It is one of the most repeated disappointments in mid-market business: a significant ERP investment, a difficult implementation, and two years later the company still runs on spreadsheets attached to emails. Management concludes the software was wrong, or the implementation partner was weak, and begins evaluating a replacement — which will fail in the same way, for the same reason.
The reason is rarely technical. An ERP is an obedient machine: it automates the operating model it is given. If approvals were informal, if item masters were nobody's responsibility, if processes lived in the heads of long-tenured employees, then the ERP faithfully digitised that informality — and informality at digital speed is chaos with better fonts.
What the ERP actually exposed
A failed implementation is best read as a diagnostic, because it reveals precisely where the operating model was undefined. Every workaround the team built is a process that was never truly standardised. Every report nobody trusts marks data that has no owner. Every module abandoned after go-live shows a decision the organisation was not actually making in a structured way, however the workflow assumed otherwise.
Seen this way, the disappointment holds real value. The business has paid — admittedly at a painful price — for a complete map of the gap between how it believes it operates and how it operates.
Discipline precedes digitisation
The repair sequence matters and is routinely inverted. Data ownership comes first: item masters, vendor masters, customer masters, each with a named owner and rules for creation and change, because no system survives corrupted masters. Process definition comes second — not thick documentation, but clear answers to who decides, who approves and what happens on exception, for the twenty processes that carry the business. Only then does reconfiguration make sense, and it is usually reconfiguration of the existing system rather than replacement; the fastest ERP recovery is almost never a new ERP.
The test of recovery is blunt: the month the parallel spreadsheets die. As long as the real business runs outside the system, the system is an expensive ledger. When exceptions become visible inside it — when a blocked order or a breached credit limit surfaces in the workflow instead of a phone call — the investment starts returning.
The management implication
Technology spending amplifies the operating model it lands on. A disciplined business gets compounding returns from modest systems; an undesigned one converts sophisticated systems into shelfware. Which means the ERP question that boards should ask is not which platform, but a harder one: is this organisation defined enough to be automated? Answering that honestly, before the next licence is signed, is worth more than any vendor evaluation.