How many roles do you work for free?

More than half of them, if you work mainly in open competition. A role handed to several agencies is filled in 40 to 55% of cases, against 70 to 80% on an exclusive basis and 75 to 90% under a retained mandate. Everything else is work delivered to the client and never invoiced.

None of that is news to anyone in the business, and yet the figure almost never appears in the decision to take a role on. You look at the salary, the fee rate and how much you like the client; you rarely look at the probability that this assignment will produce an invoice, even though that is what sets the real price.

What is the real price of a contingent role?

Your fee rate multiplied by your fill rate, and that multiplication changes everything downstream.

At a 20% fee and a 45% fill rate, an open role returns an expected 9% of the position’s annual salary. This is not a criticism of the contingent model, it is its arithmetic: the client does not only pay for the placement they receive, they also pay, inside that rate, for the searches that produced nothing for anybody. Except the sharing is wrong, since the client whose role went unfilled paid nothing at all.

The practical consequence is that you need to change unit. The fee percentage compares nothing, because it says neither how much work was done nor how likely it was to conclude. The only unit that makes two assignments comparable is revenue per hour worked: expected fee, multiplied by the probability of filling, divided by the hours that kind of role usually takes. That calculation fits on one spreadsheet line and generally reorganises a portfolio in an afternoon.

An honest caveat about the figures themselves. Public sources disagree: some put open competition at around 20 to 30% filled, others between 40 and 55%, depending on whether they count roles abandoned along the way or only those actually worked. So use your own, which are in your software, rather than these ranges. What matters here is not the exact value, it is that it is below 100% and that it appears nowhere in your decision.

Which roles get filled least often?

Those where seniority and technicality compound, and the spread between categories is wider than intuition suggests.

On administrative or entry-level roles, fill rates sit between 70 and 85%. On a mid-level profile, between 55 and 70%. On an experienced manager, between 45 and 60%. On an executive position, between 30 and 45%, and on a highly technical specialist, between 35 and 50%. In other words the roles that pay most per placement are also the ones that conclude least often, which makes the average misleading in both directions.

The mode of engagement moves those numbers as much as the nature of the role, sometimes more. A need given to a single agency on an exclusive basis goes from 40-55% to 70-80%, and preferred vendor status, which merely reduces the number of competitors, already gains about ten points. There is nothing magic about it: on an exclusive basis the agency can afford the long search among people who are not looking, and that is exactly the work open competition makes irrational.

That is the mechanism explaining the rest. Paid on success and put in competition, an agency mechanically optimises for speed and probability of close: it presents candidates already searching, who reply fast and accept fast. The client, who wanted access to the hidden market, receives profiles they would have found alone. Both parties are right separately and the outcome is bad.

Why the exclusivity conversation fails, and how to hold it

It fails almost always for the same reason: it is asked as a favour, at the moment the client has the least reason to grant it.

A client opening a role to three agencies believes they are multiplying their chances. Asking for exclusivity at that moment amounts to asking them to give something up in exchange for a promise. The conversation holds far better when it is about what exclusivity buys: a committed timeline, a guaranteed number of profiles by a date, a weekly check-in, a search among candidates who will never apply spontaneously. It is an exchange, and it should be framed as one.

It holds better still if you can show your own numbers. An agency able to say “on this kind of role, exclusively, we fill in eight cases out of ten and within thirty days; in competition, in one case out of two and within sixty” is having a completely different conversation from one invoking the quality of its network. It is the same principle we argue about industry statistics repeated without being checked: your data is worth more than a generic figure, and it cannot be disputed.

Declining a role is a commercial decision

This is the part nobody enjoys, and the only one that changes the year’s result.

A role with a low predictable fill rate does not just cost the time it consumes. It costs the time it takes away from the roles that do get filled, and that substitution is invisible because work not done leaves no trace. A consultant spending fifteen hours on a role open to four agencies does not see that they have just taken fifteen hours away from an exclusive mandate they would have closed.

Declining cleanly happens in three steps, and the refusal is almost never final. You give the reason in numbers rather than as an excuse, you propose the shape in which you would take the same need on, and you ask to be called back if the role is still open in six weeks. That last sentence turns a refusal into an option, and difficult roles often come back.

What an agent actually changes about this arithmetic

It does not raise your fill rate, and that has to be said before anything else, because it is exactly what will be sold to you.

What it changes sits upstream. It makes the calculation available at the moment of the decision rather than at the end of the quarter: on this kind of role, with this client, in this mode of engagement, here is what we did over the past twelve months. The data already exists in your system and nobody has time to pull it out at the right moment, which makes it a verifiable task whose result you can check at a glance, and therefore a good first candidate for delegation.

We have Balt say it in the conversation where the need arrives, rather than in a dashboard nobody opens at six on a Tuesday. The difference is not technical, it is one of place: a correct figure arriving after the decision is worth nothing.

It also changes what happens during the assignment you did take, where delay decides the rest. A competitive role is won largely on response speed, and that delay is played out in the waiting rather than in the work: the candidate called back the same day, the client feedback chased at the right moment, the restart after ten days of silence. None of that requires judgement, all of it requires memory, and it is the one part of this job that delegates without losing anything.

That leaves the question that comes next, once your numbers are finally on the table: what do you do with the candidates you worked on the roles that never filled, and who are still in your database. They represent most of the work already paid for, and the silence they get costs more than you think.