How to Work Out What Each Payroll Client Really Costs You to Serve
Most payroll bureaus can name their heaviest clients and prove nothing. A practical method for working out true cost to serve, and what to do with the answer.
Ask anyone in a payroll bureau to name the client everybody groans about and you will get an answer immediately. Ask what that client pays, and whether the fee has ever been set against what they actually cost to serve, and the room goes quiet.
Payroll client profitability is the most commonly discussed and least frequently measured number in a bureau. Almost every owner believes two or three clients are underpriced. Almost none can prove which.
Why cost to serve is invisible
A payroll fee is usually built from things that are easy to count. Number of employees. Frequency of the run. Perhaps a starting assumption about complexity.
None of those are what makes a client expensive.
What makes a client expensive is friction, and friction is distributed across a bureau in a way that no single person sees. The chasing happens in one person’s inbox. The queries land in another’s. The re-keying happens quietly at somebody’s desk because it is quicker than explaining. The month-end reconstruction of what actually happened falls to whoever does the billing.
Every one of those costs real hours. Not one of them appears on the fee note.
The four things that actually drive cost
In most bureaus, cost to serve is dominated by four variables, and headcount is not among them.
Change volume, and change quality
A client sending forty changes cleanly through one route is cheaper to serve than a client sending eight, in three separate emails, with the effective date missing from two of them. The first costs keying time. The second costs keying time plus interpretation, plus a chase, plus a re-check.
Query volume
Client questions are genuine service work, and in most bureaus they are entirely unmeasured because they arrive in individual inboxes and are answered between runs. A client generating fifteen queries a cycle can consume more of the team’s week than one running three times the payroll volume.
Lateness
Lateness is the most expensive client behaviour there is, because its cost is non-linear. A change arriving before cut-off costs a few minutes. The same change arriving after cut-off costs a judgement call, a possible re-run, an explanation, and sometimes a correction in the following period.
Rework
Corrections, adjustments and reissued reports. Rework is doing the same job twice, and it is almost always caused upstream rather than by the team doing it.
A practical method
This can be done roughly in a fortnight, and properly with a system.
Step one: pick five clients. Two that everybody suspects are heavy, two that feel straightforward, and one in the middle.
Step two: count four things for one full cycle. Changes received, and how many arrived incomplete. Chases sent. Queries answered. Corrections made after the run.
Step three: attach a rough time to each. Not a precise one. A shared estimate from the team is accurate enough to be useful, and arguing about whether a chase takes six minutes or nine will stop the exercise from finishing.
Step four: compare the total against the fee. Express it as an hourly recovery per client.
The result is usually uncomfortable and always useful. In most bureaus the spread between the best and worst recovering client is far wider than anyone expected, and the worst is rarely the biggest.
What to do with the answer
Repricing is the obvious move and often the wrong first one.
The more productive question is whether the cost is inherent to the client or caused by how work arrives from them. A client who sends changes in three emails is not being difficult. They have never been given a better route, and they cannot see what the current one costs.
Give that client a structured way to submit changes, with the required fields built in, and a substantial share of the cost disappears without a single awkward conversation about fees.
Where the cost is genuinely inherent, the fee conversation becomes straightforward, because it stops being an opinion. “This client sends three times the change volume of a comparable client and pays less” is a sentence nobody argues with.
The reason most bureaus never do this
Not reluctance. Arithmetic. The data needed to answer the question does not exist in a countable form, because it is spread across inboxes, phone calls and memory.
That is the practical case for running payroll operations on a platform rather than around one. When every change is captured at the point it arrives, every chase is logged and every query is tracked against the client that raised it, cost to serve stops being a fortnight’s project and becomes a report.
Changepen works alongside the payroll software a bureau already runs, as the operational layer where that information actually lives.
If it would be useful to talk through how other bureaus have approached this, the conversation is payroll-focused rather than a sales demo.