Many companies are so large or handle such complex businesses that it becomes difficult to deal with everything as a single unit. The number of decisions and the difference between them makes the business almost impossible to manage effectively single-handedly. Work needs to be divided with specialized people focusing on different things. Therefore, these companies are separated into subunits, usually named as departments, or areas.
Of course, each division handles its own businesses, usually responding to a global management level. However, this management level will not interfere in most decisions and practices, especially those related to daily operations, and will try to supervise based on local KPIs, representing the results of each area.
That creates a traditional problem in many organizations: every department will try its best to optimize its decisions, which happens on a local level. The scenario is usually reinforced by some management issues, such as the lack of communication between different departments and the mentality of area managers, who are mainly focused on their own KPIs.
This problem results from the fact that the "sum" of the local optimum does not correspond to the global optimum. This statement is applied to many contexts, such as the Theory of Constraints, notably when we think of production bottlenecks, and the Mathematical Optimization, where there is a universe of solutions, and some of them may be a local optimum but are worse compared to the global optimum, such as in the figure below. It is also applied to this management context, where the optimal solutions for each department combined do not correspond to the best for the company.
To exemplify this problem, think of a steel company. Normally, making a profit is its first objective and the company will try to maximize it, but let us suppose that sales are already defined and, therefore, the revenue is fixed and the company needs to minimize the total cost. Due to the complexity and differences of the several processes of steelmaking, this company has a fragmented structure, with different areas for each production process. For instance, a Coking department responsible for the production of coke to be used as one of the inputs at the Blast Furnace, a Blast Furnace handling the production of hot metal that will be one of the raw materials for the Steel Plant, and a Steel Plant dealing with the production of steel slabs.
Each of these departments will translate the company objective into its own: the Coke plant will try to spend the least possible money when producing coke, and the same will happen with the Blast Furnace and the Steel Plant. Since these processes interface, they will impose some constraints between them, and each one of them will try to minimize its own cost.
A classical example of constraints is the Blast Furnace imposing a minimal quality to the coke. Locally optimizing, the Coke Plant will try to produce the demand of coke with that exact minimal quality, avoiding "unnecessary" costs. Likewise, the Blast Furnace will produce the hot metal requested by the steel plant for the least possible cost.
However, these locally optimal decisions do not necessarily represent the best decision for the company as a whole. In the mentioned example, an optimal global solution might be the result of the production of higher-quality coke, with a lower content of ashes (which implies an increase of costs to the Coke Plant), that when used in the Blast Furnace will reduce the fuel rate and the need for other expensive materials such as pellet. This is one example of how a worse solution in one area can generate a better solution for another department and for the entire company, but many others can be identified: it is usually the result of the impact of one area into another, summed with process characteristics such as non-linearities.
It is important to notice that these decisions are dynamic, and will be modified depending on the context. Therefore, the global optimum may be constantly changing, and it is necessary to readjust each department’s decisions based on the best for the company at that moment.
A very common origin of such local optimum decisions is the definition of constraints and targets defined from one area to another when they interface, as mentioned above. Usually, the area will impose one desired value for a specific factor as a target, and this will eliminate a huge set of possible solutions that could generate better global results.
The solution to prevent this problem is to think globally when making decisions, especially those that might affect other areas’ processes and results. It usually starts by changing the management mentality to accept counterintuitive decisions and a local "loss" to obtain global gains. The usage of an integrated optimization model, which will consider the relations between different departments and the implications of each variable, can be the perfect supporting tool to provide better-coordinated decisions. It can also be used as evidence and argument to prove those counterintuitive decisions.
Cassotis has expertise in identifying these gain opportunities for large companies dealing with complex processes. Some examples are our solutions to the metals and mining industries. Do you think that your company might be facing this problem? Do not hesitate to reach out to our team!
Author: Cassiano Lima - Senior Consultant at Cassotis Consulting
Co-author: Fabio Silva - Senior Manager at Cassotis Consulting