Your revenue comes from clients you never call

Both, but not in the current proportions. 79% of recruitment agencies make at least half their revenue from clients who come back, and in nearly every organisation I know, almost all structured commercial time goes to winning new ones. Effort is allocated as the inverse of where revenue comes from.

That imbalance is not a lapse of judgement, and that is what makes it interesting. It comes from what gets measured: prospecting produces calls, meetings and opened accounts, which is to say rows in a table. An existing client you do not call produces nothing at all, and an organisation reacts only to what leaves a trace.

Where does your revenue actually come from?

From clients you have already served, in a proportion most owners underestimate until the day they run the numbers.

The available measurement puts 79% of agencies drawing at least half their revenue from recurring clients. The calculation takes an hour in your own systems: revenue invoiced last year, split between accounts opened during the year and accounts that already existed. Most owners who run that exercise find a proportion higher than the one they had been quoting, often considerably.

The second calculation is more uncomfortable and takes the same time. Take the list of clients who invoiced at least once two years ago, and see how many invoiced anything in the past twelve months. The gap between the two lists is your dormant portfolio, and it is usually the same order of magnitude as your annual growth target.

Why commercial effort always goes elsewhere

Because the incentive structure asks for it, and nobody ever wrote it down that way.

A business manager is judged on open roles and placements. Opening an account is an event, it gets told in meetings and celebrated; calling back a client served eight months ago is an event for nobody, even when it produces the same assignment. The first behaviour is rewarded, the second is assumed to happen by itself, which is the definition of work that does not happen.

There is a real awkwardness on top, worth naming rather than moralising about. Calling a client after eight months of silence means having to account for the silence, and nobody wants to open a commercial conversation there. The usual solution is to wait for a pretext, an exceptional candidate, a piece of news, which postpones the call indefinitely since the pretext never arrives at the right time.

It is the same mechanism, on the client side, that makes a consultant on assignment invisible to their own employer. In both cases the relationship you already have produces no alert, and it eventually loses the arbitration against the one you do not have yet.

What a dormant account is not

It is not a lost account, and one question settles the distinction: was there an incident?

A lost account has a nameable cause. A dispute, a placement who left after three months, a competitor signed at group level, a change of contact who arrived with their own partners. Those accounts are worked differently, and some cannot be recovered.

A dormant account has none of that. It has a last assignment that went well, an invoice paid, a contact who would probably take your call, and no conversation since. It did not decide to leave, it decided nothing at all. It is the largest and least worked category, and the only thing separating it from an active client is elapsed time.

I owe a concession here, because the optimistic version of this reasoning is false. Some of those accounts will not come back whatever you do: the company brought recruitment in house, it changed size, the need is gone. The sorting happens on the call and not before, which means part of the work described here consists of getting fast noes rather than yeses.

In what order should you call them back?

Not from largest to smallest, which is the reflex and the wrong criterion.

Three signals beat historical revenue. A change of contact first, because a new manager arrives with needs and no incumbent supplier, and that window closes within a few months. The end of an assignment you delivered next, which is the only moment your work is still visible inside the client. And the opening of a role you could have covered, which proves the need exists and that the reflex to call you has gone.

The amount once invoiced, by contrast, says almost nothing about the probability of return. It says what you were once able to do, inside an organisation that may no longer exist in that shape.

The opposite risk, which nobody writes down

Living off recurring clients has real drawbacks, and skipping them would make everything above suspect.

The first is concentration. An agency with half its revenue in four accounts no longer has a negotiating position, and it shows first in the fees, then in the terms, then in what it agrees to work in open competition. Profitability per assignment on a recurring account is often lower than on a new one, precisely because the relationship lasts.

The second is more insidious: a client you have served for years stops being challenged. You take the brief as given, you stop questioning their interview process, you do not tell them their feedback time is costing them the best candidates. It is comfortable for everybody and it is exactly how an account degrades without any incident.

So the conclusion is not to stop prospecting, it is to stop treating new business as the only commercial work that deserves a process. Both need a list, a calendar and an owner; only one of them has all three today.

What delegates here, and the figure not to believe

The list and the date delegate. The message does not, and that boundary matters more here than almost anywhere else.

Knowing which accounts invoiced eighteen months ago and nothing since, which have had a change of contact, which have posted a role you could have covered, and preparing the call with what actually happened last time: that is memory work, it is verifiable, and it requires no judgement. It is also the kind of task whose result you can check at a glance, which makes it a good candidate for a first trial.

That is the split we settled on for Balt, and it fits in one sentence: it keeps the list and prepares the call, a person makes it. That division looks timid on a brochure and it is the only one that survives contact with use, because the half it keeps is the half nobody had time for.

What does not delegate is the call itself, and the opposite temptation is strong because it is easy to sell. A generic reactivation campaign sent to three hundred former clients produces precisely the opposite of the intended effect: it teaches each of them that the relationship was a row in a file. That is work that looks finished and costs the person receiving it, with the aggravating factor that it is addressed to people who know you and will spot the difference immediately.

A last word on a figure you will run into, because it illustrates exactly what this article argues. The Bullhorn study published in February 2026, covering nearly 2,300 professionals, finds that top-performing agencies are four times more likely to use AI, and that 78% of those growing more than 25% use tools embedded in their software. Read the sentence in the direction it is written: it is the top performers who adopt, not the adopters who become top performers. The correlation is real, it is measured by a software vendor, and it says nothing about what the tool produces for somebody else. An agency with no list of dormant accounts will not have one after buying a tool.

That leaves where the freed time should go, and the answer is not obvious: between a dormant account to call and one more role to work in open competition, the arithmetic of fill rates settles it fairly brutally.