Where we start
Business session 1 turned a vague ask into a sizing question with a decision and a threshold. Business session 2 read a sizing deck for the hero assumption. Tonight is about people: who writes which part of a sizing, who signs it, and what goes wrong when the wrong person holds the pen. One seam to say out loud first. The general map of what a data team owns versus what marketing and BD own, and the intake process that routes a request, are taught in Marketing Analytics: the ownership map ↗ and intake and the service menu ↗. Market sizing appears there as a "both" row with a split contract. Tonight zooms into that one row and writes the contract down.
★ What you walk out with tonight. A filled one-page sizing brief with seven fields, the business half written by the business and the data half written by data, read aloud and signed by the other side. Plus four warning signs that the halves are blurring, and the fix for each. Business session 4 turns that brief into a plan.
Data owns the evidence, the business owns the bet
5 min liveThe evidence half answers "what is true, and how sure are we": defensible numbers, each with a source, ranges rather than points, the method, and the gap between two methods with the input that explains it. The bet half answers "what will we do about it": which category to compete in, what share to aim for and by when, and the go or no-go. The split is not about status. It is about keeping each half honest, because each one is easy to bend by the person who owns the other.
LiveWhy the split matters: two ways a size quietly changes3 min▶
When the analyst picks the category, the number grows. Nobody intends it. An analyst asked "how big is Japan for us?" without a category will reach for the frame with the most data behind it, and the frames with the most data are the broad ones. Business session 1 showed the same households and survey giving 19.63 billion yen a year for premium loose-leaf and 271.00 for wellness drinks. The analyst who picks a frame has picked a number, and has made a positioning choice that belongs to marketing.
When the bet owner edits the range, the evidence stops being evidence. A BD lead who trims 10.90 to 21.55 down to "about 20" for a board slide has not simplified the evidence; they have replaced it with their own belief and kept the data team's name on it. The board now thinks it is reading measurement when it is reading hope. The fix is not to forbid simplification. It is that only the person who signs a half can change it.
The pattern underneath both: each half is easy to bend by the person who owns the other half, because they are the one who wants a particular answer. Separate pens are a cheap guard.
Self-studyWhat "both" means in the ownership map, and what it does not2 min▶
The Marketing Analytics ownership map ↗ lists market sizing as a "both" row. "Both" does not mean "both teams edit the same number until it looks right". It means a split contract: two named halves, two owners, one page. Who is responsible, accountable, consulted and informed on each row, and how a request enters the team, are taught there and not repeated here. Tonight only draws the line inside the one row.
The slide that said "about 20". Kettle & Co's head of data sends the Japan sizing as a range, 10.90 to 21.55 billion yen a year. The BD lead, building the board pack the night before, rounds it to "about 20 billion yen" because a range "looks unsure". The board approves a Japan team sized for the top of the range. Six months later the online pilot is tracking nearer the bottom, and the head of data is asked why the number was wrong. It was not wrong; it was edited. Constructed Kettle & Co example; the shape is common.
Seven fields on one page
5 min liveThe brief is the split contract made physical. Seven fields: the question, the decision, the owner, the category, the method, the range, and the assumption that would change the call. The first four are the business's and are written before any number exists. The last three are data's and come back with the evidence. One page is a hard limit: if it does not fit, somebody has written an essay where a decision maker needed a line.
LiveThe seven fields, filled for Kettle & Co3 min▶
1 · Question. From Business session 1: how big is premium loose-leaf tea in Japan, bought online, next year, as a range. 2 · Decision. Enter Japan online next year, yes or no, against a threshold of about 10 billion yen a year of SAM. The threshold is constructed for Kettle; yours comes from your own owner. 3 · Owner. Two names: the BD lead answers for the bet, the head of data answers for the evidence. 4 · Category. Premium loose-leaf, chosen by marketing, with the competitive alternative it implies: specialist tea shops and other direct-to-consumer leaf brands.
5 · Method. Top-down (census households, survey spend per household, premium and online shares) and bottom-up (buyers, spend per order, orders a year), built separately so that one can check the other. 6 · Range. The two methods overlap from 10.90 to 21.55 billion yen a year, and their middles sit 1.50x apart. 7 · The assumption that would change the call. Share bought online. At the edge of its range it closes 76 percent of the gap between the two methods, more than any other input, so it is the number to measure next.
Every number in rows 6 and 7 is computed by the bench in Analyst session 6 from real survey and census rows and constructed Kettle shares. Change an assumption there and your numbers will differ.
Self-studyWhy row 7 matters more than row 62 min▶
Kettle's whole range, 10.90 to 21.55, sits above the constructed 10 billion yen threshold, so row 6 already says "go" on paper. But the bottom of the overlap is less than a billion yen above the line. Row 7 tells the decision maker what would push the range below it: if the share of premium leaf bought online is at the low end of its assumed range, the picture changes. That turns a go into "go, and measure online share in the first quarter", which is a far more useful sentence than any point estimate.
A range without a call-changer leaves the decision maker with a number to admire. A call-changer gives them something to do.
The brief that arrived as a 30-slide deck. Kettle & Co's marketing lead asks for "the brief" and receives thirty slides: methodology, survey tables, sensitivity charts. Nothing in it is wrong. The leadership meeting spends forty minutes on slide 4 and never reaches the input that would change the call, which was on slide 27. The second attempt is one page with seven rows; the meeting takes ten minutes and ends with a named owner measuring online share. Constructed Kettle & Co example.
Four ways the wrong person ends up holding the pen
5 min liveThe split rarely breaks in one dramatic moment. It erodes in small, reasonable-sounding steps, usually under deadline. Four blurs account for most of it. Three move the evidence into the bet; one moves the bet into the evidence. Each has a cheap fix, and each fix is a field on the brief written at the right time by the right person.
LiveThe four blurs and their fixes3 min▶
1 · The threshold is set after the number is seen. The size comes back at 13, and the threshold quietly becomes 12. A threshold drawn under a known number is not a test; it is a caption. Fix: the threshold goes in row 2, dated, before any analyst starts work.
2 · The category is chosen by whichever is biggest. Three frames are sized "to see", and the biggest one ends up on the slide. That is the analyst making marketing's positioning call by default, and Business session 1 showed the cost: a 20.75x gap with the pilot under the wellness frame. Fix: marketing ticks the category in row 4, with the competitive alternative, before sizing starts.
3 · The range is edited to a point for the board. "Boards do not like ranges" is a real pressure and a bad reason. Fix: the range stays, and row 7 goes next to it, so the board sees both what is known and what would change the answer. If a single number must appear, it is labelled as the bet owner's planning figure, signed by them, not by data.
4 · The analyst presents the go or no-go. Fix: the bet owner says go or no-go and why; the analyst says what the range does and does not support. Quiz question 3 asks why this one is the subtlest.
Self-studyA real disagreement: should the analyst sit inside the bet?3 min▶
Practitioners disagree on this, and both sides have a point. Some teams embed the analyst in the bet. The analyst sits with marketing or BD, hears the strategy as it forms, and turns a question around in days rather than weeks. Their sizing is closer to the decision because they were in the room when it was framed. The risk is that they absorb the team's hopes and their ranges drift towards the answer the team wants.
Others keep the analyst separate. A central data team sizes for whichever team asks, reports to someone who does not own the bet, and has no stake in the answer. Their evidence is more independent and comparable across teams. The cost is speed and context: they can size the wrong question well because nobody told them what the decision really was.
No study settles which works better, and the answer probably depends on how often a team bets and how much it costs to be wrong. What both camps share is the brief: whether the analyst sits in the room or across the building, the seven fields and two signatures keep the halves apart on paper. How the reporting lines themselves are drawn belongs to the ownership map ↗.
When every analyst is inside a bet. Damodaran's big market delusion posts (Musings on Markets, 2015 and 2019 posts) describe the market-level version of a bet writing its own evidence. In 2015 he took each listed online advertising company's market value, worked out the revenue it implied, and added them up: the total came to more than his own optimistic estimate of the whole market. Each company's case, built by people inside its own bet, can look reasonable alone; together they exceed the market. He calls the habit behind it pricing rather than valuing: "young companies tend to be priced, not valued". Analyst session 5 runs that collision test on Kettle and its rivals.
Fill the one-page brief, one half at a time
★ 22 min · everyone works it throughNo code, just a page that passes between two desks. In pairs, write the business half first: decision, owner, threshold, category. Then hand it over and fill the data half, from the Analyst bench if your team has run it, or as a placeholder if not. Finish with a two-minute read-aloud where each side signs the half the other side wrote. Kettle & Co's filled example sits under every step; your answers will differ, and that is the point.
Step 1 · Business half, in pairs: decision and owner. One person plays the bet owner, one the analyst. The bet owner writes; the analyst may only ask questions. Start from your Business session 1 sentence if you have it.
Step 2 · Business half: threshold and category. Write the threshold into row 2 and date it. Tick the category and name its competitive alternative. If the bet owner hesitates, that hesitation is the finding: a sizing without these two lines will be redone.
Step 3 · Hand it over: the data half. The analyst now fills rows 5 to 7. If your team has run the Analyst bench, copy its method, overlap and best-explaining input. If not, write placeholders that say what will go there; an honest blank beats an invented number.
Step 4 · Two-minute read-aloud, then sign. The bet owner reads rows 5 to 7 aloud in their own words; the analyst signs only if it was heard correctly. Then the analyst reads rows 1 to 4 aloud; the bet owner signs. Anything that could not be said in plain words goes back for a rewrite.
Step 5 · Check for blurs. Run the four questions from Part 3 against the page you just filled. One "yes" means a pen moved; fix it before the brief leaves the room.
Keep the signed page. Business session 4 starts from row 7 and the bet half: how a range becomes a plan, and what evidence a team should demand from its own numbers before anyone believes the share in the bet half.
Before Business session 4
30 min- Find one real sizing in your organisation and mark each statement in it as evidence or bet. Count how many sentences mix the two.
- Rewrite it as a one-page brief with the seven fields. Leave any field you cannot fill blank, and write who should have filled it.
- Run the blur check from step 5 against it. Note which of the four blurs, if any, happened, and when.
- Read the "both" row in the Marketing Analytics ownership map ↗ and check that your two signatures match who your team says is accountable. Bring any mismatch to Business session 4.
Three questions
Clear all three and Business session 3 gets stamped in your passport.
1 · Who should choose the category for a market sizing, and why?
The category is part of the bet. Left to the analyst, it drifts to the broadest frame, and the number grows without anyone deciding that it should.
2 · Which field of the one-page brief is the most useful to a decision maker, and why?
For Kettle the range already clears the threshold on paper; row 7, share bought online, closing 76 percent of the gap, is what tells the decision maker what could change that.
3 · What goes wrong when the analyst presents the go or no-go?
The analyst says what the range does and does not support. The bet owner says go or no-go, and owns it.
What this session teaches, and where it came from
Every fact on this page, and its verification tier, is recorded in the course's source map.
Business session 3 cheat sheet · pin this
You have a signed brief with the evidence and the bet kept apart. Business session 4 turns it into a plan: realistic penetration and ramp, why most entrants in a crowded category end small, and the evidence gate a team should pass in its own numbers before anyone believes the share it plans to win.