Your doctor calls: "Your haemoglobin is 11.2." Is that bad? You've no idea, and neither does the number. A lab report answers by surrounding it. There's the reference range for your age and sex, which is a sector norm. There are your last five readings, which make a time series. Related markers on the same page make a common-sized panel. And there's the level below which treatment starts, a hurdle. The number was never the information. The comparisons were.
Now try "Asian Paints' operating margin is 18%." Good against what? You could hold it against its own ten-year band of 15 to 21%, or against Berger's 16%. You could hold it against the 8% a cement company would kill for, or the 30% a software company would apologize for. Or you could ask whether 18% on this capital base clears the cost of that capital. Each of those comparisons returns a different piece of the answer, and no subset substitutes for the rest. There are five moves in all:
| # | Comparison | The question it answers | Blind spot if used alone |
|---|---|---|---|
| 1 | Time series (5–10 yr) | Is the company getting better or worse? Is the number stable or lucky? | The whole industry may be drifting; a stable mediocrity looks fine |
| 2 | Cross-section (3–5 true peers) | Is it winning or losing against the same game? | The whole peer set may be value-destructive; "best" of a bad lot |
| 3 | Common-size (vertical + horizontal) | Where in the statements is the difference coming from? | Structure without judgment: shows the gap, not whether it matters |
| 4 | Cost of capital (hurdle) | Is any value being created at all? | Says nothing about direction, cause, or peers |
| 5 | Sector norms | Is this level normal for the business model? | Norms describe, they don't excuse; a whole sector can sit below the hurdle |
Advertisement
How do the five comparisons work together?
Run all five and a number becomes a verdict. Skip one and you've got a guess wearing a decimal point. The table hides a precondition, though. Both sides of every comparison have to be measuring the same thing. Think of a margin polluted by a one-off gain, an EBITDA inflated by a lease standard, or a "company" that's secretly a consolidation of three unrelated businesses. Comparing those raw is noise transfer, however analytical it looks. So the work splits in two. There are the five comparisons, and there's the normalization-and-segmentation hygiene that has to run before them. After both, everything comes together in a complete worked study and a repeatable protocol.
Install one habit now, because every section uses it: the gap decomposition. Any difference between two numbers, whether it's this year against five years ago or Company A against Company B, is the sum of three things. There's accounting, meaning different policies. There's definition, meaning different formulas or bases. And there's economics, a real difference in the business. Experts close the first two gates before touching the third. Amateurs read every gap as economics, and that's how a lease standard gets mistaken for a moat.