Production Capacity Is Not A Number

If you have any experience with production planning, I believe you are frowning after reading this title. To demonstrate what I mean, I will first differentiate between planning systems and scheduling systems.

The first (planning systems) are those MRP (Materials Requirement Planning) type, developed in the early 50's to help in the management of GE and Rolls Royce manufactures, later expanded to other manufacturing companies and transformed into shelving systems, being commercialized since then; today they belong to the heart of ERP systems. Obviously, MRP systems have evolved and become extremely powerful tools to perform demand explosion into purchase requisitions, production orders and material requirements, performing a primitive sequencing of productive resources. To this end, manufacturing capacity is grouped into “time buckets”, as they are commonly referred to, and for each “time bucket” a numerical value is assigned that represents the capacity, usually in hours. To better understand this concept, take a look at the diagram below where we have two months with thirty days each, with a work regime of eight hours per day, from Monday to Friday:

First month has a total of 160 hours and the second has 176 available hours; these hours can be understood as the total capacity of a given productive resource in the MRP view. Generally, "time buckets" have constant value of hours to represent the capacity in that period, and this value is either decided by the management or calculated by the MRP based on the work shifts, as demonstrated in the above example. Considering a shift of eight hours a day and five working days a week, 40 hours a week or 168 hours a month is commonly defined, but obviously that depends on the grouping taken by the MRP. The planning system, when it considers the capacity, starts using these 168 hours according to the delivery date of the activities, allocating the highest priority order first, then the second and so on, performing backward planning from the delivery date until fulfilling 100% of those 168 working hours. When this happens, the next highest priority operation is allocated at the previous “time bucket”. Below, we have 4 graphs to exemplify the difference between planning and sequencing. As a simplification, we use a “time bucket” with 160 hours and orders that requires 20 hours to be processed (adding the preparation and processing time) and, in addition, we have 3 distinct attributes that are characterized by the colors yellow, red and blue:

MRP systems understands, then, that we can perform 8 different activities. This can be seen in Graph 1, where the operations of production orders (named A to H) were allocated until they occupied 100% of the capacity for that “time bucket”, regardless of the attributes associated with the production orders. A sequencing system cannot apply those simplifications; activities should be allocated to the productive resource as they will be performed on the shop floor, or at least as close to it as possible. We have graph 2 with this representation, where the black bar identifies preparation time (also often called changeover or setup) for attribute change, calculated at 10 hours for any change, except from red to yellow, where we have 30 hours.

Still in graph 2, we can see that it would actually take 180 hours of the productive resource availability that month to be able to execute the sequence suggested by the MRP. Graph 3 would be a sequencing alternative where production orders are grouped by attribute, avoiding the excessive need for changeovers. In this case, we would have a “leftover” of 50 hours to absorb other production orders or even to allocate preventive maintenance or testing of new products/equipment. Graph 4 goes further in the sense of trying to exemplify manufacturing restrictions, such as stoppages due to lack of labor, tools and raw materials. These situations are commonplace when production planning and control does not plan considering these and other restrictions, in addition to constraints that may be inherent to its production process.

Coming back to the statement that "capacity is not a number" and looking at the examples above, we can extrapolate and ask questions like: What if we only had yellow orders to execute? What if the mix of products to be manufactured were prioritized, in such a way that we would have only 1 operation per attribute? Could we anticipate operations with less critical delivery dates to try to mitigate any problems, without impacting the delivery of priority orders? What if a key customer required prioritization on your delivery for a production order that is planned to start only in the next period? What if the region where the factory is located was going through a pandemic, temporarily removing a good portion of its workforce? What if critical equipment was running below its standard efficiency due to technical issues caused by a lack of sufficient predictive maintenance? There are so many questions whose answers are not easy to obtain that simplifying the planning of a manufacturing unit to a number does not seem to me the best way to obtain a plan that adheres to reality.

Let's go now to a second example. We still have our “time bucket” of 160 hours and operations with a total duration of 20 hours per operation; however, now we have 2 periods: current (period 0) and next month (period 1). In addition, it will be necessary to allocate 9 operations, all of which must be completed by the end of the next month (period 1), according to the delivery date agreed with the customer. The current period has already occurred, so it is not possible to allocate operations in this "time bucket", which is represented by the gray color in the scheme below:

For this situation, MRP starts allocating orders from period 1 and when it allocates 100% of the 160 hours, which is the capacity defined for the monthly time bucket, it tries to allocate the 9th and last operation, but there is no availability, since the previous period has already occurred. A production sequencing, in turn, will obviously not sequence activities in the past, so it delays the yellow production order, as shown in the above schema. As MRP systems have evolved, there are different configurations that make it possible to reduce some of the problems mentioned in this article; however, it will always apply simplifications that harm when you want to develop a process that adheres to reality.

As MRP-style planning systems are mostly used for medium and long-term plans, the arguments that "we don't need precision" and that "sequencing that consider this level of detail do not bring positive results for the industry" are common. The above statements are based on two key arguments:

• The further away the production order is to be started, the lower the quality of the data;

• Long-term plans do not need to be detailed, as the environment will vary greatly until these production orders enter a short-term horizon.

On the first point, I no longer believe that this argument is valid for at least 10 years now. Although you still find many manufacturers defining production routing and bill of materials manually, this is no longer an excuse for not wanting to improve. There are systems today, both for engineering and configuration / quotation, that can solve this problem. Even for ETO (Engineer to Order) companies, this argument is no longer valid; think like this: a customer demands a solution, the engineering team designs the product to meet that solution, this design goes to a quotation team that has the role of pricing this product, and then the planning team has the role of planning and saying when it is possible to deliver that order. Modern solutions can, with proper configuration, translate the engineering design and generate the production routings and bill of materials needed to produce the order and, consequently, generate the final price. So, if a design change is necessary, the process is repeated, but the quality of the data will always adhere to the design made by engineering. The problem is much simpler for CTO (Configure to Order), MTS (Make to Stock) or MTO (Make to Order) industries.

As for the second point, it is easier to answer it with a contrary argument: the greater the planning adherence with the reality of the shop floor, the greater the management's ability to make assertive decisions. In my point of view, the main differential for continuous and quality management lies in being able to look at a detailed production plans and knowing what needs to happen to improve customer satisfaction. This is only possible if production plans and schedules are trustworthy.