From our work

One product in five was being sold below cost.

A regulated, high-volume manufacturer with several factories put every price next to its real cost and found that 20% of its products were being sold below cost. It now knows what each product costs and which ones earn, without replacing the system that runs production.

The situation

The system described itself, not the operation.

The company had grown for decades on a long-serving, on-premises ERP. Like most systems of its generation, it recorded what it was given, and over the years its data drifted away from the operation it was meant to describe. Identical products carried different references, so sales were missed and stock was double-held. In one of the main product families, about half the records lacked the dimensions needed to work out how much material a product consumed, so they could not be costed. Costing ran on a hand-typed workbook with no register of indirect costs, quoting ran in spreadsheets that saw neither the cost model nor the sales history, and product and customer margins were invisible.

Our diagnostic mapped these problems on the factory floor and in the accounts and sized each of them. It put the margin opportunity from knowing the real cost of each product at more than a million euros a year, and the working capital tied up in duplicate references in the hundreds of thousands of euros. The leadership team used that ranked picture to decide what to fix first.

What we did

A backbone, built one capability at a time.

The operation could not stop for a big-bang migration, so nothing was ripped out. We installed the integration backbone alongside the existing ERP and moved ownership of the data across one capability at a time, each with a single, unambiguous system of record. The old system keeps doing what it still does well, and is retired at the pace the operation can absorb.

Three of us built it. Work began in May, the first capabilities were in production by the end of July, and by August the platform ran to several hundred thousand lines of code with close to four thousand automated tests. AI agents did much of the routine engineering, under our direction and review; nothing an agent produces reaches production without a person approving it.

  1. 01

    Items first

    One governed item master replaced thousands of drifting records. Every product received a single identity, a defined owner and the attributes the operation actually uses.

  2. 02

    Then bills of materials and routings

    Versioned, governed bills of materials and routings replaced the workbook, so production starts from the right materials and costs roll up from live operational data.

  3. 03

    Then costing and margin

    Full absorption costing with indirect costs included, rolled up line by line. Every price was put next to its real cost: 20% of the products were being sold below cost, and the list of them, with how much each one lost, became something the leadership team could act on. Gross margin by product became a dashboard.

  4. 04

    Then planning and execution

    Production planning against real capacity and real routings, customer orders promised against what the factories can do and work tracked from release to dispatch with actual costs captured as they occur. Built, and being brought into use.

Measured so far

The numbers are the point.

The backbone is live in production for the item master, bills of materials, routings, costing and margin. These figures are measured results of the data consolidation and the costing work.

20%
Of products were being sold below their real unit cost.
12,480 → 4,980
Legacy article records consolidated into governed items across 76 item types.
553
Duplicate product groups (1,562 records) found hiding in the legacy data and resolved.
€29M
Of yearly revenue whose products are now described truthfully, end to end.
A note on restraint

The boring route, chosen on purpose.

Connecting the backbone to a legacy ERP can be done in clever ways and in careful ways. The most sophisticated change-capture route available to us carried a small but real risk of overloading the old system's database, and that database runs production. We designed the clever route, understood its failure modes and chose the careful one: read-only access to the ERP, no message broker in the path, every import passing through the same business rules as a manual entry and every write back to the ERP going through its own interface.

Every rollout is rehearsed on a copy of production with a rollback ready before it touches the real one. That discipline cost us some engineering elegance and bought the client certainty that nothing we do can take their operation down. We consider that the right trade in every case. Technology, AI included, earns its place in a factory by the value it delivers and the risks it avoids.

Where it stands

Toward independence.

The engagement continues: capability by capability, the backbone is taking on more of the operation, and the client is building its own software team to run and evolve it. We drafted the first engineering role, are running a search that has drawn over 130 applications and wrote the onboarding path a new engineer follows, from the domain model to their first delivery. That is the destination we work toward on every engagement: an operation the leadership team can see truthfully and run without us day to day, with Grandhelm remaining at their side as advisors.

This is what steps 02, 04 and 05 of our sequence look like in practice: margin as a fact, the business on systems and a team that runs them.

Next step

Do you know what each of your products earns?

If parts of this story sound familiar, a conversation is the right first step. We will tell you honestly whether we can help.

GrandhelmValue Creation Partners

Value creation partners to the owners and investors of mid-market companies. We increase EBITDA through operational improvement, with AI that gets there in a fraction of the time, and we prefer to be paid on the result.

Where we work

  • Eindhoven, The Netherlands
  • London, United Kingdom
  • Porto, Portugal

© 2026 Grandhelm. All rights reserved.

Privacy