The Callback Playbook: What to Do When a Job Comes Back

Every service business has a version of it. The roof leaks again after the first heavy rain. The system that was serviced on Tuesday is blowing warm air on Thursday. The filler settled unevenly and the client is back in the chair, unhappy. Most shops handle these one at a time, by whoever picks up, in whatever mood the week has left them in. That is expensive twice over — once in the labour you give away, and again in the review that shows up a week later. A callback playbook does not stop jobs from coming back. It decides in advance who owns the response, how fast it goes out, and what you are willing to do before anyone starts negotiating.

Sort it in the first hour: callback, complaint, or new job

Three different things arrive through the same phone call, and they need three different answers. A callback is work you already did that did not hold. A complaint is work that held but disappointed — the tech left a mess, nobody explained the aftercare, the result is fine but not what the customer pictured. A new job is a separate problem the customer has decided is your fault because you were the last one there.

Sorting these is not about deciding who pays. It is about routing. A callback goes to the crew who did the original work, or the closest person if they are booked. A complaint goes to whoever owns the customer relationship, not to a technician. A new job gets quoted like any other, with an explanation of why it is separate that you can say out loud without sounding defensive.

Write the criteria down before you need them. A team that has to guess will guess in the direction that avoids conflict today, and you will give away paid work because a Thursday afternoon was busy. The categories also let you count — you cannot fix a callback rate you have never measured.

Answer faster than the customer expects, and say less than you want to

The response window on a callback is shorter than on a new lead. Someone reporting that the work failed is already deciding whether you stand behind things, and every hour of silence answers that for them. A same-day acknowledgement with a named person and a time costs almost nothing and defuses most of the anger.

What that first message should not contain is a verdict. Do not tell a customer over the phone that this is not covered, that the leak is coming from something else, or that the result is normal at three weeks. You do not know yet, and if you are wrong you have to retract it, which is worse than saying nothing. The opening line is an acknowledgement and an appointment: we hear you, here is who is coming, here is when.

Get the evidence early and get it from the customer: a photo, a short video, the time the problem appeared. It takes thirty seconds to ask by text and it changes the visit — the tech arrives knowing what they are looking at, and you have a record of the condition on the day it was reported rather than the day you got there.

Decide the free-fix threshold in advance, in dollars or in hours. Below the line, fix it and stop talking about it. Above the line, the visit is a diagnosis and the pricing conversation happens after. Teams that have not set that number argue about it in front of the customer, which is the most expensive place to have the argument.

Close the loop after the fix, and find out why it happened

The callback is not finished when the technician drives away. It is finished when someone who was not on site confirms with the customer that the problem is gone. That call is short and it does two things: it catches the fix that did not actually work before it becomes a second callback, and it is the moment when a recovered customer is most willing to say something good about you.

That is also the only honest place to ask for a review — from a person who knows what happened, never from the sequence that runs on ordinary jobs. Suppress the automated review request on any job flagged as a callback, or you will send a five-star prompt to the one customer most primed to answer it honestly.

Then look at the pattern once a month rather than once a crisis. Group callbacks by cause, by crew, and by job type. Some of what surfaces will be training. Some will be a supplier or a product. A meaningful share will turn out to be expectation-setting — work that was done correctly and sold badly, where nobody told the customer what week three looks like or that the first rain would wash silt through. That last category is a marketing problem wearing a service problem's clothes, and it gets fixed in your quotes and aftercare instructions, not in the field.

What to automate, and what stays a human call

Automate the plumbing. Flag any inbound message from a customer whose job closed in the last ninety days and surface the original job record before anyone answers. Send the acknowledgement text with the assigned name and window. Request the photo. Create the callback record, tag the category, and hold the job open until the follow-up confirmation is logged. Suppress the review sequence on flagged jobs. Push a monthly rollup by cause and crew into a report someone opens. None of that requires judgement, and none of it happens reliably when it depends on someone remembering during a busy week.

Keep the judgement human. Whether an edge case sits above or below the free-fix line, when to send your most experienced person instead of the next available one, when to refund rather than repair, and when a customer has moved from disappointed to unreasonable and the right answer is a clean, polite exit. Those calls depend on context no system has — how the original sale was made, what was promised on site, how much of the fault is honestly yours. Automate the speed and the record-keeping. Decide the money yourself.