The consultant nobody has called in six months

By talking to them while they are there, which sounds obvious and almost never happens. A consultant placed at a client sees their assignment manager every week, has lunch with their teams, learns their subjects and their vocabulary; they see their employer at the annual review and at the Christmas drinks. They have two employers, only one of the two talks to them, and the other is surprised when they resign.

Median attrition in French IT services firms is 18% in 2026, rising to 23.4% in firms of 15 to 50 consultants, five points above the median. Those figures usually get read as a market problem, a pay problem or a shortage. They read better as a calendar problem, and that is good news, because a calendar can be fixed.

When does a consultant decide to leave?

Not at random, and not on the day they tell you. Departure risk concentrates at two moments, and knowing them is worth more than any general retention policy.

The first falls between the sixth and twelfth month. It is the end of the novelty effect: the assignment is mastered, the learning has slowed, and the question “what next?” arises for the first time. Over that same window, around 30% of new joiners leave their job within the first six months across all sectors, a reminder that the promise made at hiring gets tested very early.

The second arrives around the third year, and it is of a different nature. The consultant is good, the client keeps them, the firm touches nothing because everything is fine, and it is precisely that “everything is fine” which becomes the problem: three years with no visible change looks like a ceiling. This is no longer boredom with the assignment, it is the absence of a trajectory.

In both cases the leading stated reason is not pay but the absence of a path. Salary comes behind, and it is over-represented in exit interviews for a human reason: it is the one answer that puts nobody on the spot and closes the conversation politely.

Why an assignment at the client makes the employer invisible

Because the structure of the business organises that invisibility, without anybody having decided it.

The business manager who placed the consultant is judged on placements, so their attention goes mechanically to open roles and needs to cover. The consultant in post produces no alert: they bill, the assignment runs, the client does not complain. In every management system I know, a consultant who is happy and a consultant preparing their exit look exactly alike.

It is the same mechanism we describe about the consultant on the bench, which is not a sourcing problem, with the sign reversed. Bench time is visible, it costs every day and it shouts; a happy consultant on assignment is silent and costs all at once, on the day they leave. An organisation reacts to what shouts.

There is an asymmetry nobody enjoys spelling out. The client has every reason to talk to your consultant, to value them, sometimes to hire them. They see them daily, they know their work better than you do, and they have no interest in telling you they are considering an offer.

What a departure costs, in days before euros

The real cost is counted first in delay, and it is known: 45 days median to place an equivalent profile, with a considerable spread between the fastest and the slowest firms.

During those 45 days the assignment is uncovered. Depending on the contract it is billed short, suspended, or picked up by a competitor the client already had on its vendor list. The cost of recruiting the replacement is added to that, it does not substitute for it, and the margin lost on the assignment generally dwarfs the recruitment fee.

Compare it to the cost of what would have prevented it. A forty-minute check-in, three or four times a year, is about three hours per consultant per year. Across a portfolio of thirty consultants that is ninety hours annually, roughly two weeks of work spread over twelve months. I know of no other line of spending in this business where the ratio between the cost of prevention and the cost of the incident is so favourable, and so consistently decided the wrong way.

The check-in at three, six and twelve months

What is missing is almost never the business manager’s judgement, who knows perfectly well what to say when sitting across from the consultant. What is missing is the date.

The three-month check-in verifies that the assignment resembles what was sold, on both sides. The six-month one matters most, because it lands at the entrance of the first risk window: that is where you ask what the consultant wants to learn next, and answer something other than “we will look at it next year”. The twelve-month one is about trajectory rather than the current assignment, and it loses all meaning if it turns into an annual performance review.

Three rules make these meetings useful rather than decorative. They happen on the planned date, including when everything is fine, because a check-in postponed twice becomes a check-in cancelled. They are held away from the client site, which changes what can be said. And they produce a written trace whose useful part is not the summary but the next date, with a name against it.

Should you retain everybody?

No, and a retention policy that fails to say so ends up costing more than it returns.

Some departures are good for both sides. The consultant who wants to join the vendor whose product they have used for two years, the one whose interests have moved to another discipline, the one leaving to build something: keeping them means promising what you cannot deliver, and an unkept promise costs more than the departure, because it gets retold internally. What those departures deserve is a clean exit and a door left open, which is a commercial calculation as much as a courtesy: former consultants become clients, and that is one of the rare sources of new business that cost no prospecting at all.

What should be avoided is narrower and more precise: the departure nobody saw coming, by somebody you would have kept by answering a question asked six months earlier. It does not show up in the annual statistics, it shows up in a conversation, and that is what makes the date of the meeting more decisive than the compensation policy.

What an agent can do here, and what it must not

It can carry the calendar, which is exactly the work nobody does and nobody wants to do.

Knowing which consultants enter their sixth month next month, which have had no check-in for over six months, which are approaching their third assignment anniversary, and preparing the meeting with what was said last time: those are verifiable tasks whose result you can check at a glance, which is precisely the criterion that decides what you delegate first. It is also a recurring task in the strict sense, with a named owner, which requires knowing who owns a routine and what happens when it fails silently.

Here it takes a deliberately narrow shape: a recurring task owned by a named person, which says every Monday who is entering their sixth month, prepares the meeting with what was said last time, and stops there. Balt does not pick up the phone.

What it must not do is just as clear. It does not run the conversation, it does not detect a flight risk in place of a human, and it does not write to the consultant without a person having read and released the message. A consultant who receives an obviously automated nudge mostly learns that their employer takes an interest in them through a system, which produces the opposite of the intended effect.

The boundary is the same as everywhere else here: nothing outbound leaves without a person releasing it. In this case it is not merely a governance rule, it is the subject itself. The value of that meeting lies in it being human; the only thing that can be delegated is that it happens at all.

That leaves what happens when the assignment does end, and the consultant comes back available with nobody having anticipated it. That is the other end of the same calendar, and bench time is decided there, several weeks before the assignment finishes.