Fast Growth Copies Everything, Including the Mistake

The par sheet in your first kitchen was built in a hurry, three weeks before opening, by a manager doing four jobs at once. It was wrong in one place. Not badly wrong. Wrong the way a number is wrong when it was set before the menu settled and nobody went back.

That kitchen ran anyway. The team learned to work around the number without ever quite naming it. The workaround became muscle memory, the muscle memory became how we do it here, and when unit two opened, how we do it here is what got taught.

You now have six restaurants. The wrong number is in all of them. It is in the training materials. The managers enforcing it are the good ones, the ones who follow the standard, because as far as anyone alive can tell, it is the standard. Nobody decided any of this. Replication decided it.

That is what growing too fast means. Not too many units, too quickly, by some number on a chart. Growth is copying, and copying is faithful. It reproduces the systems you built on purpose and the defects you never caught, at the same speed, with the same fidelity, and it does not know the difference.

What growing too fast scales

The comfortable story about growth problems is that new units get worse. The tenth restaurant misses the standard the first one set, the culture thins, the founder cannot be everywhere. All of it happens. But it flatters the original.

The less comfortable story is that new units are accurate. They faithfully run what the first restaurant does, which was never quite what its owner believes it does. Every operation carries a gap between the standard on paper and the standard in practice, and one restaurant absorbs that gap without anybody needing to see it. The founder is there. The instinct that would catch a drifting number walks the line four nights a week.

Copy the operation and you copy the gap. Except now the instinct that absorbed it is standing in one restaurant out of six, and the other five are running the practice, not the paper.

So the question that governs a growth plan is not the one that usually gets asked. It is not how fast can we open, or how many are funded, or whether the pipeline has sites. It is a plainer question: when something is wrong, how fast do we find out?

The loop that sets your real growth rate

Every operating mistake lives through two intervals.

The first is detection: the time between a wrong practice starting and someone with the authority to change it knowing. Not suspecting. Knowing, specifically, which number or practice is wrong and where.

The second is correction: the time between knowing and the last unit running the fix. Not the first unit. The last one, plus the training materials, plus the opening playbook, so the next restaurant opens without the defect instead of inheriting it on day one.

Add those two intervals together and you have the only growth governor that matters. A group that opens a unit every four months while carrying a detection loop that runs nine is manufacturing defects faster than it can find them, and that gap compounds in one direction. This is why grow slower is such useless advice on its own. Some concepts are built for speed, and the market rewards them for it. Speed is not the disease. Speed with slow detection is.

A mistake caught at two units is an edit. One conversation, one revised sheet, done by Friday. The same mistake at twelve units is a retrofit: a rollout plan, a retraining calendar, twelve managers who have to be talked out of a practice they were once graded on. The mistake did not get worse. It got copied, and every copy multiplied the cost of the fix.

Detection slows down exactly when you speed up

Here is the part that catches good operators, and it catches them because it is built out of their own strengths.

In the first restaurant, detection was free. The founder was the detection system: on the floor, tasting the food, reading the room, catching the drifting number before it survived a full week. That radar is why the first restaurant earned the right to grow at all.

Then growth does three things to it at once. It moves the founder away from the floor, because someone has to sign leases and hire directors. It puts the new units in the hands of people trained on the copy, who cannot flag the defect because the defect is the standard they were taught. And it pushes the first signal downstream into the P&L, where a wrong practice finally surfaces as a bad month, weeks after it started and one full accounting cycle after it mattered.

So the operator who was the fastest detection system in the building becomes, at six units, the slowest. Not through any personal failing. One set of eyes, six restaurants, and a signal that now has to travel through a spreadsheet to be seen at all. The instinct did not scale because instinct does not copy. Systems copy.

The replacement is not a better founder. It is detection that works without one: numbers reviewed on a cadence tight enough to catch a practice inside the week it drifts, a standard written precisely enough that a manager can tell the difference between the paper and the practice, and a reporting line where the person who spots the gap is rewarded for saying so rather than left to absorb it into another workaround.

Trace one mistake and time your loop

You can measure your own loop this afternoon, with a mistake you already fixed.

Pick the last operating error your group corrected. A wrong par, a broken prep sequence, a service step that was being skipped. Then put three dates on it.

The date it started, as best you can reconstruct it. The date someone with authority to change it knew. The date the last unit was running the fix.

The first interval is your detection speed. The second is your correction speed. Together they are the answer to how fast your group can safely open, because every unit opened inside that window is a unit built on whatever is currently wrong and not yet found.

Most operators who run this trace find the first interval is the ugly one, and that it ran not on any system but on coincidence: the founder happened to visit, the right person happened to mention it, a bad month happened to ask the question. If your detection history is a list of lucky catches, the loop is not fast. It has been fortunate, and fortune does not replicate either.

Growth multiplies what you have not found

The groups that scale fast and survive it are not the ones with flawless operations, because there are no flawless operations. They are the ones whose detect-and-correct loop runs faster than their opening calendar, so a defect gets caught at two units and edited, instead of caught at twelve and excavated.

Before the next lease, time the loop. If the answer is that you find problems when you happen to be standing in the building, the constraint on your growth is not capital, and it is not sites. It is that your operation only learns at the speed you personally travel, and the next unit is a bet that nothing goes wrong anywhere you are not.