What a recruitment tech stack costs per recruiter
Most agencies know their CRM bill and almost none know their stack cost. Here is a worksheet that gets you a defensible number per head in about an hour, including the three costs that never reach an invoice and are usually larger than the ones that do.
Ask an agency owner what their software costs and you will usually get the CRM figure, because that is the one that arrives as a single invoice with a recognisable name on it. Ask what the whole stack costs per recruiter per month and the room goes quiet, and then somebody says they would have to check with finance.
The number matters for a plain commercial reason. It is a fixed cost per head. It sits directly against your margin on every placement, it does not fall when billing falls, and it is the one line in the P&L that grew quietly every year for the last five while nobody was ever made to justify it.
This is the worksheet we use when an agency asks us to look at theirs. Get your card statement out before you start, because the subscriptions nobody remembers are always on it rather than in the accounting system.
Step one: the licences you know about
List every tool with a recruitment purpose, its total monthly cost, and how many people it is billed for. Not how many use it. How many it is billed for, which is almost always the larger number.
The normal categories: the CRM or ATS, job board access and posting, a sourcing or contact-finding tool, professional network licences, email sending and sequences, a phone system, a video meeting tool, meeting notes or transcription, electronic signatures, background and right-to-work checking, assessments, scheduling, and a reporting or business intelligence tool. Most agencies find between eight and fourteen once they actually list them.
Two things surface at this step, every time. There are tools being paid for that nobody has opened in months, usually bought by somebody who has since left and renewed silently ever since. And there are seats being paid for on behalf of people who no longer work there, because removing a seat requires somebody to remember to do it and nothing prompts them.
Both of those are free money and they are the reason to do step one properly even if you go no further.
Step two: the costs that arrive per use
These do not show up as a fixed subscription and they are real money. Job slot purchases and premium postings. Contact credits on sourcing tools. Call minutes and phone numbers. SMS and WhatsApp messages. Signature envelopes. Background check units. Any AI usage billed by consumption.
Take three months of invoices, average them, and divide by heads. Do not take last month, because usage is seasonal and you will either flatter or scare yourself.
Agencies routinely find this is around a third of the total, and that they had been thinking of it as an occasional expense rather than a running cost. It is also the part that scales with activity rather than headcount, which means it grows in exactly the quarters when you are busy and feeling good about the numbers.
Step three: the three costs nobody invoices
This is the part the exercise exists for. It is where the real number lives, and it is the reason a licence-only comparison never changes anything.
Retyping. Every detail that exists in two systems gets entered twice or corrected twice. A candidate's mobile is fixed in the phone system and stays wrong in the CRM. An address is typed into the compliance tool because the export did not carry it. Sit with a consultant for two hours and count the times they type something that already exists somewhere else in the building. Multiply out across the week and the team, and cost it at their hourly rate including on-costs rather than at salary.
Reconciling. There is always somebody whose week includes making the systems agree. Exports, a report assembled by hand every Monday, a spreadsheet that is quietly the real source of truth for something. This person is usually good at it, which is precisely why nobody has noticed it is a job. Ask them honestly how many hours, and expect to be surprised by the answer.
Switching. Moving between tools is not free even when nothing goes wrong. Each move costs a few seconds of navigation and a longer moment of remembering where you were, and a consultant does it many times an hour. This is the softest of the three numbers and I would still rather have a rough version of it than pretend it is zero, because pretending it is zero is what makes a nine-tool stack look cheap.
Step four: divide, then look at it two ways
Add all four groups, divide by fee-earning heads, and you have a monthly cost per recruiter. Now do two things with it, because the raw number on its own does not change anybody's mind.
Express it as a share of your average placement fee. This is the conversation that changes behaviour. A per-month figure always sounds small, and the identical figure set against a fee does not. It also puts the cost in the unit your desk actually thinks in.
Split it into capability and glue. Capability is what the tool does for you. Glue is anything that exists only because two other tools do not talk: the integration subscriptions, the middleware, the reconciling, the retyping. Glue is pure loss. It buys you nothing a client would ever pay for, and it is the part you can actually remove.
| Line | Where to find it | The usual surprise |
|---|---|---|
| Licences | Card statement and finance | Seats for people who left; tools nobody opens |
| Usage | Three months of invoices, averaged | About a third of the total, and lumpy |
| Retyping | Two hours sitting beside a consultant | Nobody had ever counted it |
| Reconciling | Ask whoever keeps the spreadsheet | It is a named person's half week |
| Switching | Count moves in one hour, extrapolate | Rough, real, and never zero |
What to do with the answer
Three moves, in order of what they return rather than how easy they are to start.
Cancel what nobody opens. Free money, no trade-offs, and usually the first ten per cent. Do it this week.
Remove glue rather than tools. Find the one pair of systems generating most of the retyping and either fix that join properly or bring one of them inside the other. The saving here is in hours rather than licences, which is exactly why it never gets prioritised, and it is almost always the largest number on the page. It is also the only one that improves your data quality as a side effect, because the reason two systems disagree is that a person is keeping them in step by hand.
Then, and only then, compare licence costs between vendors. Most agencies do this first. It is why the number never really moves, and it is why the annual software review feels like theatre.
The part that decides whether it stays fixed
A stack cost that comes down and then drifts back up over two years has not actually been fixed, and drifting back is the normal outcome. The mechanism is always the same: a gap appears, somebody buys a small tool to cover it, nobody adds it to a list, and three years later you are running an integration project you never agreed to.
The defence is unglamorous. One person owns the list. New tools go on it before they go on a card. Anything unopened for ninety days gets a conversation rather than a renewal. And when you next choose a major system, the question is not only what it does but how many of the eight other categories it removes, because a system that costs more per seat and takes out two other tools is usually cheaper and always simpler.
We built ours so that the phone system, video, e-signature, sourcing and campaigns are part of it rather than joined to it, which removes most of the glue by construction. That is our commercial interest and I would rather state it than hide it. The worksheet above is worth an hour of your time even if you never speak to us, because the number it produces is yours either way.
Gowri is the CEO of Recruitly.



