The systems bill nobody put in the model

A roll-up model is straightforward on paper. Buy the sites, consolidate the back office, apply the multiple to a bigger number. The consolidation line is the thesis, and it is usually the line with the least detail behind it.

What happens in practice is that the sites keep their systems.

Not by decision. By accumulation. The first two acquisitions are integrated properly, because there is time and attention. The third arrives while the second is still settling. By the sixth, integration means giving the new site an email address and a login to the group reporting pack and leaving everything else where it is. Every one of those calls was reasonable on the day. The result is a group running several booking systems, several charts of accounts, and a finance team closing the month by hand.

You can operate like that for a long time. The trouble starts at the point the thesis depends on, which is when you want to do something to all of the sites at once.

Change a price list. Report the same measure across the group. Move a clinician or a driver between sites without retraining them. Answer a buyer's question about performance by site, in the same format, for every site. Each of those is a small job in one system and an unbounded one in nine.

The bill arrives late and in the wrong place. It shows up as a finance team that cannot close on time, as a group report four people quietly distrust, as an acquisition that takes longer to onboard than the one before it, and eventually as a diligence process in which a buyer discounts the numbers because nobody can show them consistently. That last one is the expensive version, because it lands on the exit multiple rather than on the operating line.

The mistake is not failing to integrate. It is deciding when.

Integration is cheapest at the third site and dearest at the thirteenth, and the reason is not technical. At the third you are setting a pattern. At the thirteenth you are unpicking twelve exceptions, each of which somebody depends on and none of which is written down. The work is the same in kind. The volume is not.

So the useful question early in a roll-up is not whether to standardise. It is what the smallest set of things is that every site must do the same way from the next acquisition onward.

Usually that set is short. One chart of accounts. One way of recording the thing the group is measured on. One identity system, so that a leaver actually leaves. One written definition per reported measure, with a name against it.

Everything else can wait. A site can keep its own booking system for years without hurting anybody, as long as what comes out of it lands in the same shape as everywhere else.

The groups that get this right are rarely the ones that spent the most on systems. They are the ones that decided early which small number of things would never be allowed to differ, and then held that line through every acquisition afterwards, including the ones that were in a hurry.