Is your problem integration, or process?

The request almost always arrives phrased the same way: "we need to integrate the ERP with the shop". And in roughly half the cases, integrating the two systems solves nothing, because the real problem is not the cable but both ends of it.

The signal that separates the two

One question distinguishes the situations fairly reliably: who owns this data?. If you ask about an item’s price, or an order’s status, and two people from two departments give different answers while both are convinced they are right, you do not have an integration problem.

You have a data governance problem. And if you automate it without resolving that first, all you achieve is propagating the disagreement faster and to more places.

What to look at in week one

1. Who creates each piece of data and who modifies it afterwards. Not on the org chart: in practice. The system that "owns" a field is often not the one management believes.
2. Where things get corrected by hand. Every recurring manual correction is a business rule that is written down nowhere. Integrate without capturing it and the integration will break it.
3. What happens on conflict. If the ERP says A and the CRM says B, who wins? If nobody can answer, that decision will end up being made by whoever writes the integration — the worst possible person to make it.
4. How many times the same data gets re-entered. It is the most honest measure of the real cost of the mismatch, and it usually surprises people.

When it genuinely is integration

There are clear cases. Everyone agrees what each field means and who owns it, but the transfer is manual, slow or fragile: Excel exports, retyped orders, a supplier API that changed without warning. There, engineering genuinely solves it, and usually solves it quickly.

Why this gets skipped so often

Because assessment is hard to bill and integration is not. It is more comfortable to start connecting systems — there are tasks, there is visible progress, there is something to show — than to sit two departments down and decide who is right about the definition of "active customer".

But that conversation is going to happen anyway. The only variable is whether it happens in week one, while it is still cheap, or in week twenty, once things have been built on top of it.