31 October 2026
In many plants, the story is the same. The MRP runs every night, proposes purchase and production orders, and in the morning planners rework them one by one in a spreadsheet. After a few months, nobody trusts the system any more. Management concludes it needs a new tool, or an APS.
Before going that far, look at what you are feeding the calculation. An MRP simply multiplies bills of materials by requirements, shifts them in time using lead times, and rounds them with lot-sizing rules. If those parameters are wrong, the result is wrong, whatever the vendor.
The five data items that derail an MRP
- Supplier lead times, entered when the item was created and never reviewed. The supplier delivers in six weeks; the system thinks two.
- Bills of materials that do not keep up with product changes. A replaced component stays in the calculation.
- Stock, when movements are recorded late or not at all. The MRP calculates on stock that does not exist.
- Lot-sizing parameters: minimum order quantities, multiples, safety stocks inherited from another era.
- Duplicate items: the same component created twice, on two sites or by two people, with two stocks and two replenishments.
What I saw on a Master Data programme
At Sanofi, I was PMO of a Master Data Management programme covering 30 of the group’s 60 plants. The work was not about cleaning files once and for all. It was about answering four questions for each piece of data: who owns it, which system is the reference, which workflow for creation, update and retirement, and which flow to the systems that use it.
One side effect illustrates what is at stake: identifying duplicate items and pooling purchases between plants reduced stock by €15M. Nobody had launched the programme for that. It is what happens when data becomes reliable.
Where to start
- Measure the gap between what the MRP proposes and what planners actually order. It is the best indicator of trust in the system.
- Choose a limited scope: the items that cost the most when they run short, or those that generate the most manual corrections.
- Name an owner for each data item, on the business side. IT owns the systems and flows, not the content.
- Put a few simple checks in place: lead times not reviewed for over a year, items with no movement, bills of materials changed without updating the calculation.
- Then expand, site by site or family by family, with the same rules.
What about the MDM tool?
It becomes useful when volumes and the number of systems require it. Choosing it before setting the rules and responsibilities means automating the mess. Governance first, tool second.
If your planners correct the MRP every morning, master data governance is probably the most profitable project you can launch. Let’s talk.