The Variance Spectrum

To illustrate the idea I plotted a spectrum. The left side represents zero variance—the realm of manufacturing and Six Sigma, where the whole point is that every output is identical to the last. The right side is 100 percent variance, where R&D and innovation reign supreme, and each output differs fundamentally from what came before.

The poles help explain it, but it's what you place in the middle that makes it powerful. Finance sits far left since so much of its output is “governed”—quite literally the government sets GAAP accounting standards and mandates specific tax forms. Sales lands somewhere in the middle: a pretty good mix of process and methodology plus the “art of the deal.” Marketing sits off to the right, just behind R&D.

That placement explains the system-of-record problem. The variance of marketing's output—the fact that each campaign and piece of content is meant to be different than the one that came before it—makes for an environment that at first seems opposed to the basics of systemization that the rest of a company has come to accept.

But that's just the first layer. Companies are hierarchical, and at any point on the spectrum you can drill in and find a whole new spectrum of activities ranging from low variance to high variance.

Finance may be “low variance” on average thanks to government standards, but forecasting and modeling is most certainly high variance: something that must be imagined in original ways depending on the company, its products, and its markets. Zoom into marketing and you find the same thing—brand governance sits far to the low-variance side while creative development clearly occupies the other pole.

Another way to articulate these differences is that the low variance side represents the routine processes and the right the creative. The left side is where you have the most clarity about the final goal—in manufacturing you know exactly what you want the output to look like when it's done. The right side holds the most ambiguity—the goal of R&D is to make something new.

While I haven't seen anyone else plot things quite this way, the underlying idea—that there are fundamentally different kinds of tasks within a company—is not new. Organizational theorists Richard Cyert, Herbert Simon, and Donald Trow noted this duality in a 1956 paper called “Observation of a Business Decision.”

Simon, of course, went on to develop the concept of satisficing—the idea that humans don't optimize, they find something good enough and move on. The connection to the variance spectrum is direct: on the left side, where problems are well-defined, you can optimize. On the right side, you can't even define what optimal looks like. You satisfice.

For that reason, high variance tasks should also fail far more often than their low variance counterparts. Nine out of ten new product ideas might be a good batting average, but if you are throwing away 90 percent of your manufactured output you've massively failed.

Even though it may be tempting, that's not a reason to focus purely on the well-structured, low-variance problems. Richard Cyert laid out the danger in a 1994 paper called “Positioning the Organization”:

This is the gravitational pull of the left side of the spectrum. Structured activity drives out unstructured. It's easier to answer email, check boxes, attend meetings—the routine stuff—than to sit with the ambiguity that high-variance work demands. Most organizations optimize for the left because it's measurable and comfortable. But the value disproportionately lives on the right.

This is also, incidentally, why bureaucracy metastasizes. Every low-variance process that gets added—every approval chain, every governance layer—crowds out the space for high-variance creative work. The sabotage manual didn't need to invent a way to destroy organizations. It just described what happens when you let the structured side win.

Going a level deeper, another way to cut the spectrum is based on how you should actually solve the problem. For routine tasks on the left, you want a single way of doing things—push down the variance of the output. On the high-variance side you need the freedom to try different approaches. In software terms: automation and collaboration respectively.

This explains so much about the SaaS industrial complex. Most SaaS tools are designed for the left side of the spectrum—they encode a single way of doing things and ask you to conform. That works brilliantly for payroll processing. It's a disaster for content strategy. Their mirror becomes your mold.

AI is different. AI is malleable—it can operate across the entire spectrum. It can automate the routine stuff on the left and serve as a creative collaborator on the right. That's why it's the first technology with a genuine shot at being marketing's system of record. Not because it imposes structure, but because it can work with whatever level of variance the task demands.

This is primarily a framework for thinking about process, but there's a more personal use I keep coming back to. Employees over- or misinterpret the feedback of more senior people all the time. I experienced this many times as CEO—an aside about color choice in a design comp gets misconstrued as an order to change when it wasn't meant that way.

The variance spectrum can make feedback explicit: is this a low-variance directive you expect to be acted on, or a high-variance comment that is simply your two cents? I found it helped avoid ambiguity and made it clear I respected their expertise. “This is high-variance feedback” became a useful shorthand for “I'm thinking out loud, not giving an order.”