Sample frameDay 310·Operations · Variation

Morning. Most of what your dashboard flags as a problem this week is just the process breathing.

Operations.

Common Cause, Special Cause

5 min read·Apply by lunch

The question

The last time a weekly number moved the wrong way, did anyone check whether that figure had ever sat outside its ordinary range before you changed something — or did you adjust, watch the next week land back where it always lands, and quietly credit the change you'd made?

The idea

Walter Shewhart was a physicist in Western Electric's inspection engineering department when, on 16 May 1924, he sent his superior a one-page memo with a sketch on it. The sketch was the first control chart. The group moved to Bell Telephone Laboratories the following year, and Shewhart set the argument out properly in Economic Control of Quality of Manufactured Product (1931). It turns on a single division: every process produces two kinds of variation. Chance causes are the ordinary jitter of a process working exactly as built — there's nothing in any one of them worth investigating, and they'll all be back tomorrow. Assignable causes are real events: a new supplier, a worn tool, one person doing the job differently this week. You'll know the pair by their later names, common cause and special cause. What made Shewhart unusual is that he didn't treat any of this as a statistical question. He treated it as an economic one — the chart exists to stop you spending money on the wrong kind of variation. React to an ordinary point as though it were an event and you tamper: you adjust a stable process in response to noise, and the adjustment contributes variation of its own. The trap is that tampering looks exactly like management. Something moved, you responded, it came back.

The framework

01
Plot it over time before you form an opinion about it.
One number held up against last week's number is not information; it's two points and a mood. Take the last fifteen to twenty-five values of the measure in the order they actually occurred, draw the average as a line through them, and mark the range the process ordinarily produces. Almost everything you have been calling a good month and a bad month will sit inside that range. The chart isn't decoration — it's the only instrument that tells you whether a point deserves a meeting, and it takes about ten minutes to build in a spreadsheet.
02
Two mistakes, and you have to decide which one you'd rather make.
Shewhart's framing was economic rather than statistical. You can react to noise as though it were a signal, which wastes effort and adds variation; or you can miss a genuine signal and let a real problem run. No way of setting limits eliminates both, so he settled on three standard deviations from the average as the practical compromise — wide enough that ordinary points are left alone, tight enough that real events still break through. Every organisation leans one way, and most lean hard towards the first mistake, because reacting is visible and restraint looks like negligence. You don't get credit for the meeting you didn't call.
03
What actually counts as a signal here?
Write the rules down in advance, or you'll find a signal in whichever direction you were already worried about. Western Electric's Statistical Quality Control Handbook codified the short list in 1956: a single point outside the limits, and eight consecutive points on the same side of the average. Add a steady run of increases or decreases and you have the working set. Any of those earns an investigation. Nothing else does — including the point your board asked about, and including the one that moved in your favour, which is exactly the same class of evidence and gets a fraction of the scrutiny.
04
A stable process that isn't good enough is a design problem.
This is the half people skip. If the chart shows nothing but common cause and the results still aren't acceptable, the process is doing precisely what it was built to do, and no amount of attention paid to individual points will move it. Predictable and acceptable are different words, and a chart can only ever speak to the first. Improving a stable process means changing the process itself — the handover, the tooling, the form, the rule — and then plotting the next twenty points to see whether the range genuinely moved. If it didn't, what you changed was decorative.

Members only · 4 principles + template + AI mentor

364 more frameworks are waiting.

Common Cause, Special Cause is Day 1 of 365. One framework every morning for a year — across strategy, sales, negotiation, leadership, and more.

$1 /day

Billed $365/yr · cancel any time

365 frameworks

12 topics · templates · AI mentor

Reserve your place — $365/yr →

One payment a year · the whole library · no auto-renewal.

More Operations frameworks

All 24 operations frameworks →