A staged replacement for an ERP whose vendor had disappeared

ACME manufacture industrial handling equipment from a single site in Wakefield, with 340 employees. Their ERP vendor was acquired in 2022 and support for their version ends in 2027 — a deadline far enough away to keep being deprioritised, and close enough that a failed replacement would leave no second attempt.

The board had watched a previous system replacement over-run badly. The brief was explicitly about de-risking: they wanted to see the thing working on their own data before committing to a full programme.

What we did

We proposed three separately-funded phases with a genuine stop point after each. Phase one was a proof of concept: ERPNext deployed to a single environment, loaded with a slice of real ACME data, configured for one manufacturing cell. Eighteen days. The deliverable was a decision, not a system.

Phase two was an MVP for finance and stock — the two functions with the hardest 2027 deadline. This is where the eleven separate legacy data sources surfaced, several of which were departmental spreadsheets nobody had mentioned because nobody thought of them as systems.

Phase three is the full rollout, extending into production planning, purchasing and projects. It is running now, on the reference architecture the MVP established.

The EDI integration with ACME’s two largest customers was flagged as the highest-risk item during phase one and scheduled early in phase three, rather than being left to the end where a problem would have nowhere to go.

The outcome

Finance and stock went live on ERPNext ahead of the internal target, with the legacy system kept read-only for historic enquiry rather than migrated wholesale.

The phased structure did what it was meant to. After the proof of concept ACME changed two significant assumptions about their own processes, at a cost of a few days rather than a few months.

The 2027 support cliff is no longer a programme risk. It is a decommissioning date.