Statistical Process Control & Strategy

Being a Theory-Driven Strategist

Source: Shutterstock, 2024

I have gotten into an enjoyable conversation with readers on the topic of statistical process control and strategy, and they have encouraged me to write a Playing to Win/Practitioner Insights (PTW/PI) piece on it. So, I have done so in Statistical Process Control & Strategy: Being a Theory-Driven Strategist. All previous PTW/PI can be found here.

What is Statistical Process Control?

Statistical process control (SPC) is defined by the American Society for Quality (ASQ) as the use of statistical techniques to control a process or production method. The founding father of SPC is considered to be American engineer Walter Shewhart, who developed the core methodologies in the 1920s.

[I do need to point out that the ASQ would undoubtedly prefer me to use the term statistical quality control (SQC) not SPC in this article. They define SPC as dealing solely with process inputs — which like me, they see as the independent variables under the company’s control. They define SQC as dealing with process outputs (dependent variables) — which I am most interested in. I choose to use the term SPC because it is more widely known, but I recognize that ASQ would think of me as talking about SQC.]

Importantly for this discussion, SPC holds that there are two kinds of variation in the outcomes of a process (in their words):

1. Common cause variation, which is intrinsic to the process and will always be present

2. Special cause variation, which stems from external sources and indicates that the process is out of statistical control

That is to say, no matter how carefully and thoroughly you have designed a process, it will produce variation. Some of that is random statistical fluctuation around the designed mean output of the former sort. But other variation comes from things not anticipated in the design of the process and are of the latter sort.

In SPC, you set upper and lower control limits within which we can safely assume that the process is experiencing only common cause variation. If a process is operating within control limits, it should be left alone — keep your hands off the dials! However, when the outcomes range outside the control limits, something unexpected is occurring — special cause variation — and the process needs to be improved to achieve desired outcomes on a consistent basis.

Application to Strategy

Recall that strategy is an integrated set of choices that compels desired customer action. The company makes choices on the various relevant independent variables under its control to produce outcomes on the key dependent variable — desired customer action (DCA). There will always be common cause variation in response to any set of strategic choices. Strategy can’t produce something with precise, unvarying outcomes. But it is important to distinguish in strategy between common cause and special cause variation. If it is common cause variation, the strategy is working. If it is special cause variation, it is likely that it isn’t.

If you consider strategy rigorously, you will have a logic structure behind your set of choices, what I always refer to as the what-would-have-to-be-true (WWHTBT). That is, WWHTBT for your set of choices to produce the DCA?

The WWHTBT can be utilized to set control limits for your strategy. That is, for each important industry, customer, company and competitor WWHTBT, within which control limits would you remain confident that we are tracking toward our desired outcomes? For example, if a WWHTBT is that within five years we can build the strongest brand in our industry, what would give us confidence in the next (say) two years that we are tracking toward that outcome? Let’s say we would have to achieve parity with the leading brand within two years to be on track to be the demonstrably best brand in five years. Or, with the increasing scale anticipated in our strategy, we can get our costs down 50% in the next three years might mean that we must get costs down (say) 12% within the first year. For each WWHTBT, what are our control limits?

Setting control limits in advance is critically important because human beings have an infinite capacity for ex post rationalization. Humans attempt to rationalize horrific genocides — and do so at war crimes trials. They will claim under oath that they had no choice when they truly did. Criminals have very detailed rationalizations of their crimes. There is no limit!

I have watched this ex post rationalism phenomenon over and over in the business domain. For example, a management team can declare that if we build this factory or if we double our salesforce, our sales will increase by 50% in five years. Then, five later, when sales are up 25%, the same team will say to itself — believing every word — that the results were exactly as planned. Often, the actual growth rate doesn’t produce an acceptable rate of return for the investment — but that is never considered because the original ‘control limits’ were never written down and have faded into the mists of time.

The only protection against your own ex-post rationalization is to set control limits ex ante. That will give you the ability to stick to your strategy — long enough, but not too long. If the elements of WWHTBT are within control limits, keep running the strategy. As soon as they are out of control limits, get to work on changing it now — right now. If you don’t set control limits, you will convince yourself that the strategy “just needs a bit more time to work,” or you will start tweaking your strategy even if it is within what would have been reasonable control limits.

Why WWHTBT & Not Just DCA?

An argument could reasonably be made that you shouldn’t have to set control limits for all the elements of WWHTBT because one element — what would have to be true about customers — should take care of the dependent variable, DCA, on its own. If DCA is within control limits on its own, shouldn’t that provide enough confidence that your strategy is on track, and you can chill?

That is not an unreasonable argument. However, I often find that the customer element of WWHTBT is a lagging indicator. Other elements get out of control limits — and only later do you feel it with customer actions because customers are often sticky. Industry dynamics might be shifting for the worse, the distribution channel may not providing the necessary support, our costs may be getting outside control limits, and/or competitors may have started initiatives that run against the WWHTBT assumption.

These can all happen in advance of seeing DCA get out of control limits. I have watched it happen. And that is why I advocate setting control limits ex ante on all elements of WWHTBT. It can serve as the canary in the coal mine for likely future adverse customer action.

Isn’t SPC Purely and Narrowly Quantitative?

When I talk about using SPC to monitor, manage and adjust strategy, I often get challenged about the purely quantitative nature of SPC. They think of classical quantitative control charts, with numbers to the last decimal point and point out that much of strategy is qualitative and not reducible to precise numbers.

I agree with the characterization of strategy as having important qualitative dimensions. I also understand that SPC was created as a very precise mathematical/quantitative technique and is taught to this day with that mindset.

But I use it more conceptually. For me, in the work of strategy, SPC is a tool not for quantitative precision but for logical precision. Are competitors behaving within a band we anticipated in our WWHTBT? For example, did they immediately launch retaliatory moves against out core business when we launched our strategy? Or did the distribution channel give our new line the kind of support we were counting on. It is not as though the latter one can’t be quantified at all. But it is not subjectable to precise quantification, and that is not necessary to make an assessment as to whether the WWHTBT for the distribution channel appears to be remaining within control limits or not. What is imperative is having a very clear strategy logic.

Practitioner Insights

You should aspire to be a theory-driven strategist. That is, your strategy should be based on a clear theory of the future — about what we can do, how our industry will evolve, how competitors will behave, and, most importantly, what actions customers will take. Don’t just think about these elements of your theory casually in passing. It is critically important that you make your logic very explicit to yourself.

A critical attribute of making it explicit is to write it down to protect yourself from ex post rationalization. If you don’t, you will ex post rationalize and won’t be a theory-driven strategist.

Being explicit about and writing down your strategy will enable you to determine — and again to write down — how far from your predicted outcomes the real outcomes need to stray for you to tweak the dials. That is, you need to set the control limits on your theory — ex ante.

If outcomes stay within your preset control limits, then keep your hands off the dials — with confidence. But if they stray outside, start rethinking your theory — again with confidence. In this way, you will be a theory-driven strategist.