Nothing breaks at two branches. Everything breaks at four.

The second branch goes well. That is the problem.

It goes well because you can hold both of them. You walk in, you look at the shelf, you know the number is wrong before anyone tells you, you fix it in the corridor. Nothing is written down because nothing needs to be. The system is you, and at two sites you are an excellent system.

So you open a third, and it is harder but manageable. Then a fourth, and inside about two months everything is on fire at once: the cash does not reconcile, two branches are quoting different prices, stock says one thing and the shelf says another, and nobody can tell you which day it started.

Nothing about your staff changed. The arithmetic did.

Output adds. Coordination multiplies.

Fred Brooks worked this out about software teams, and it is one of those results that turns out not to be about software at all.¹ If you have n things that must stay in agreement, the number of pairs that must be kept in agreement is n times n minus one, over two.

Two branches: one pair. Three: three pairs. Four: six pairs. Six: fifteen. Eight: twenty-eight.

Your revenue went up four times when you went from two to eight. The coordination went up twenty-eight times. Nobody planned for the second number because nobody was ever shown it.

Why walking between them stops working

What you had at two branches was not a small system. It was no system, and it worked because informal coordination is genuinely excellent up to a point and genuinely impossible past it.² The way a group stays aligned by everyone roughly knowing everything has a ceiling, and the ceiling does not announce itself. You do not get a warning that you have crossed it. You get a Tuesday where nothing adds up.

You did not lose control at four branches. You ran out of a mechanism that had no four-branch version, and nobody told you it had a limit because it never looked like a mechanism.

What gives way first

This ordering is reasoning, not measurement, and I want that clear before you plan around it. We have not followed a set of businesses and recorded the sequence. What follows is an argument from how often each thing is touched and how fast a mistake in it becomes permanent.

1. Cash goes first. It is touched every day, by the most junior people, at the busiest moment, and a discrepancy from Tuesday is effectively unrecoverable by Friday. There is no shelf you can go and look at.
2. Price and promise go second. Two branches start quoting differently, or one honours something the other does not. Customers find this out before you do, and they compare.
3. Stock goes third. It breaks loudly but it is recoverable, because the truth is physically present and can be counted.
4. People and rosters go last.Painful, visible, and the one everybody blames first — usually while the cash problem that started three months earlier goes unnoticed.

If that ordering is right, it has an awkward consequence: the thing to fix first is the one with the least drama attached to it.

The bill for waiting

Almost everybody builds this at four branches, and almost everybody should have built it at two. The reason is not stupidity, it is that at two branches the payoff is distant and uncertain while the pain is immediate, which is the exact shape of a decision that gets deferred.³

The difference in price is large and it is not the software. At two branches you are writing down how things already work, calmly, with one person’s help, in a quiet month. At four you are doing the same work during an emergency, with staff who have lost confidence, while somebody is chasing a cash difference from six weeks ago. Same project. Completely different month.

What to do at two, if you are at two

1. Write down the one that would be worst to get wrong. Usually the daily cash close. Not a system — a page. What gets counted, by whom, by when, and where the number goes.
2. Give every figure one home. For each number that matters, name the place that wins when two places disagree. Say it out loud to staff. Most branch chaos is two sources and no rule.
3. Count your pairs before you sign the next lease. Work out n times n minus one over two for the branch count you are about to have. That number is the coordination you are buying along with the rent.
4. Test it by leaving. Take a week off. Whatever breaks is the thing that was living in your head, and now you have a list instead of a feeling.

If you are already at four, the order above is your order. Cash first, and do it before the roster, however loudly the roster is complaining.