How good is your production scheduling?

It’s one of those questions that seem simple enough – but aren’t.

When I ask supply chain managers how good their production scheduling is, I’m often met with incomprehension. “Well we have good planners who know the business well. What exactly are you talking about?”

Those managers are right to throw my question back at me. ‘Good’ is a vague term. In fact, there are at least three ways of looking at what ‘good’ might mean in the context of production scheduling.

When supply chain managers say they have ‘good’ planners, they usually mean they have planners who create feasible schedules. This is no mean feat. It isn’t easy to take all the relevant constraints into account to create a schedule that can actually be executed. But isn’t there more to a good schedule than that? What if those feasible schedules could be improved?

This brings us to another common definition of a ‘good’ schedule: A schedule is good if an experienced planner can’t improve it. After all, if an experienced planner can quickly and easily spot ways to make a schedule better, it can’t be very good, can it? But what if many of the opportunities to improve a schedule aren’t easy to spot – even by an experienced planner? Even worse, what if those invisible opportunities mean that millions of dollars are being left on the table?

This scenario is more common than many realize. There are an almost infinite number of ways to create the average production schedule. Mathematicians call such problems NP-Hard, which simply means that there are so many possible options that it’s impossible to compute the optimal solution. Even experienced planners regularly miss opportunities that could lead to improvements of up to 20%. This 20% ceiling exists for a reason – which I’ll come to later. Before going any further though, I’d like to consider the steps that need to be taken to answer the question I posed at the beginning of this post: How good is your production schedule?

First, you need to create metrics that can be used to determine how good or bad a production schedule is. It’s easy enough to say that a schedule should maximize profit or minimize total cost. But breaking down high-level financial goals into measurable benchmarks raises more questions.

• What maximizes profit?
• Should you aim for the lowest possible inventory or should you try to ensure that SLAs are always met?
• Should your planners attempt to minimize energy consumption, avoid scheduling overtime, or aim to achieve something else entirely?
• What about the fact that a schedule that minimizes costs may not be very robust?
• How do you put a monetary value on the risk that even small deviations from the schedule will result in chaos?

These fundamental questions must be discussed and answered before you can even begin to evaluate production schedules.

Once the right operational KPIs are in place, you need a way to benchmark schedules. Here again, there are no easy answers. As there’s no way to determine the optimal solution to an NP-Hard problem, there are no absolute benchmarks to aim for. What you can do is to apply certain tests to uncover hidden opportunities for improvement – what I call your hidden optimization potential.

At Quintiq, we do this by applying supply chain planning and optimization (SCP&O) to historical data, and comparing the results with historical schedules. While after-the-fact scheduling naturally yields better results, we can make allowances for this to arrive at a ballpark figure. What we generally find is that our approach to SCP&O produces improvements of between 5% and 20% in planning challenges ranging from manufacturing and logistics, to workforce planning. The reason for the 20% ceiling is obvious once you think about it. A plan that’s more than 20% worse than what our optimizers achieve tends to be noticeably bad and gets revised by experienced planners. Anything less than that, and human planners are often incapable of spotting the difference between plans that save millions and those that leave millions on the table.

This isn’t because they aren’t good planners. It’s because even the best planners are no match for the complexity of the average production schedule.