Many of our clients have fast-paced operations and are experiencing rapid growth. They frequently struggle, in the midst of all of this, with figuring out their operational capacities and production schedules.
Here is a letter we recently sent to just such a client.
Dear [client EVP]:
I recall [from our conversations] that some of the items that concerned you were the matters of capacities and scheduling regarding your production. You said, if I recall, that you had several people who were attempting to tackle the whole matter of the capacity/scheduling conundrum with less than the success you would like.
First of all, let me say that this is not at all unusual. And there are many reasons for such poor outcomes.
Unfortunately, most folks today are still trying to schedule using methods and algorithms that were designed for an entirely different set of market circumstances—namely, for the 1950s and 1960s. In those days, in most industries, market demand still outstripped production capacities; hence, whatever was produced could generally be sold in the not-too-distant future. It was a world where product variety was typically small; supply chains were typically short; and the consumer’s buying options were fairly limited.
In today’s world, all of that has dramatically changed: with extended supply chains; huge product variety and customization offerings; and customers can buy from nearly anyone in the world who has a product to meet their requirements.
Nevertheless, even in “the good old days,” production scheduling had its challenges. These challenges are well articulated in Factory Physics:
[U]nfortunately, most real-world [factory scheduling] problems violate the assumptions made in the classic scheduling theory literature [and their implementations in software] in at least the following ways:
- There are always more than two machines. Thus [the] algorithm for minimizing makespan and its many variants are not directly useful.
- Process times and demand are not deterministic. …[We have] learned that randomness and variability contribute greatly to congestion in manufacturing systems. By ignoring this, scheduling theory is based on an unrealistic model of system behavior.
- All jobs are not ready at the beginning of the problem. New jobs do arrive and continue arriving during the entire life of the plant. To pretend that this does not happen or assume that we "clear out" the plant before starting new work is to deny a fundamental aspect of plant behavior.
- Process times are frequently sequence-dependent. Often the number of setups performed depends upon the sequence of the jobs. Jobs of like or similar parts can usually share a setup while dissimilar jobs cannot. This can be an important concern in scheduling the bottleneck process.
[Hopp, Wallace J., and Mark L. Spearman. Factory Physics. Long Grove, IL: Waveland Press, 2011.]
Using traditional approaches to capacity management and production planning, your schedulers need to know—with reasonable accuracy—the following factors for every order that will pass through your production floor:
- Queue time – the average amount of time a production unit (or job) spends waiting for processing at each work center, or to be moved to the next work center
- Setup time – the average amount of time a production unit (or job) spends waiting for the work center to be set up for processing the unit (or job)
- Move time – the average amount of time a production unit (or job) spends being moved from the previous work center to the current work center in a routing
- Process time – the averaged amount of time a production unit (or job) spends actually being processed by the work center
- Wait times
- Wait-to-batch time – the average amount of time a production unit (or job) spends waiting to form a batch for either simultaneous processing or moving
- Wait-in-batch time – the average amount of time a production unit (or job) spends waiting in a process batch for its actual processing
- Wait-to-match time – the average amount of time a production unit (or job) spends waiting for other components for an assembly operation
Furthermore, you will note the repeated use of the term “average amount of time.” This highlights the fact that, for these times to be meaningful in calculating schedules and capacities, your processes must be in statistical control—that is, variations from the mean are small and clustered around the mean, as opposed to variations being large and not clustered around a mean.
In most cases, we find that 1) organizations do not know these “average times,” and 2) their processes are not in statistical control. (And, as you so accurately pointed out while I was on-site, time-studies are of little value as they generally do not reflect the real-world scenarios of day-to-day production cycles.)
While traditional scheduling methodologies attempt to produce a “precisely right” schedule based on calculations made from these supplied times (see above), the result is a schedule that is always precise and always wrong!
As a result of these recognized failing in traditional methods, by the mid 1980’s a new scheduling methodology was emerging. This method is commonly referred to today as “DBR” or “Drum-Buffer-Rope,” and it was articulated in the well-known business novel entitled The Goal—A Process of Ongoing Improvement by Eli Goldratt and Jeff Cox.
All DBR requires in order to provide you with an accurate assessment of your capacities and schedules in an “approximately right” (maybe, slightly liberal) estimate of the processing time on your system’s CCR(s)—capacity constrained resource(s). The CCR becomes the “drum,” setting the pace for all other operations, and a “rope” ties the drum to other operations. The “rope” is actually a time buffer, which becomes a buffer of inventory (in some cases). (Inventory is simply a way of storing the combination of time and capacity it took to produce the inventory.)
In a simplified schematic, DBR takes your operations, and they turn out looking like this:
A quick assessment of your shop floor has suggested to me that you have a flexible workforce that may be assigned to various work centers or work orders on an as-needed basis. If that is true, then the “Shipping Buffer” model would probably be most appropriate in your circumstances (provided, of course, that your wave solder machine, for example, does not become an internal CCR at any point).
Using the “shipping buffer” model, your total capacity could be calculated as the total number of working hours (man-hours) available for production work over any given period of time.
The shop floor load can be calculated as the sum of number of units * drum-hours per unit of production time. In your case, your “drum-hours” would be an “approximately right” or “safe estimate” of the time it takes for shop floor processing, plus any other task-related times.
Calculation of the load is pretty simple and, since [your ERP system] doesn’t do this natively, it could be done in Microsoft® Excel™ using data extracted via Excel’s data access capabilities from [your ERP database]. It might look something like this:
In your case, “drum-hours” would be the total estimated load on your shop floor (as depicted in the “Constraint is the Market” scenario in the first diagram).
There’s more to this, of course, but I think you can get the general picture about how being “approximately right” using a DBR approach beats all attempts to be “precisely wrong” using traditional approaches to capacity planning and scheduling.
This approach gives you the wonderful combination of being both simpler and more accurate in assessing your real capacities to take on new work and setting estimated delivery dates.
We can discuss this further if you are interested. Just let me know.
I would also highly recommend the following reading. I think you will find some or all of it valuable and full of fresh insights for your management team.
Is your internal supply chain too complex? Are you facing big challenges trying to figure out production floor capacities and schedules that actually work?
Don’t feel alone. There are many in this same boat, and there are excellent and well-proven solutions available. We can help.
Leave your comment below, or feel free to contact us directly, if you prefer.