Callbacks: what they really cost, and how to count them

There is a category of work every shop does and almost none of them measures.

A callback is a return visit to something you have already completed and been paid for. It costs a full slot — the drive, the diagnosis, the fix, the apology — and earns nothing. Worse, it takes the slot from a job that would have paid, so the real cost is roughly double what it looks like.

The trade press generally treats somewhere around two to three percent of service calls as an acceptable callback rate. Whether that number is right for your shop matters less than this: most owners cannot say what their number is, and a number nobody knows cannot improve.

Why they go uncounted

Not laziness — the record simply is not shaped to answer the question.

A callback does not arrive labelled as one. It arrives as a phone call from an annoyed customer, gets booked as an ordinary job, and is done by whoever was free. Nothing in the paperwork connects it to the original visit. By the time anyone wonders whether callbacks are a problem, the evidence is scattered across three months of unrelated job records.

There is also a quieter reason. Counting callbacks means naming them, and naming them can feel like accusing your technicians. That instinct is understandable and it is the main thing standing between most shops and a straightforward improvement.

Counting them costs almost nothing

You need three things, all of which you already have:

Same customer. Same equipment. Inside a window. Pick the window — thirty days is a common choice, ninety catches slower-emerging faults. Any second visit meeting all three is a candidate.

That is it. It will over-count slightly, because some are genuinely new faults on the same unit. Over-counting is the right direction: a list with a few false positives is reviewable, and a list nobody keeps is not.

Do it for one quarter and you will have something almost no competitor has — an actual number, and a list of the jobs behind it.

What you will probably find

Callbacks rarely spread evenly. They cluster, and the clusters are usually one of these:

One fault type. A particular component, a particular installation practice, a particular model that keeps coming back. This is the best possible finding — it is a specific, fixable thing.

One customer or one site. Sometimes the building is genuinely difficult. Sometimes the expectation was set wrong at the quote. Either way it is a conversation, not a training problem.

One technician. This is the finding everyone fears and it is usually less dramatic than expected — most often a gap in exposure to one kind of job rather than carelessness. It is also the one where having the number matters most, because without it you are managing on impression.

The paperwork. A meaningful share of callbacks are not faults at all. The work was done and the customer does not believe it was, or does not remember agreeing to what was left undone. Those are dispute callbacks and they are solved with a record rather than a spanner.

What actually reduces them

Photograph before and after. The single highest-return habit in the trades, and it costs seconds. It settles dispute callbacks entirely, and it gives you the evidence to see fault patterns you would otherwise argue about from memory.

Write down what you did not do. "Left the second valve — customer declined" prevents the return visit where you are blamed for it. Most shops record work performed and nothing about work deliberately skipped.

Say what happens next. "This will drip for a day or two, that is normal" converts a callback into a non-event.

Review the list monthly, not defensively. Fifteen minutes with the actual jobs. The point is finding the cluster, not finding the culprit — and a review that feels like a hunt will produce technicians who stop reporting them.

The honest limits

Not every return visit is a failure. Some faults genuinely emerge later, and some customers call about something unrelated on the same unit. The counting method above cannot tell these apart on its own — a human has to look at the list.

A low rate is not automatically good. It can mean customers are not bothering to call you back, which is worse and much harder to see.

Two to three percent is a commonly cited figure, not a law. Commercial work, older stock and emergency-heavy books all behave differently. Your own trend over time is far more informative than any published benchmark.

Where to start this week

Pick your window. Look at last quarter. Find every second visit to the same customer and same equipment inside it.

Whatever that number is, it is now a number — and the shops that improve at this are simply the ones that started looking.