Turnover Cost, Calculated Roughly

A rough figure for what a departure cost does not win the retention argument. It makes both sides of it the same kind of argument, which is a smaller thing and most of what was missing.

A retention argument arrives without a number attached

The exchange has a familiar shape. Somebody says the practice is losing medical assistants faster than it can train them, that whoever stayed is covering for the ones who left, and that this is going to break something. All of that is true. The answer comes back as a figure: what lifting a band would cost across a year, what another position costs once payroll taxes and benefits sit on top. One side of the conversation is precise and the other is a feeling, and precision carries a room whether or not it is right.

Owners are not especially hard to persuade. Turnover simply has no line. Nothing on a monthly statement carries that name, and the cost of it is spread across categories that each arrive with an explanation of their own, so the only person who sees the whole is whoever watched it happen.

Most of what a departure costs is already recorded, under other names

The pieces are mostly in the building already. Overtime sits on the wage line while a seat is open, and agency cover lands somewhere else entirely, alongside contract labour. The posting has a price. The hours spent screening and interviewing do not, because salaried people did them between other work, and that piece is almost never counted. Then there is the training already spent on the person leaving, and the same training spent again on whoever replaces them.

None of it arrives labelled. The overtime line moved and we were short for two months, which is accurate and closes the subject. Assembling the pieces into one figure takes no new accounting, only the same figures sorted by the event that caused them rather than the category they were posted to.

The expensive part is the ramp, and it has no receipt

A new hire is paid in full from the first week and works at some fraction of the role for months after that, and nothing meters the gap. At the front desk it surfaces as registration detail that becomes a denial six weeks later, by which point it is filed as a billing problem. In billing it is somebody appealing what a longer-tenured person would have caught before the claim went out. Whoever is teaching them is doing it on top of a full day, so a share of the cost is a second person's work finishing late.

This is the point where the arithmetic wants to become a time study, which is what kills the exercise. What is knowable is coarser. How long it took the last two hires to stop asking is a question the people who trained them can answer in an afternoon.

Rough is the point rather than a concession

The temptation is to build something defensible to the decimal, and precision is what these die of: a figure carried that far invites an argument about method, which is a comfortable place to leave a decision unmade. What survives a partner meeting is a range assembled out of the practice's own material, its wage rates, how long the last few vacancies stayed open, what overtime did in those months. An outside multiple of salary sounds more authoritative and is the weaker version, since the first question anyone sensible asks is whether that figure describes this practice, and nothing inside it answers. A number built from this payroll and these vacancies is less impressive and harder to wave off.

The second number is the one that gets underestimated

A comparison needs both sides, and the retention side is harder to state honestly. A raise that keeps one person is rarely a raise to one person. Pay travels, the way it always does, and what ends up adjusted is a band. So one side of the comparison happened once and the other recurs every year and reaches everybody in the role.

Setting it out that way is what makes the exercise worth doing, because it lets the comparison come out the other way. A departure can be genuinely expensive and still cost less than the standing price of whatever would have prevented it, and a version of this that can only land on one conclusion was not worth assembling.

One blended figure averages away the difference between roles

The cost differs by role, by enough that a single practice-wide number conceals the decision sitting inside it. Replacing a front desk hire in a market that still has candidates is not the event that replacing a biller is, and neither is replacing an advanced practice provider, where whoever arrives sits through credentialing before the work can be billed and the vacancy runs months past the start date.

None of this decides anything by itself. What it changes is the subject, from whether turnover is expensive, which nobody disputes and nobody can act on, to which departures are expensive here and what the alternative costs. The failure mode is a model built backwards from a conclusion somebody already held. Those are recognisable by how neatly they land, and the price of one is paid by the next number carried into that room.