Reading a sizing deck

Analyst session 1 builds Kettle & Co's top-down as a chain of four links: Japanese two-or-more-person households (34,553,907 in the 2020 Census), times the 5,737 yen a year such a household spends on leaf tea (Family Income and Expenditure Survey, 2023-2025 average), times the share of that spend that is premium (0.33, constructed), times the share of premium bought online (0.30, constructed). Result: 19.63 billion yen a year.

"Every tea drinker is our customer" deletes the last two links. Not premium tea, not online: all leaf tea, every household. The remaining two links are both real statistics, so the result looks more solid than the honest version, not less. That is the trap: a hero assumption makes a slide feel better sourced while making it less true.

Top-down with the hero: 198.24 billion yen a year, against 19.63 without it: 10.1 times larger, which is just the two missing shares multiplied back out.

Against the bottom-up: 15.18x. Kettle's analysts also sized the market from the other end: buyers, spend per order and orders a year, which gives 13.06 billion yen a year. Without the hero the two methods sit 1.50x apart and their ranges overlap. With it, the gap is 15.18x and the ranges do not touch.

Best single input closes 7%. The bench tries moving each remaining input to the edge of its range to see how much of the gap it can explain. The best manages 7 percent. When no reasonable input can close a gap, the gap is not uncertainty. It is a missing piece of logic, and the conversation should be about the sentence, not the decimals.

All of these are printed by the bench in Analyst session 6 at its default settings. Your numbers will differ as soon as you change an assumption there.

The headline is a TAM. The biggest number sits in the biggest type, and nothing on the slide narrows it to the buyers the product reaches through the channel it sells in. Business session 1 called TAM the least useful of the three sizes; a hero assumption is how TAM gets promoted to the plan.

There is no second method. One chain, one number, no bottom-up beside it. A second method is the cheapest hero detector there is: if the bottom-up from your own buyers is ten times smaller, somebody has to explain the difference, and "every tea drinker" does not survive being said aloud.

The phrase itself. "Anyone who", "every", "all", "the whole category". Listen for words that make a population the same size as a country.

In Business session 1 the same households and the same survey gave 271.00 billion yen a year for "wellness drinks" and 19.63 for "premium loose-leaf". Both are defensible descriptions of what Kettle sells. Only one of them matches what Kettle's pilot customers actually buy instead of Kettle: other leaf tea, not mineral water or sports drinks.

The tell on a slide: the category is broader than the product, and nobody on the slide can name the competitive alternative it implies. Ask "if we did not exist, what would this buyer use instead?". If the honest answer is a specialist tea shop, the frame is premium loose-leaf, whatever the slide's headline says.

The case for it. As a sanity anchor, a small share of a well-defined market can be useful: if a business would be viable on a share far smaller than its rivals already hold, market size is probably not the constraint, and the conversation can move on to execution. Some founders and investors use it exactly that way, as a quick check that the opportunity is not too small to matter.

The case against it. Its critics say it hides the work that would win the 1 percent: which buyers, through which channel, at what cost to acquire, against which alternatives. It skips the bottom-up entirely, so the number never meets the company's own evidence. And a percentage applied to a vanity frame inherits the frame's problem: 1 percent of a category your buyers are not choosing from is not a plan.

Where that leaves you. The disagreement is real and the literature has not settled it. A reasonable practice: accept a small-share figure as context only when it sits beside a bottom-up that names the buyers, and treat it as a warning sign when it is the only reasoning on the slide.

The 1 percent argument has a group version. If five entrants each pitch a small share of the same big market, each sounds modest alone and the plans together can still claim more than the market holds. Analyst session 5 runs this as the collision test on Kettle and three constructed rivals, and Business session 4 turns it into a check every plan should pass before anyone believes its share figure.

The anchor camp. Some teams use analyst-firm figures as an outside view: a number built by people with no stake in your plan, and often with access to trade data you do not have. If your own sizing lands within shouting distance of it, that is mild comfort; if it lands far away, that is a prompt to look for the reason.

The unusable camp. Others treat any figure without a published method as unusable for a decision. Their reasons: the method usually sits behind the paywall or is not published at all; the definition of the category varies between firms, so two reports on "the same market" often size different things; and the firms sell the reports, often to the companies in the market, which gives everyone a reason to prefer a large and growing number.

Where that leaves you. Both camps agree on one thing: a figure without a method cannot be checked. Whether it is still worth having on the slide as context is a judgement your team has to make and write down. No firm is named here and no real analyst figure is quoted, because the useful lesson is the shape of the problem, not any one report.

Triangulation means two methods built from different evidence, compared, with the gap explained by the inputs. Analyst session 6 has a break button that replaces Kettle's top-down with a constructed analyst figure of 50.00 billion yen a year. Against the bottom-up of 13.06, the gap becomes 3.83x, and the bench prints "cannot close".

Not because 3.83x is large. The hero gap was larger. It cannot close because the bench has nothing to move: no population, no rate, no share. A figure with no method has no inputs, so it cannot be one side of a triangulation. It can only sit beside one.

The same three patterns turn up when a market size is part of somebody's pitch to you. How sizing fits inside a consulting proposal, and how to read one as the client, is taught in AI Consulting ↗. The five questions below work there unchanged.