Where the margin goes in a trades operation

A trades business that is growing but not widening its margin is a specific and frustrating shape of problem, because there is usually nothing to point at. One bad job is visible. One bad hire is visible. Eleven percentage points of drive time spread evenly across every technician, every day, for three years, is not visible at all — it is just the texture of the operation.

That is the general rule for margin in this industry: the costs large enough to matter are distributed thinly enough to be invisible. Which means they survive every attempt to fix the business by working harder, because working harder does not change a ratio.

Take your technician headcount. Multiply by the share of the day that is drive time rather than wrench time. Multiply by fully-loaded cost.

On 40 technicians running 34% drive time against a 23% target, that is an 11-point gap — 4.4 technician-equivalents, or roughly A$420,000 a year at A$95,000 fully loaded.

The arithmetic is easy. The middle term is not. Most operations cannot state their drive-time ratio, because dispatch data records where people went and when jobs closed, not how much of the day was spent getting there versus working. Extracting it is a few days of work against data you already hold, and it is worth doing before anything else, because it is the term the whole case is most sensitive to.

The target is the harder half. The 23% used above is an input to this model — it is not a published industry standard, and we are not asserting it about your operation. The comparison that means anything is against operations genuinely like yours: same geographic density, same job mix, same emergency-to-planned ratio. Building that comparison set is part of the assessment, and it is what turns the gap from an opinion into a number.

An emergency callout and a scheduled visit are frequently the same labour sold at different margins. The emergency carries a labour premium you pay rather than charge, it disrupts a day that was already committed, and it arrives with a client conversation nobody enjoys.

So the question worth asking is what share of your emergencies were predictable — not in principle, but from data you were already collecting. Modern plant emits condition data continuously, and most of it is retained by somebody and read by nobody.

We want to be careful here, because this is where trades AI marketing tends to overreach. A model that usefully predicts failures is not a given. It depends on how many years of service history you hold, how consistently the failures were coded when they were recorded, and whether the sensors exist at all on the equipment that actually fails. Some operations have five clean years. Some have a filing cabinet and a WhatsApp thread. Which one you have is a question with a definite answer, and it is answerable in days rather than months.

The problem is rarely that an owner has not tried. Route optimisation gets bought. Scheduling software gets bought. Consultants deliver documents. The common failure is that none of it started from a number.

Not ‘drive time is probably an issue’, but a measured percentage, a defensible comparison, and an annual dollar cost. Without the percentage you cannot decide what to fix first. Without the comparison you cannot tell whether your operation is genuinely poor or merely average — and those call for completely different responses.

Run against your dispatch and service records. Measure the drive-time ratio directly instead of estimating it. Build a comparison set from operations with a similar density and job mix. Test what share of emergency callouts were predictable from the history you actually hold. Then rank what surfaces and pick one thing.

If the numbers come back small, we say so. A tight metropolitan operation running 20% drive time with a planned-maintenance book already in place has a materially weaker case than the one modelled above — and it is better for everyone that this shows up in week one than after an implementation.

It does not tell you your numbers. Every figure above is arithmetic run on illustrative inputs, chosen because they are plausible — not because they were measured at a particular business. No client engagement underlies any of it.

The inputs that matter most are the ones you would have to supply: your drive-time ratio, a defensible target, and the state of your service history. Change any of them and the case changes with it. That is the honest position, and it is why the first step here is a measurement rather than a proposal.