For about four months I reviewed every inspection my crews ran. Every one. I told myself it was quality control. What it actually was: jobs sitting in my phone until nine at night, homeowners waiting on me instead of on the inspector they met, and me getting worse at reviewing because by the twelfth one I was skimming. That is the trap in trying to review inspection quality at scale. The instinct is to check everything, and checking everything is how you become the slowest part of your own company.
Sampling fixed it. Not because sampling is more rigorous, but because sampling is the only version I actually kept doing.
Sample rates that hold up
How often you review inspection quality should track risk, not seniority or feelings.
A new inspector gets three of every ten reviewed for the first month. Somebody established gets one of ten. Any inspector who has a callback, a complaint, or a record I had to ask a question about goes back up to three of ten for two weeks, no discussion and no hard feelings, and it comes back down on its own.
On top of that, I review everything on two categories regardless of who ran it: anything where the homeowner never watched or opened the record, and anything on a property where somebody is going to be looking at this months later. Those are the ones that come back to bite.
On a shop running three inspectors at roughly six inspections a week each, that is about four reviews a week once everybody is trained. Half an hour of my time. Compare that to reviewing eighteen.
A six-minute way to review inspection quality
A review that takes twenty minutes will not happen consistently, so build one that takes six.
I scrub the recording rather than watching it end to end, and I am looking at five things.
Route order. Did the walk run in the sequence on the card, or did it wander. Wandering usually means time pressure, and time pressure usually means the schedule, which is my problem and not his.
Wide shots. Every slope, before any close-up. This is the single most common miss and the one that hurts most later, because a close-up with no context is the same as no photo at all.
The boring captures. Penetrations, boots, flashing, ridge, whether or not anything was wrong there. Undamaged is a finding. An inspector who only documents problems has produced an argument instead of a record.
Narration. Not polish, just continuity. Long silences mean whoever was watching lost the thread. If the homeowner was live on the session, silence is worse than a stumble.
Language. Did he describe condition in the standard terms, and did anything come out of his mouth that is not ours to say. That last one I check on every single review without exception, because it is the one with real consequences.
Five things, six minutes, and I write two sentences.
The line I check on every review
Every inspection I pull, I listen specifically for what happened when the homeowner asked an insurance question, because on maybe a third of jobs somebody asks.
My inspectors document condition. They do not tell a homeowner whether something is covered, what a claim would be worth, whether to file, or what their policy means, and they do not speculate about what any insurer will do. We are contractors, not licensed public adjusters, and in most states unlicensed public adjusting is a crime rather than a technicality. Coverage is between the homeowner and their insurer.
So the review is not just checking that he did not say something wrong. It is checking that he had the replacement sentence ready, in the moment, with his hands full. When somebody handles that cleanly on camera, I tell him specifically, because the behaviors you praise out loud are the ones that spread to the rest of the crew.
This is also why I treat the do-not-say rule as a process item rather than paperwork, which I got into in the walk order, shot list, and language rules.
Feedback that changes anything
Two sentences, same day, in text. That is the entire format.
One thing that was right and one thing to fix, and the fix has to be a behavior rather than a quality. "Back slope had no wide shot, start each slope wide before you close in" changes Tuesday. "Be more thorough" changes nothing and makes people defensive.
Same day matters more than depth. A note two weeks after the fact is a performance review. A note that night is coaching, and the guy remembers the roof.
What I stopped doing was the monthly sit-down where I stacked up six inspections and went through them. It felt organized and it was useless, because by then he could not remember the house and the whole thing landed as criticism instead of instruction.
I also send the good ones around. When an inspector runs a clean walk with a homeowner watching live and handles a hard question well, that recording goes to the other two. Nobody learns the standard from an index card as fast as they learn it from watching somebody they work with do it right.
When the sample tells you something about you
The useful part of sampling is not catching individuals. It is the pattern.
When three different inspectors all skip the same capture, the card is wrong, or that capture is genuinely awkward from a ladder, and the fix is on my end. When route drift shows up across the whole crew in the same week, I overbooked the week. When one guy's records are consistently thin on a specific roof type, he needs an hour on that roof type and not a talking-to.
I would not have seen any of that reviewing everything, because when you review everything you stop seeing anything. The point of a sample is that it stays small enough to actually think about.
All of this only works if reviewing is cheap. If checking a crew's work means driving back to the property or reconstructing a walk from a folder of photos, you will not do it past the second week, and the standard quietly reverts to whatever each person feels like doing. Being able to open the actual session and scrub it is what makes this sustainable, and it is why the sample rate is a real dial instead of a good intention. The use cases page has a few shops running this at different crew counts.
The wider system this sits inside is in running inspections consistently across crews.
If you are working out how to review inspection quality at your size, start with sample rates and the five-item checklist before you buy anything. Our pricing is per inspection, so the review habit does not cost extra as crews grow, and I am happy to talk through how to phase it in.
InspectStream records what an inspector observed. It is not a public adjusting service and does not prepare, negotiate, or advise on insurance claims.