Main content


S&OP, Master Scheduling and Optimization

Chris Gray, Gray Research


If it wasn’t bad enough that companies think they can simply summarize the MPS and call it the S&OP, sometimes companies are lured into implementing some type of advanced planning system (APS) to optimize the master schedule prior summarization.  And they mistakenly assume that they can re-optimize both the detailed plans and the aggregated S&OP supply plan each and every S&OP cycle. 

The basic theory is that since the detailed data – demand, supply, equipment and labor availability, setup constraints, machine and labor capacity constraints, transportation distances and costs, production cost data, and other constraints, etc. – exist in the system, the computer and clever programming can produce or reset all the demand, distribution and supply plans effectively taking humans out of the equation.

Then of course – with no humans in the loop – the detailed plans can be adjusted in almost real-time as new customer orders are received, forecasts are updated, equipment downtime is reported on the factory floor, inventory is updated or adjusted, bills of material are changed, etc.  Since the detailed master schedules have been optimized by advanced planning logic, it’s clear that the aggregated supply plan must be optimized as well. 

But this is all built on several false premises –

·         That humans can be completely eliminated from the planning, production and support processes:  from the evaluation and analysis part of demand and supply planning; from the evaluation of the consequences of various plans; from the process improvement activities associated with existing plant and equipment; from execution activities related to material sourcing; from the value-adding process that transform material into usable products; from any judgment calls when there are trade-offs that aren’t easily quantified, etc., etc.

·         That people are willing to be held accountable for executing a plan where they can't reconstruct the calculations (or sometimes, even understand it changed from the last plan that they agreed to). 

·         That all the data, parameters, constraints, rules and factors are known and can be expressed in a way that the computer can use them to produce an optimal plan; that all this data, parameters, etc. can be updated fast enough when change occurs in the marketplace, in the factory, in the supply chain; and that there is no need for human evaluation or judgment with respect to trade-offs.   And additionally that changes to the same factors needed for planning near term demand and supply can also be predicted in advance) and can be quantified in a way so as to be usable in the optimization calculations.  (This is the same weakness that exists with any software that attempts to predict the future:  the factors and elements in the past can be clearly defined, but will they continue in the future, and how and when will they change?)

·         That the primary emphasis is on planning production levels up to the known and fixed constraint, not on predicting problems in the future and then eliminating them.  This is especially true in “finite loading” type APS systems. 


Assumption 1:  Humans can be completely eliminated  

In fact, humans are in the loop in planning, support and execution in sales, marketing, planning, purchasing and manufacturing.  Even if they could be “engineered out” of the planning and support roles, without a fully automated and autonomous factory (and supporting supply chain), people will still be largely responsible for execution of the plan.  Given that fact, and the implications of frequent updates and recalculations, how could people ever keep up with real time schedule changes?  How could they evaluate whether the changes are reasonable and doable in the time available?  And assuming that they did approve the revised plans, how could they ever redeploy the equipment resources necessary to react to the constant changes in plans?    

If you accept the idea that the best approach is to predict and solve problems in advance, rather than just scheduling around them, then you may recognize that solutions are basically human, and the judgment and evaluation of a master scheduler, planner or shop person is the best way to developing them. For example, many of the ways to solve capacity problems will probably never be programmed into a computer. Reducing setup time until it is inconsequential, rethinking the work and eliminating wasteful steps, reducing unnecessary machining or assembly processes, running overtime or putting on extra shifts, subcontracting, running a job on a different piece of equipment, changing (reducing) order quantities, reducing the lead time for an item, improving flows and reducing the feedback loop associated with finding defective material, putting in place statistical process controls to reduce or eliminate scrap (and therefore reduce capacity required), running jobs early, and many others are common ways to solve capacity problems. For the most part, these are human decision and not ones that can be relegated to an automatic computer calculation. 


Assumption 2:  People are willing to be held accountable for the results of an APS plan

The primary problem with APS systems seems to be the lack of accountability inherent in the system.  A master scheduler, capacity planner, dispatcher or shop foreman looking at a schedule developed in an APS calculation is often unable to verify that the plan makes sense. Because of the volume of calculations and the different interacting options that are taken in the computer program, the people using the system are unable to duplicate the calculation and verify that it is correct. They have to take it on faith: faith that the computer used the right numbers, and faith that the computer took the right options to develop the schedule.

Few people are willing to be held accountable under these circumstances, and those who are willing are just buying time until something goes wrong. Since the calculations cannot be duplicated by the planners using the system, quite a few things could go wrong without becoming obvious. Yet, when the results of such an error manifest themselves in missed shipping schedules or out of control inventories, people are still going to be blamed for the results even if they aren’t “responsible”.  

The situation with APS is no different from a pilot who flies with an “autopilot”.  When an autopilot is used, the instruments are still on.  The pilot always has the option to switch the autopilot off.  This is for a very simple reason: regardless of whether the autopilot is on or off, the pilot is still responsible for flying the airplane, and he will be accountable if something goes wrong.


Assumption 3:  All factors affecting the calculations, both at the present time and in the future, are known and quantified. 

All advanced planning systems assume that the factors needed for planning future demand and supply are known (or can be reasonably foreseen in advance) and can be quantified in a way so as to be usable in the optimization calculations.  This is the same weakness that exists with any software that attempts to predict the future:  the factors and elements in the past can be clearly defined, but will they continue in the future, and how and when will they change?

For these reasons, and because of the inherent complexity of the tools, many of these software packages have been purchased, but only a very few have actually been implemented and used on an ongoing basis. 


Assumption 4:  The primary emphasis is on planning production levels to the known and fixed constraint, as opposed to predicting problems and eliminating them

Capacity requirements planning and resource requirements planning are tools that show predicted capacity requirements.  Neither tool attempts to solve capacity problems, regardless of whether they are overloads or underloads. As explained above, solving capacity problems is a job for people, typically the capacity planner working with people in the factory.  In capacity planning, people are provided with information showing any capacity problems, however, the computer does not automatically attempt to solve these problems.

A finite loading system is a type of APS that that attempts to solve capacity problems is a finite loading system, typically by rescheduling work to earlier or later dates. The term “finite loading" means that the work centers are loaded to their stated capacities, at which point the computer begins to take action to solve projected overload conditions. It schedules the work that creates the overload in a period to an earlier date, or failing that, later in time. 

Do finite loading systems produce results? Yes and No.

·         In some companies, the effect of the “finite” planning calculations is to do what the people in the company are unwilling to do:  unload an overloaded schedule.  So in the sense that the system now is producing a “doable” schedule, the answer is yes. 

·         But more typically, it’s something else that produces the improved results.  This is because, in addition to the finite planning that’s being done, there are four or five other things changing.  As Ollie Wight said years ago – “if you change five things at once and things get better, you’ll probably attribute the improvement to the most complex thing that was changed. In fact, the results are typically due to the simplest thing that changed.” The improvements in companies with finite loading systems often come from better basic material and capacity planning and from manufacturing process improvements rather than finite loading.  So in that sense, finite loading does not produce results.   


What is the Value of APS in S&OP? 

Where APS has been effective (and where it will likely be effective in the future) is as a simulation or what-if planning tool.  Load some factors, and test out the implications of a given set of numbers.  Adjust the factors and try it again.  Let knowledgeable, experienced humans prescreen the significant factors and the potentially effective solutions, rather than have the computer test every possible combination of factors and data.

It seems unlikely that a fully automated APS will ever effectively replace human managed communication and decision-making processes like S&OP and ERP/MRPII.  Rather APS probably is best seen as a set of decision support tools that can be used in circumstances where major change is occurring or seems imminent.  But generally speaking it will be run less frequently, not more, and will likely be limited to a small number of constraints that are being evaluated. 

In fact, this seems to be validated by the “Best Practice” companies studied in “Sales and Operations Planning Best Practices” (Dougherty and Gray, 2006, Trafford Publishing Company).  Of the thirteen S&OP best practice companies, only a single one uses APS extensively – to optimize against a total of two constraints.  In their case, they are doing S&OP against a global supply chain encompassing every civilized continent, dozens of factories, and a multi-tiered supply network.  They state that using the system to optimize any more constraints or criteria in this kind of supply and demand network would result in information that was so complex that it would be impossible for the responsible planners and managers to utilize it effectively.