ERP systems

The defining risk is the cutover. Unlike most software, an ERP cannot easily run alongside the thing it replaces — inventory cannot be authoritative in two systems, and a finance period cannot be half-closed in each. That forces a moment where the business commits, usually over a weekend, with a rehearsed fallback and a period of parallel data entry that everyone hates and nobody should skip. Projects that fail generally failed here, not in the build.

The second trap is customisation. Every department will ask for the system to match its current process exactly, and granting all of those requests produces something that cannot be upgraded, cannot be supported and encodes habits that existed because the old software forced them. The discipline is to distinguish a genuine competitive difference in how you operate — worth building — from a workaround that has been mistaken for a requirement, which is most of them.

Integration is where the effort actually goes. An ERP is only useful if it agrees with the warehouse system, the e-commerce platform, the bank feed, the payroll provider and whatever the accountants use. Each of those is a contract about what a record means and when it is considered final, and disagreements between them show up as an inventory number nobody can explain. Modelling those boundaries carefully is the difference between one system of record and five arguing.