiter's hour at $40 to argue that automatable work is eating $250,000 a year off a five-person agency, which sells consolidation onto its platform. Reqcore's ATS cost model prices staff time at $50 an hour and half an hour a month to argue that self-hosting maintenance is close to free, which sells self-hosting.
Same rate, opposite conclusions, both convenient for the party publishing them.
For a salaried recruiter, $40 an hour is roughly right. Glozo's own market data puts the US median for the recruiting profession at $72,800, and loaded with employer taxes and benefits that lands near $91,000, or about $45 an hour across a full working year. If you want the underlying market read, it is in our recruitment industry salary report, and the pay picture by model is in recruiter salaries for freelance and agency desks.
For a solo biller that $91,000 is not what your hour is worth. It is the floor beneath which the desk is not worth running, because you could earn it working for somebody else and carry none of the risk. Use it as the cost of your labor, not as its value.
Glozo's market data is built on 10M+ job market signals processed monthly, sourced from 30+ public and partner data signals.
The tool stack, and the minimum that punishes you for being one person
A solo stack is not a five-person stack divided by five. LinkedIn Recruiter Corporate carries a three-seat minimum, so the industry-default sourcing product is not available to you at one fifth of the price. Recruiter Professional Services, the version built for agencies, runs $6,000 to $10,000 per seat per year in Execue's 2026 breakdown. An ATS built for agency work runs $85 to $199 per user per month on the vendors' own published pages, checked 28 August 2026: Crelate at $85 on Essentials and $119 on Business, Loxo at $149 on Core and $199 on Professional. Contact data adds $49 to $119 a month. Then outreach, scheduling, email verification and an AI assistant.
Call it $14,000 a year for one person, and note what that means. Execue budgets $15,000 to $22,000 per recruiter per year for a five-person agency. You are paying almost the same per head with none of the volume discounts, because the minimums do not scale down.
At ten placements a year, your stack costs $1,400 per placement. At six, it costs $2,333. The stack does not get cheaper when you have a bad year, which is the first thing the model will teach you.
The reqs that died are part of the cost of the ones that closed
This is the line that breaks most desk models. You do not spend your year on the placements you made. You spend it on every req you accepted, and most of them paid nothing.
Say a filled search takes you 90 hours end to end: sourcing, outreach, screening calls, submissions, client management, offer stage, paperwork. A req you eventually walk away from still costs you 40 hours before you admit it is dead. If you work 25 reqs and fill 10, you spent 900 hours on the ones that paid and 600 on the ones that did not.
Fifteen hundred hours produced ten fees. That is 150 hours per placement made, not 90. Any model that counts only the hours you spent on closed business will tell you your desk is about 40 percent more profitable than it is.
Three desks, same arithmetic
Everything below uses the same assumptions: 90 hours on a filled search, 40 hours on a req that dies, $91,000 for your own labor at a loaded market rate, and the tool and overhead figures above. Change any of them and the shape of the answer holds.
| Line | Desk A | Desk B | Desk C |
|---|---|---|---|
| Placements made | 6 | 10 | 16 |
| Average fee | $18,000 | $22,000 | $28,000 |
| Revenue | $108,000 | $220,000 | $448,000 |
| Reqs worked | 20 | 25 | 34 |
| Reqs that died | 14 | 15 | 18 |
| Delivery hours | 1,100 | 1,500 | 2,160 |
| Tools | $14,000 | $14,000 | $16,000 |
| Insurance, accounting, entity, banking | $5,000 | $5,000 | $6,000 |
| Your labor at a loaded market rate | $91,000 | $91,000 | $91,000 |
| Total cost | $110,000 | $110,000 | $113,000 |
| Cost per placement | $18,333 | $11,000 | $7,063 |
| Profit above a market wage | -$2,000 | $110,000 | $335,000 |
| Profit as a share of revenue | -1.9% | 50.0% | 74.8% |
| Breakeven placements | 6.1 | 5.0 | 4.0 |
Desk A bills a fee above the industry floor and still loses money, because six placements a year cannot carry a fixed cost base of $110,000. Desk C lands inside the 60 to 80 percent band the answer box quoted, which is why that number exists and why it keeps getting repeated by people who have only ever run a Desk C.
Two things move the answer, and only two. Volume, and fee size. Your cost base is close to flat, so cost per placement falls from $18,333 to $7,063 across the same desk doing more of the same work.
Desk C also runs 2,160 delivery hours, which is 43 hours a week for 50 weeks with nothing left for business development. Desks at that volume are almost always running repeat clients rather than cold outreach, and that is a structural fact about the model rather than a work-ethic story.
Your breakeven placement count
The cost per placement tells you what a fee has to clear. The breakeven count turns that into a target. Divide your annual cost base by your average fee. That is the number of placements below which the desk is a worse deal than a job.
Hold the cost base at $110,000 and an $18,000 average fee needs 6.1 placements a year, a $22,000 fee needs 5.0, and a $28,000 fee needs 3.9. Desk C in the table shows 4.0 rather than 3.9 because its heavier stack pushes its own base to $113,000. Raising your average fee by $10,000 removes two placements a year from your breakeven, which is a far cheaper lever than adding two placements.
That is the case for specializing, stated as arithmetic rather than as advice.
The guarantee is a line item, and nobody publishes the rate
Most placement agreements carry a replacement obligation. In a 2019 Top Echelon survey of its own recruiting network, whose members average fifteen years in the profession rather than the industry at large, 61.4% offered replacement with no money back. No one has measured it since. On those numbers a fall-off usually costs you a delivery cycle rather than a refund. Ninety more hours, no new fee.
The rate at which that happens is your number, and you will not find it published anywhere. There is no credible industry benchmark for how often placements fall off inside the guarantee window. What exists is vendor blog assertions with no source behind them.
So measure your own. Count the placements you made in the last three years, count the ones that triggered a guarantee, divide. Then apply it: every percentage point of fall-off adds about 0.9 hours to the true cost of every placement you make, because you are amortizing replacement work across your whole book. On Desk B at an 8% rate, that is 72 extra delivery hours a year for zero extra revenue, and it raises cost per placement by roughly 8%.
If you want the full model behind pricing that clause, including refund ladders and the client profiles worth declining, that is a separate piece of arithmetic.
Cost per hire is not cost per placement
You will find plenty of published numbers for cost per hire. Ignore all of them for this exercise. Cost per hire is an employer-side metric: what a company spends to fill its own seat, including job ads, internal recruiter time, referral bonuses and your fee. Your fee is an input to their number.
Cost per placement is what you spend to produce that fee. Different payer, different scope, different denominator. Putting one into a model built for the other is the single most common way these calculations go wrong.
The same trap catches the revenue side. Your fee percentage is not your margin, and the fee models are covered properly in recruitment agency fees and, for the retained versus contingency question, in contingency versus retained recruiting.
What to do with this
Run the table on your own last twelve months. You need four numbers: placements made, average fee, reqs worked, and an honest hours estimate for a filled search and a dead one. Your bank statements give you the cash costs. The labor line is the market rate you would have to pay a replacement.
If your cost per placement is above your average fee, you have a volume problem or a fee problem, and the breakeven number tells you which lever is cheaper.
The largest controllable input is hours per req, and most of that time goes into finding and qualifying people before anyone talks to a human. The channel-level version of that arithmetic, from outreach volume through to one signed offer, is in the InMail cost and response rate math.