Five mistakes that keep derailing ERP integration projects

After dozens of integration projects we can say this with confidence: when a project slips or falls short of expectations, the cause is almost never a limitation of the API. The problems come from the same five places every time.

1. Leaving data cleanup out of scope

Duplicate debtor accounts, inconsistent SKUs, abandoned item records. All of this existed before the integration too, but nobody minded because a human eye tolerates it. Automation does not. Auditing the existing data before the project starts removes most of the surprises that would otherwise appear later.

2. Not deciding who owns what

Is the product description edited in Shopify or in Exact? If a customer address is updated in both systems, which one wins? Leave these questions unanswered and the two systems will start overwriting each other, and data will be lost.

Assign exactly one owner to every data field. Shared ownership makes conflicts inevitable.

3. Skipping the test environment

Building the integration straight into production looks like a time-saver. In reality you lose that time several times over the moment a bad bulk operation corrupts your customer records. An Exact test environment and a Shopify development store cost nothing; there is no reason not to use them.

4. Postponing returns and cancellations

The happy path — order, invoice, shipment — is always the first flow anyone builds. But what actually strains an operation is a returned order becoming a credit note, a partial shipment, or a cancelled payment. Leave those scenarios until later and you inherit a pile of manual corrections after go-live.

5. Not planning for monitoring

An integration doesn't finish on the day it goes live. API versions change, limits are revised, new product types appear. Without something watching sync health and reporting failed operations, you find out about a problem only when a customer complains.