20% vs. 25%: Why Your Recruiting Fee Model Is the Wrong Thing to Optimize

A CTO at a Series B company asked me last quarter, "Can you match 8%?" He had a competing proposal from a larger agency at that rate. I said no, and I told him why. Not because 20% is sacred, but because he was asking the wrong question. He was optimizing the cheapest variable in the equation. The one that barely moves the outcome.

Here is what actually moves the outcome: whether the senior engineer you hire ships anything in the first six months. That is the line item worth obsessing over. The gap between a 20% fee and a 25% fee on a $200,000 base salary is $10,000. The gap between a hire who ships and a hire you fire in 90 days is somewhere between $80,000 and $300,000, depending on seniority and what the role was supposed to build. The fee is rounding error.

I have been doing technical recruiting for 20 years, mostly on the engineering side. I spent a decade in engineering myself before that, and I have placed over 1,000 engineers into companies like Agoda, Hearst, Con Edison, and Trilogy. I have watched companies pay 30% and get garbage. I have watched companies pay 8% and get gold. Fee percentage is a data point. It is not a strategy.

This piece is for anyone about to sign a recruiting agreement and feeling pressure to drive the number down. Do that work. Negotiate. But first, understand what you are actually buying.

The real cost of a bad technical hire

Start with a concrete scenario. You hire a senior backend engineer at a $200,000 base. You pay a 25% fee up front, which is $50,000. Three months in, it is clear the hire is not going to work. Maybe the technical skills were oversold. Maybe the person cannot ship in your stack. Maybe they are not a cultural fit and the team is avoiding them. Whatever it is, you need to let them go.

Let us add up the actual cost of those three months.

That number does not include the less legible costs. A bad senior hire damages team trust. Reports they managed start hunting. The person who interviewed and recommended them loses credibility. The roadmap slips in a way you cannot easily reconstitute. If the hire was meant to lead a greenfield project, the project may be dead before a replacement starts.

The fee you paid is a small fraction of the hole. If the same role had been filled at a 20% fee with a worse candidate, you would have saved $10,000 on the fee and spent $145,000 on the mistake. The math is uncomplicated. The cheap fee was the expensive decision.

Now run it the other way. Imagine the same role filled once, at 25%, with a hire who stays three years and ships. Fee: $50,000. Value delivered over three years at a reasonable productivity multiple: somewhere between $600,000 and $1.5 million, depending on role and company stage. Fee as a percentage of value: 3% to 8%. Suddenly the 25% fee looks like the bargain.

This is the reframe. Recruiting fees are not an expense you minimize. They are a cost of acquisition on an asset that will compound or decay over the next 24 months. Optimize for the asset.

Why recruiting fees exist at all

A good agency is not a job board with a markup. Or it should not be. Here is what you are actually paying for when you engage a recruiter.

• Sourcing. Active, targeted outreach to people who are not on the market. The best engineers are almost never applying. Someone has to reach them, frame the opportunity, and get them to take a call.
• Screening. A first technical and cultural filter before the candidate consumes your team's time. Done well, this is the single highest-leverage thing a recruiter does. Done badly, it is worse than useless.
• Pipeline management. Keeping five candidates warm while one negotiates, so you do not start over if the first offer falls through.
• Market intelligence. Realistic comp data, counter-offer patterns, what other companies are selling candidates on.
• Close support. Managing the window between offer and start date, where candidates ghost, counter, and second-guess.
• Guarantee. Most reputable firms replace the hire for free if they leave or are fired within 90 days. This transfers real risk off your books.

You pay an agency when some combination of those things is worth more than doing it in-house. The honest question is not "what is the cheapest agency." It is "which of these jobs is the agency actually good at, and am I being sold the whole bundle or just a Rolodex?"

Why 25% is the standard

The industry standard of 20% to 30% is not a conspiracy. It is a structural output of how most agencies are built.

A typical mid-sized technical recruiting agency has 20 to 60 people on staff. The organization chart usually looks like this:

• A sales or business development team that closes new client relationships. They are compensated on deals signed, not placements made.
• Account managers who own the client relationship after the sale. They write job descriptions, manage expectations, and coordinate the recruiting team.
• Senior recruiters who own the search, do some of the sourcing, and run the candidate relationship.
• Junior recruiters and sourcers who handle volume outreach and initial screens.
• Operations, office, tools, legal, marketing.

When a placement fee lands, it gets split. The sourcer might take 10%, the recruiter 30%, the account manager 20%, the salesperson 15%, and the firm keeps the rest to cover overhead, draws, and profit. Some agencies run commission splits as high as 70% to the recruiter, but those firms typically operate on thinner team margins and compensate with volume.

On a $200,000 hire at 25%, the firm bills $50,000. After the splits and the overhead load, the firm might net $10,000 to $18,000. That is not outrageous. That is the cost of running a 30-person organization that has to keep the lights on whether this particular search closes or not.

A big agency is selling you a machine. The machine has advantages. It has scale, redundancy, and the ability to run 40 searches in parallel. It also has friction. Your search has four people touching it, and the incentives at each layer are not aligned. The salesperson wants to sign you. The account manager wants to make the client happy. The recruiter wants a placement. The sourcer wants to hit a dial count. Somewhere in the middle, quality leaks.

How a 20% fee is possible without cutting quality

The 20% fee is not a discount play. It is not a loss leader. It is not sustainable because I am desperate for business. It is sustainable because the cost structure is different.

Engineers in AI is small by design. No sales team, no account managers, no junior recruiters. When a client engages us, the person doing the search is the same person who answered the phone, which is the same person who wrote this post. The work does not get handed off four times.

The economics of that look very different.

The client pays 20% less and the operator nets roughly twice what an agency principal would after splits. Both sides end up better. The reason the math works is structural, not promotional. If I had to run a 30-person agency, I would charge 25% too. I do not, so I do not.

There are things a big agency can do that a boutique cannot. If you need to hire 40 engineers in a quarter, you need a team, and you should pay for one. If you are filling one to ten hard technical roles a year, a boutique with the right background is faster, sharper, and cheaper. Not because it is a better price. Because the overhead is not in the way.

What you should actually optimize for

Fee is one input. Here is the rest of the list, roughly in order of how much they move the outcome.

1. Shortlist quality. How many candidates does the firm put in front of you before you make a hire? Three-to-hire is excellent. Five-to-hire is fine. Fifteen-to-hire means the screening is broken, regardless of the fee.
2. Screening depth. Before a candidate reaches you, does the recruiter understand the role technically? Can they tell you what the candidate has actually built, not just what is on the resume? Ask for specifics on the last person they placed.
3. Time to shortlist. From kickoff to first qualified candidate in your inbox. For a senior engineering role this should be seven to fourteen days, not six weeks.
4. Close rate. What percentage of offers does the firm close? Industry average is around 70%. Good firms run 85% or higher. A low close rate usually means the recruiter is not managing expectations, comp, or the counter-offer conversation.
5. Guarantee strength. Ninety days is standard. Six months is better. A full replacement at no cost (not a prorated refund) is best. If a firm hedges on the guarantee, they are not confident in the screening.
6. Fee. Yes, last. Because at this point, you are comparing similar outcomes and can optimize on price.

The cheapest hire is the one that stays three years. The most expensive hire is the one you fire in 90 days, regardless of what you paid to source them.

The questions to ask any recruiter before signing

If you want a fast filter on any firm you are considering, ask these. The answers matter more than the rate card.

1. Who will actually do the search? Get a name. If the person selling you the engagement is not the person running it, ask to meet the person running it. If you cannot, that tells you something.
2. What was the last senior technical role you placed, and what did the candidate build in their first six months? A firm that can answer this has real technical context. A firm that pivots to "we have a great network" does not.
3. How many candidates do you typically put forward before a hire? Three to five is the answer you want. If the number is ten or more, the firm is outsourcing the screening to you.
4. What is your 90-day fall-off rate? That is the percentage of placements that leave or are terminated within 90 days. Below 5% is good. Above 15% is a warning.
5. How do you assess technical fit? Listen for concrete methodology. "I ask candidates to walk me through a recent system they designed, and I dig into the trade-offs" is a real answer. "We do a cultural screen" is not.
6. What happens if the hire does not work out? You want a free replacement, not a prorated refund. And you want it in writing.
7. How many searches are you running right now? Be skeptical if the answer is more than ten per recruiter. Your role is competing for attention.
8. What roles are you bad at? A firm that will not name anything they are bad at is either lying or has not thought about it. Either way, you have your answer.

If a firm answers these cleanly, the fee is negotiable and reasonable either way. If the answers are vague, the fee could be 5% and it would still be too much.

One more thing on the 20%

To be direct: our fee is not the reason to hire Engineers in AI. It is a consequence of how we are built. We charge 20% because we do not have to feed a sales layer, three account managers, and a bench of junior recruiters. Every search is run by a principal with 20 years of engineering and recruiting context, end to end. That is the product. The price is just what the product costs to run.

If a competitor can legitimately beat our work at 8%, hire them. If they cannot, do not let the fee conversation distract you from the actual question: who is going to put the right person in the seat, on time, and have them still be there a year from now.