Different ERPs. The same wall.

Thirty thousand codes in the system, of which a thousand actually move. The bills of materials are loaded with their levels but without quantities, and the availability column is empty on every row. To know what is in the warehouse you look at a sheet kept separately, and shortages are noted on paper forms.

To look up a component you need the exact code: one wrong capital letter or an extra space, and it finds nothing. The technical drawings sit on a separate server, and are opened by hand.

The vendor has put in writing that no standard APIs exist in the version in use. The system stores data in a 1990s database format. Read access requires an additional paid module, and the vendor normally doesn’t open everything: it builds boundary tables.

No scheduled data flow is provided. The production data lives in a third application, from yet another vendor.

The documented interface exposes the data read-only: how to write from outside is not documented, and the request to the vendor has gone unanswered.

It covered twenty per cent of the real needs, and the changes that came through had been designed for another customer: they slowed the work down instead of speeding it up.