The management of capacity has always been a contentious area of production planning. This page is intended to provide a brief background, covering the key elements and developments since the appearance of the forerunners of today’s ERP systems in the nineteen-seventies.
Capacity Requirements Planning (CRP)
The earliest systems conforming to the Manufacturing Resource Planning (MRPII) model employed the CRP approach. The mechanism is in fact relatively simple. CRP looks at all production orders and planned orders from MRP, calculating start and finish dates for each operation using either a ‘backward’ pass from order due date or ‘forward’ pass from today or the order start date. It then accumulates planned time by work centre by period and reports the totals for planners to address mismatches between load and capacity.
Some of the time elements in the routing for each item are determined within the process definition – set-up time, run time and possibly some waiting allowance after completion of an operation (to allow, for example, for cooling). Obviously, the calculation of operation dates also has to allow some time for movement. In the days of the earliest systems, the approach was aimed primarily at large businesses with large manufacturing sites and so movement from one work centre to another had to be scheduled as it could take a significant amount of time, typically held in a matrix within the system.
Even this is insufficient allowance since no plan can be based upon the assumption that an order arriving at a work centre will start work immediately upon arrival. The system has to allow for some time spent queuing at a work centre so the packages hold a queue allowance field on each work centre record. These have to be established by planners – who should be able to recognise the likely length of queue at each work centre. Those that are heavily loaded are more likely to have lots of orders waiting to be worked on and so should perhaps have a longer queue allowance defined. Of course, when setting the operation dates the CRP logic allows for all orders to arrive ahead of the planned start date and uses the queue allowance for this. If we plan to have a 12-hour queue at a work centre, then all orders will be planned to arrive 12 hours early and there will be a 12-hour queue. The queue allowance is a self-fulfilling prophecy.
Since each order has been planned independently (no effort has been made to allocate the available capacity), then some work centres are, inevitably, overloaded at times. Because a CRP system takes no action other than to report overloads the approach this approach has often been quoted as being based on infinite capacity. It is often referred to as ‘infinite planning’.
Its detractors claimed that MRPII systems were based around infinite capacity and the approach was therefore wholly invalid. Its proponents argued that MRPII was in no way based on anything as ludicrous as infinite capacity. They pointed out that CRP systems warned of overloads specifically so that planners would take steps to maintain a valid plan – that is, one which recognises the finite capacity of manufacturing resources. They also argued that presenting information to planners who could use judgement was better practice than trying to define rules within a system and effectively trusting our futures to some clever ‘black box.’
The finite camp came back and argued that MRP is based around fixed lead times for items when in fact lead times are anything but fixed. They are wholly dependent upon the workload of the plant and the nature of the work at any particular time. To avoid too much contention planners had to build excessive queuing allowances into the lead times used by MRP. This in turn extended lead times to market and increased the level of work-in-progress, which in turn led to even more contention. Lead times, they said, are like queue allowances; they are self-fulfilling prophecies. (And, of course, in this they were absolutely correct!)
Some software companies, as the processing power of computers increased, began to develop capacity management systems based around finite capacity. This, of course, required a fundamentally different approach. Where infinite CRP simply accumulated times against a work centre, a finite approach had to identify when a particular order would be using each work centre and this required some logic to determine which order took priority in the event of contention.
Finite planning thus required some form of ‘forward pass’ scheduling whereby for each work centre the system looked at the orders waiting for this resource and decided which should be processed first. The system then performed this task for all the other work centres and looked ahead to the next time any of these work centres would be available. It then identified which orders would be in the queue at that work centre at that time and decided which should be the next to be worked on. It ran through this process for all the work to be carried out and generated schedules (‘dispatch lists’ or ‘work-to’ lists) for each work centre.
Some system providers fell into one camp or the other whilst others offered both approaches. IBM, whose COPICS product had been the forerunner of the standard MRPII approach, offered CAPOSS (Capacity Management and Operation Sequencing System) as an alternative to the capacity module in their own package, or a as a ‘bolt-on’ to other business systems.
Of course, the difference at the end of all this was that where we had an unachievable plan and CRP produced graphs showing overloads, the finite systems showed no overloads (if this were possible given the demand), but reported that unfortunately some orders were inevitably going to be completed late. Textbook MRPII followers argued that both systems lead, in reality, to the same thing – an action on a human planner to go away and sort out the problem. This could involve a complete change in the manufacturing plan or may simply involve a few extra hours capacity which can be achieved through overtime. (Or may necessitate some movement of work from one piece of plant to another, or perhaps to a sub-contractor.)
Thus, they argued, CRP was the better approach in that the cause of the problem (the overloaded resource) was immediately visible. In a finite scheduling system further detective work is required to move from a list of orders which are going to be late to the identification of the overloaded piece of plant. They also pointed out that we could define priority rules for a scheduling system to follow blindly but sometimes a problem could easily be averted by a simple piece of judgement on the part of a planner. One thing computers systems don’t have, they argued, is judgement. This argument prevailed within the MRPII community and CRP became the established capacity management technique within MRPII, and hence later in ERP and standard Supply Chain approaches – though the debate went on for many years and sold many books as well as filling seats at numerous conferences.
Optimised Production Technology (OPT)
In the early 1980s, however, Eliyahu Goldratt challenged all the accepted wisdom of planning and control. He argued that there are some fundamental rules that were being ignored. This subject in itself is worthy of considerable exploration since he challenged some of the basic approaches to managing and measuring manufacturing operations. He pointed out that time saved at a bottleneck is time saved and that time saved elsewhere is a mirage. We can all see this in other walks of like – for example on toll roads. The Dartford crossing on the M25 undoubtedly limits the volume of traffic passing between Kent and Essex to something like two vehicles per booth per minute. If we wanted to increase the traffic flow would we increase the width of the motorway in the five miles leading up to the crossing, or would we provide more booths? Hopefully we can all see that widening the motorway would simply get the traffic to the crossing more quickly and increase the length of the queues at this bottleneck.
Goldatt also argued that the only point at which an operation adds value is when it leads to cash for the business. Continuing to run a machine when the items being produced will go nowhere because of a bottleneck further down the line adds no real value at all. This, of course, was contrary to all the accepted wisdom within the accounting conventions in use. Manufacturers would often keep producing items that were going nowhere because doing so they were ‘recovering overheads’.
‘The Goal’ (ISBN-13: 978-0884270614) by Eliyahu Goldratt and Jeff Cox and published in 1992 explores these issues in the form of a novel and remains an excellent read.
Another of Goldratt’s arguments was that the MRP and MRPII approaches were inherently wrong in that they scheduled materials and capacity independently. MRP calculated the requirement dates for in-house components and their raw materials; CRP then looked at the manufacturing resources needed to make these components. In many cases a potential overload would be reported and the cure may be to pull the order forward into a period when capacity is available. However, by this time MRP will have led to the ordering of the raw materials so when we have changed the production plan the next MRP run will generate re-schedule messages for the materials plan.
Goldratt’s OPT package looked at materials and resources together to establish a plan which took account of all constraints – in other words, an optimised solution. It identified the bottlenecks in our plant (which, Goldratt taught us, govern our throughput) and set our production and material plans simultaneously around best use of these resources. It was never quite apparent exactly how Goldratt’s system performed this optimisation and in fact this secret always appeared to be closely guarded.
Advanced Planning & Scheduling (APS)
The OPT package, despite its obvious attractions, did not actually sell as well as it might have, but it laid the foundations for other companies to build upon as computers became capable of ever-increasing levels of complexity in their scheduling algorithms. As more packages came into the market following these principles of optimising production schedules, making best use of bottlenecks, planning materials and resources simultaneously and allowing users to set up their own rules as to what mattered most in their environment, these packages became known generically as Advanced Planning and Scheduling, or APS, sometime in the mid-1990s.
As ever in such matters the world appeared to divide into the MRP and APS camps. The arguments against MRP are well known but its defenders raise concerns about APS, primarily that it works around pre-defined rules, which may not always give the optimum solution in all circumstances, and that people will never fully understand what the package is doing in generating the particular sequence. They argue that it encourages planners to discard judgement. (In this they have a point. If we use any optimised planning system then we need each area to follow the plan religiously; if one work centre cuts across the recommended sequence then all subsequent areas will not receive orders as planned and the whole plan is meaningless.)
So, then, where next? When respected proponents from all the various camps propound their theories with conviction and credibility what are we to do? As ever, the answer is understand the issues and use our own experiences to guide us. Of course, when we are looking at the software options available in this field and hearing details of package capabilities we can remember the infallible rule of how to tell when a software salesman is lying. It’s easy – his lips move.