Using Marginal Cost Curves to dynamically optimize your Order Book

Back to list


A steelmaking plant in Germany with an annual production of 4 million rolled products.


The objective was to find an efficient approach to derive the plant’s marginal cost curve for all steel grades and finished product families. This information is critical to assessing the profitability of new orders.


Collaboration with N-SIDE started as a regular SCOOP solution implementation, with the main focus on reducing production costs, in particular the raw materials aspects.

However, the SCOOP solution also contained detailed unit costs, detailed product compositions and a special feature called sensitivity analysis. These features soon led our client to consider using the tool to draw marginal cost curves of the plant.

Such analyses used to be performed by an entire department and took several hours to be finalized, without any guarantee that mistakes would not arise, due to the manual nature of many operations. This limited the ability of the plant’s management to explore a large range of scenarios, which is precisely what is indispensable if you want to identify the most interesting cases and build confidence around the best solution.

Marginal cost curves, while well understood from an academic point of view, remain very tricky to grasp, let alone to draw and exploit in real life within the industry.

Many managers and cost controllers tend to rely more naturally on average cost curves. These are based on a “look back” business reporting logic: the higher the volume produced, the lower the fixed cost per unit of production, hence the rough rule is that one should produce as much as possible. This reasoning is valid up to a point, but becomes potentially value destructive due to constraints within the steel plant. These can be linked to (1) limited availability of certain excellent raw materials, (2) limited internal capacities of certain equipment, or (3) limited demand for certain very profitable grades.

Thus, as production volumes rise, operations face more and more of these constraints, forcing them to switch to more costly alternatives, such as less attractive raw materials, combining internal production of coke or sinter with external purchases, etc. As a result, the additional cash costs of the later tonnage will tend to be significantly higher than for earlier production. This is the marginal cost curve concept. In fact, at some point in the curve, the marginal cost of the next ton of product could start becoming higher than its sales price. Producing beyond this point means that the plant will start losing         money. This approach is “forward looking”, focusing more on future management decisions in a given context: should I produce more or less, and if I do, what will be the impact on my profits?

If steel were a simple commodity, this problem would already be quite complex to solve. However, a plant usually produces hundreds of different grades, through a wide variety of equipment combinations.

With the average cost approach, one tends to work through allocation keys, distributing the various cost items to individual grades. The result is most often a classic inverted U-shape with a few grades making most of the profit, a weak middle not contributing much and a few trouble-makers destroying quite a lot of profit. The main issue with this approach is that it is static: because of the rigidity of the allocation keys, one cannot figure out what will happen to total profits if the trouble-makers are eliminated. Some of the costs will not be negated by eliminating these grades, but their revenues will. Worse, if one imagines increasing the production of some other more profitable grades, the “excel-approach” becomes intractable.

The marginal approach solves all of these issues by definition: when a part of the volume is added or removed, only its marginal impact on profits is computed, without the need to rely on complex allocation keys.

The target therefore was to develop a practical tool that could calculate the “direct product marginal profitability” at any given overall plant production level.



N-SIDE started with its SCOOP tool and extended the built-in sensitivity analysis.

First of all, by letting the system calculate optimal operating costs for several production levels (say from 60% to 95% capacity), one could easily derive the marginal production cost curve.

On top of that, at each optimal point on this curve, SCOOP was able to calculate the marginal contribution of each product grade, simply by calculating the derivative. A user-friendly data extract was established to enable the drawing of graphs for easy management interpretation.


The following benefits were quickly highlighted by plant managers:

  • Identification of marketing and sales priorities to redirect efforts to the most valuable products;
  • Pricing adjustments to increase direct product profitability;
  • Identification of equipment-related bottlenecks: The productivity model points out the equipment concerned. By increasing the capacity of such equipment and comparing scenarios, one is able to evaluate the ROI of a small investment project.
  • Determination of the target production levels;
  • Quick reactions to new client orders: profitability of the order can be assessed in a short period of time, enabling either sales people to negotiate better conditions or operations people to adjust process parameters;

Beside these economical benefits, SCOOP also has the main advantage of providing a common tool to help in aligning managers from very different horizons who may serve conflicting objectives: Purchasing vs. Operations vs. Sales.

Print Friendly
Back to list