Five shapes - Anish Patel

# Five shapes *Learn the five shapes a monthly number takes and most of the explaining stops* --- The monthly review is the same meeting in every business I've run. Revenue against last month, cost against budget, this line ahead and that one behind, and a story attached to every movement. Most of the stories are noise, and I don't mean the data is bad or the people explaining it are careless. I mean the movement itself usually carries no information about the machine underneath, and we explain it anyway, because we're meaning-making machines and the meeting was booked. Three rows of the pack, stylised but typical: | May-26 | Actual | vs prior month | vs prior year | |---|---:|---:|---:| | Revenue (£000) | 4,130 | −4.0% | −5.7% | | Gross margin (%) | 61.6 | +0.1pts | −1.2pts | | EBITDA (£000) | 847 | −2.6% | −29.1% | Every cell is a fact and the team that produced it did its job. But look at what it invites. Gross margin is green against April and red against last May: the same number, two verdicts, and a third if you add budget. Each colour asks for a reason, and most months there isn't one to find, because all things being equal you'll sit above your own average half the time and below it half the time. What the table cannot show you is the shape. A number that has been drifting down for six months and a number that jumps around and means nothing look identical in a variance column, and they need opposite responses. Plot fourteen months of each row against limits drawn from its own history and the shape gives itself away in a couple of seconds. Five are worth learning to spot on sight, one per row of the same pack, and the instrument runs all five off the same chart. <iframe src="https://anishshailpatel.github.io/instruments/five-shapes.html" title="Five shapes a monthly number takes" width="100%" height="1140" style="width:100%;border:0;display:block;margin:1.2em 0;" loading="lazy" sandbox="allow-scripts allow-same-origin"></iframe> **Steady** is overheads behaving. Nothing outside the limits, no run, and every colour the pack put on it was noise. Worth knowing before someone spends ten minutes of the meeting on last month's dip. **Drifting** is gross margin, and it is the one the table nearly waved through. Green against April, and below its own normal for six months and counting. No single month was ever big enough to trigger a question, which is exactly how a slow decline survives a year of packs. **Lumpy** is revenue, and it is the trap in the other direction. It moves by nearly three hundred thousand from one month to the next and means very little on any of them. Its limits come out wide because the baseline is bumpy, and that width is what stops you chasing every red cell. **Stepped** is cost of sales. It changed level once, in month nine, and has held there since. Limits set before the change flag every month after it and go on saying the same thing, which reads as six months of bad news and is one event. Find what changed, then re-baseline on where it now sits. **One-off** is EBITDA, down 29.1% on last May, which is the largest red figure in the pack. Last May was an event, and because it sits inside the baseline it has also dragged the centre up until every ordinary month since looks low. Both readings come from one month twelve months ago. A properly seasonal number is a separate problem and needs its own baseline, drawn from the same month in prior years rather than from the last fourteen. These five are what is left once you have dealt with that. ## Where the limits come from The width of the band is not a matter of taste. It comes from how much the number already moves, month to month, over the period you set the baseline on. A steady series earns tight limits and a lumpy one earns wide ones, which is the only sensible answer to how big a move has to be before you care. That is what the bounce slider is for. It scales the noise around each shape while leaving the underlying level or trend exactly where it is. Turn it up on gross margin and watch a real decline vanish inside limits that have grown to swallow it. The decline has not changed. Your ability to see it has. I personally like these charts, and I would concede that some of the formal machinery feels too mathematical for a monthly pack. The intuition is the part worth having: look at the trend rather than the point, and agree in advance what would be worth acting on, which is the argument [[Reading the numbers|the longer essay makes]] about estimates too. A threshold you reach for afterwards just backs up whatever you already wanted to do, in either direction. > [!note]- The working > The formal version is [statistical process control](https://en.wikipedia.org/wiki/Statistical_process_control), and the chart is a process behaviour chart: limits calculated from the mean and the moving range, with rules for runs and shifts. Worth knowing it exists, and probably more than most packs need. Getting this wrong costs more than the meeting. Act on a wiggle and you set machinery going: initiatives briefed, owners named, teams re-pointed. In a small business that is recoverable. Once you are mid-sized, the briefing takes a month or two to reach the people who have to do it, and by then you have held another review and started another set of things, and the whole place works flat out without moving anything forward. ---