A single market-size number is a single point of failure. If it came from one method, one wrong input moves the whole answer and nobody can tell. Triangulation fixes that by sizing the same market three independent ways and reporting not a point, but a central estimate and how much the methods disagreed — which is the most useful thing on the page.
Each method anchors to a different public number, so agreement is real corroboration rather than the same assumption restated three times:
Detail on each lives in the top-down vs bottom-up and value-theory guides.
Take the three warehouse-robotics estimates from the sample:
| Method | TAM |
|---|---|
| Top-down | $9.00B |
| Bottom-up | $7.56B |
| Value-theory | $10.80B |
The median is $9.00B; the mean is $9.12B. Close here — but the median is chosen deliberately, because it resists one method running wild. If value-theory had come in at $30B instead of $10.8B (one bad input), the mean would jump to ~$15.5B while the median stays anchored at $9.00B. A single optimistic assumption should not be able to drag the headline number up, and the median guarantees it can't.
Disagreement is measured as a spread ratio, and mapped to a band the report presents honestly:
| Spread | Band | How it's presented |
|---|---|---|
| ≤ 40% | tight | Median quoted as the headline — defensible |
| ≤ 100% | moderate | A range, not a point; the method driving the high end is disclosed |
| > 100% | wide | No single number quoted — an input needs a better source first |
At 36% the sample lands tight, so the $9.00B median is reported as the headline. Had it come in at, say, 70%, the report would lead with a range and name whichever method drove the top of it — never a false-precision point.
The instant estimate runs all three methods, takes the median, and shows your tight / moderate / wide band live.
Free instant estimate → Commission a cited report