Even one well-built, well-triangulated chain isn't finished until you express it as a range. Every input carries its own uncertainty, and a point figure hides that from whoever receives it. The standard construction is best / base / worst. Hold every input at its base-case value except the one you're least certain of, then recompute at that input's plausible low and high bounds. Worked Example 6 does this for a base case built on a 4% adoption-share assumption. Hold everything else fixed, let adoption range from 2.5% (worst) to 6% (best), and you get a stated range about 2.4x wide from worst case to best case, around a base case in the middle. A client can act on that single sentence. Hand them a spreadsheet instead and they'd have to interpret it themselves.
Which input should you stress-test first?
Here's a fact worth taking in properly, because it clears up a confusion that trips almost every learner the first time. Say your market-size build is a product of independent factors, such as population × penetration × price. Then each factor has the same mathematical leverage on the output. A 1% move in any single factor moves the final answer by 1%, full stop, whichever factor it is. In a pure multiplication, no factor is mathematically "more important" than another. What differs is how wide a range of values is plausible for each one, and that decides which assumption deserves the most scrutiny before you finalize a number. Suppose your population count is well-established public data, with a plausible range of perhaps ±5–8%, while your adoption rate is a real guess, plausible perhaps to ±40–50%. The adoption rate carries no extra weight in the arithmetic. You're just less sure of it, so it's the one to stress-test first, find better data on, or flag clearly when you hand over the range. Practice P9 computes this directly. A household-count input with an 8% plausible swing, a price input with a 15% swing, and a penetration input with a 25% swing each move the final answer by their own percentage, and penetration is simply the widest of the three.
Advertisement
How do you build and report the range?
Here's the practical routine. After your base-case build, ask which single input has the widest plausible range. Recompute the best and worst case by moving that input alone across its full plausible range, holding everything else at base, and report that as your range. The other inputs can move too. But one dominant uncertainty almost always explains most of the total spread, and naming it tells the listener where a follow-up call or a data source would tighten the estimate most. State the range to two significant figures at most and resist the pull toward false precision. "₹10,237 crore" is a worse answer than "₹10,000–10,500 crore" even if the arithmetic behind it was exact, because the inputs never supported that much precision in the first place. And always report the input you're least sure of alongside the range. That one sentence is often more useful to the decision-maker than the range itself.