The problem
Replacing an ERP is not a software project. It is replacing the system of record underneath a business that has to keep running while you do it.
An industrial SME carries the same data problems as a prime, without a prime’s back office to absorb them. One system holds clients, suppliers, invoices, orders, routings, bills of material and works orders, and the team maintaining it is small. There is no parallel-run department, no spare month, and no appetite for a cutover that half-works.
The hardest data is not the customer list. It is the routings and bills of material, because those encode how things are actually built. Where the old system’s version and the shop floor’s version disagree, migration forces the question nobody had needed to answer.
What we did
We ran it end to end as PMO and data engineering: planning through to production. Owning both halves mattered: the schedule and the pipelines were the same problem, and splitting them across two suppliers is how migrations lose their cutover date.
We designed and built the conversion pipelines that carried the full industrial scope across, structured so that business-rule changes discovered mid-project could be absorbed and re-validated quickly rather than reopening the plan.
The engagement did not end at cutover. We continue to support Deshons on industrial process improvement and on evolving the BI layer that sits on the migrated data, which is also the strongest evidence the migration was sound, since we are still the ones living with its consequences.
What made it hard
Validation, not conversion. Moving rows is mechanical; proving the target is right is the work, and “right” means the people who run the shop floor recognise their own processes in the new system.
Reconciling a manufacturer’s records means confronting the places where the recorded routing and the actual practice have quietly diverged over years. You cannot migrate your way around that. You have to surface it, get it decided, and carry the decision, which is the same discipline as reconstructing what is true from records that disagree, in a different costume.