Quality stops being about skill the moment you have more than one person doing the work. It becomes about consistency, and consistency is a system problem.
TL;DR Build a short mandatory checklist into job completion. Audit a sample of photos, not every job. Trigger an immediate alert on poor feedback. Use it to coach, never to punish, or the data goes bad.
Checklists at the point of work
Not a separate audit process. Built into completing the job.
Design rules
- Under ten items. Longer lists get tapped through without reading.
- Only things that go wrong. Not everything that happens.
- Binary where possible. Done or not done, no scales.
- Photo required on the critical items, which makes them verifiable.
- Mandatory to complete the job, so it cannot be skipped when busy.
Example
Before leaving site
1. Isolation valves accessible and labelled? [ ]
2. Pressure checked and recorded? [ ] Value: ___
3. Test run completed with customer present? [ ]
4. After photos taken, same angles as before? [ ] Photo required
5. Work area cleared, dust sheets removed? [ ] Photo required
6. Care sheet handed over and explained? [ ]
7. Anything noted for the next visit? [ ] Notes: ___
Build the list from your actual callbacks. Every recurring callback reason should have a checklist item preventing it. That is how the list earns its place rather than being generic.
Sample, do not audit everything
Auditing every job is not possible and not necessary.
A workable sampling approach
| Situation | Sample rate |
|---|---|
| New technician, first month | Every job |
| Established technician | 1 in 10 |
| After a callback | Next 5 jobs |
| High-value jobs | Every job |
| After a complaint | Increased, temporarily, and say so |
Random within the sample. Predictable audits get prepared for, which defeats the point.
Photo-based review
The photos already being captured are your audit material.
Reviewing a set of before-and-after photos takes about ninety seconds and reveals
- Work quality and finish.
- Whether the site was left properly.
- Whether the standard was met on the invisible parts.
- Whether the photos themselves meet standard, which is its own compliance measure.
What to look for
- Finish quality against your documented standard.
- Site condition on leaving.
- Correct materials used.
- Safety items visible where they should be.
- Photo angles matching before and after.
Automated image review exists and is improving, but at small-business scale a person spending fifteen minutes a week reviewing sampled photo sets is more reliable and requires no tooling. Do not buy an AI vision product to solve a problem a weekly habit solves.
Post-job automated audits
Some checks need no human at all.
System-level checks that can run automatically
- Job completed without required photos. Flag.
- Job completed without parts recorded. Flag.
- Job time far outside the type average. Flag.
- No customer signature captured. Flag.
- Callback logged within 30 days of a completed job. Flag and link the two.
The callback link is the most valuable automated check. It builds your quality dataset with no manual effort, and it is the metric that matters most.
Customer satisfaction alerts
Route feedback by score, immediately.
- Poor score: alert to a manager within minutes, phone call same day. Never an automated reply.
- Middling score: flagged for review in the weekly batch.
- Good score: review request, and note it against the technician.
Speed on a poor score is the whole thing. The window between a dissatisfied customer and a public review is short.
Attribute feedback to the technician, and use it to coach. Never publish individual scores to the team. A public quality leaderboard produces score-begging on site and defensive behaviour, both of which corrupt the data.
Coaching, not punishment
This determines whether the system works or gets gamed.
When an audit finds a problem
- Ask first. There is often a reason. Access, a customer instruction, a missing part.
- Show the standard, with the photo comparison.
- Agree what changes.
- Check again in a fortnight.
- Close it out and say so.
Never raise it in front of the team, link a single audit to pay, or accumulate findings silently for a review months later.
A team that fears audits hides problems, and hidden problems become customer complaints. The data is only as honest as the culture around it.
Root cause over individual blame
Look at the pattern before the person.
- Same failure across several technicians? Training or process gap.
- Same failure on one job type? The standard is unclear or unrealistic.
- Failures clustering on busy days? A capacity problem presenting as a quality problem.
- One person, one failure type? Now it is individual, and it is usually training.
Most quality problems are scheduling problems in disguise. Rushed work is the single most common root cause and no checklist fixes it.
Measure it
- Callback rate, overall and by technician. The headline quality metric.
- Checklist completion rate. Below 100 percent means it is skippable and should not be.
- Audit pass rate, trended.
- Time from poor feedback to human contact. Should be hours.
- Repeat findings for the same person or issue after coaching.
Build the seven-item completion checklist from your last twenty callbacks. Every item should exist because something actually went wrong, which is what makes technicians take it seriously rather than tapping through it.
Need a pro to build it? [BOOK A CALL]