The context
A metals manufacturer buys steel in standard stock lengths and cuts it into the sizes its customers order. It sounds simple, but every cutting plan is a puzzle. A single bar can be cut into many combinations of lengths, and every combination leaves a remnant at the end. Remnants that are too short to use are scrap.
In a business where raw material is the largest cost, the difference between a good cutting plan and an average one goes straight to the margin.
The challenge
Cutting plans were drawn up by hand by experienced planners. They were good at it, but the number of possible combinations grows explosively with the size of the order book, far beyond what anyone can evaluate by hand. Planners naturally fell back on familiar patterns that worked, rather than the combination that wasted least.
The result was steady, invisible waste. No single plan looked wrong, but across thousands of orders, a significant share of purchased steel ended up as offcuts.
What we did
We treated cutting as what it is: a well-studied optimisation problem, known in operations research as the cutting stock problem. Given the open orders and the stock available, the task is to find the set of cutting patterns that fills every order using the least material.
We built an optimisation model around the manufacturer’s real constraints: the stock lengths it holds, the tolerances and blade losses of its machines, order priorities and delivery dates. Rather than replacing the planning process, we fitted the model into it. Planners load the open order book, the model proposes the cutting plan, and planners review and release it as before.
Before changing anything on the shop floor, we ran the model on past order books and compared its plans with the plans that had actually been used. That comparison showed, in kilograms and in money, how much material each approach consumed for the same orders.
The result
The optimised plans cut material waste by 35%, with every order still filled to specification. Because the saving is in raw material, it shows up directly in cost on every production run, and it grows with volume.
Why it worked
- The right method for the problem. This was a mathematical optimisation problem, not a prediction problem, and we solved it with the tools designed for it.
- The planners stayed in charge. The model proposes; people approve. Adoption was immediate because nothing about the team’s role changed except the quality of the starting plan.
- The saving was measured before it was claimed. Running the model on past orders proved the result on the company’s own data.
Similar optimisation applies to sheet cutting, nesting, load planning and production scheduling. See our manufacturing and supply chain practice.