Why Your Best Location Cannot Be Copied
One branch outperforms the rest by a distance and nobody can say why. Every attempt to roll out what they do produces a manual nobody follows.
One site does thirty percent better than the others on the same catchment, the same pricing and the same equipment. Everybody knows which one it is.
Somebody visits, writes down what they do, and circulates it. Nothing changes anywhere else.
The manager is not the whole answer
The usual explanation is that the branch has a good manager, and that is usually partly true and completely unhelpful, because it implies the only lever is recruitment.
What a good manager actually does is make a hundred small decisions differently: how the day is sequenced, which jobs get taken and which get declined, when they ring a customer back, how they handle a complaint before it becomes one, what they do with a quiet Tuesday.
None of that is in a process document. Most of it the manager cannot articulate, because to them it is just how you do the job. Ask directly and you will get a description of the official process, which is the same process every other branch follows.
Why the visit produces nothing
Somebody spends a day observing and writes a summary. Three problems, and they are structural rather than about effort.
A day is a sample of one, and what makes the difference is behaviour across a quarter. The observer sees what is visible and misses what is absent, which is frequently the point: the good branch is not doing something the others are. And what gets written down is what the manager says they do, which is not reliably what they do.
Then the summary becomes a document, the document becomes a training session, and nothing about anybody’s Tuesday changes.
Look at the numbers underneath the number
Before assuming it is culture, check whether the difference is mechanical. It frequently is, and it is checkable.
Does that branch answer the phone faster. Does it quote the same day rather than in three. Does it decline a category of low margin work the others accept. Does it schedule differently so there is less travel. Does it follow up quotes and the others do not.
Any of those would produce a thirty percent gap on its own, and every one of them is measurable rather than cultural. Most multi-site businesses have never compared branches on anything except revenue, so the actual mechanism has never been looked for.
What AI can see across sites
The comparison was never impossible in principle. It was impractical because it means reading the operational detail of every branch continuously, and nobody has that time.
AI compares behaviour, not just output. Response times, quote turnaround, follow up rates, job mix, rework, how often a customer has to chase. Those are the differences that explain the revenue gap, and they sit in systems nobody cross reads.
AI finds what the good branch does not do. Absences are invisible to observation and obvious in data. The branch that never takes a certain job type, or never lets a quote sit past 48 hours.
AI reads the customer side. The tone of correspondence, the complaints that were resolved before escalating, the questions that never had to be asked because something was explained up front. That is where a lot of the difference lives and none of it is in a report.
It shows each branch its own gap. Not a league table, which produces defensiveness. One specific thing: your quotes take three days, the group average is one, here are the twelve that sat longest last month.
Roll out one thing, not a playbook
The failure mode of every best practice programme is scope. A forty page manual asks every branch to change everything, so they change nothing.
Take the single largest mechanical difference and roll out only that, with a number attached, for one quarter. Quotes go out same day. That is it.
It is measurable within weeks, it does not require anybody to believe a theory, and if it works the next change is much easier to introduce. Businesses that do this get three or four real changes a year. Businesses that write the manual get none.
The thing you may not want to hear
Sometimes the good branch is good because of something you cannot copy and would not want to.
A manager who works sixty hours. A long standing local relationship that came with them. A catchment that is genuinely different in a way the demographics do not show. Being propped up by one customer that will not be there in two years.
Finding that out is a good outcome even though it feels like a bad one, because the alternative is spending two years trying to reproduce something that was never reproducible, and blaming five managers for failing to.
Do not promote the manager out of it
The standard response to a strong site manager is to move them into a regional role so their approach spreads.
It usually costs you the good branch and does not produce five more. What made them effective was a hundred small decisions taken in one place with full context, and a regional role removes exactly that. Six months later the original site has regressed and the others are unchanged.
Better: leave them where they are, and use them to specify. They cannot articulate what they do in the abstract, but they can look at a specific gap in another branch and say what they would do about it. That is a different question and they can answer it.
The branch that is quietly worst
Everybody focuses on the top performer. The larger and easier gain is usually at the other end.
Moving the weakest site to the group average is normally worth more than moving the average to the best, because the weakest site has obvious mechanical problems rather than subtle cultural ones. Slow quotes, unanswered calls, work not followed up.
It is also less politically difficult, because you are not asking four managers to copy a fifth.
Where to start
Pick four operational measures, not financial ones. Response time, quote turnaround, follow up rate, rework.
Compare every site on those for the last six months.
Find the largest gap that plausibly explains revenue.
Change that one thing everywhere, for one quarter, and measure it.
If the four measures show no meaningful difference, the answer genuinely is people or catchment, and you have ruled out the cheap explanations in a fortnight rather than a year.
AI Optimize compares how your sites actually operate rather than what they bill, and shows each one the specific gap it can close. That work sits under Reporting & Data.
Related reading
How to Know If a Change Actually Worked
You changed the process, the number went up, and everyone agreed it was a success. There are usually four other explanations, and one of them is normally the real one.
What Multi Location Businesses Lose to Inconsistency
The second location rarely works like the first, and the gap widens quietly. What looks like a people problem is almost always an absence of anything written down.
WHAT WE BUILD