First, write each layer's definition so that someone else could apply it
Before touching a number, write each layer of the challenged figure as a rule a stranger could apply and reach the same answer. The total: who is counted, by what rule, in which places, in which unit and for which period, bounded by the job buyers need done rather than the product category, so that substitutes sit inside it. The serviceable layer: the segments, places, price points and buyer sizes the business can serve now or within the plan. The obtainable layer: the evidence on sales capacity, conversion and competitor strength behind the share. The sizing guide's counter-examples, "the global software market", a total trimmed by a round percentage, a share picked because it sounds modest, show what a layer looks like when it cannot be written this way, and that layer is where the rebuild begins.
Then read the paper against those definitions. Copy out, in the publisher's words, how the source of the headline total drew its category and set it beside yours; follow each filter down to the serviceable figure and ask whether evidence or a round percentage sits behind it; ask what the obtainable share was derived from and whether a range exists at all. Check too that the layers stayed apart on the page: the sizing guide counts a plan built on the total as if it were winnable among the commonest failures in go-to-market planning.
Two further readings sharpen the diagnosis. Ask who produced each figure and what they wanted from it, including whoever built the size, since a figure made to win funding or justify a budget has a motive to come out large; the guide's test is whether the method would have stopped an analyst who wanted a small number. And read the obtainable layer through the five forces model: rivalry, entry, substitutes and the power of buyers and suppliers, which, as the framework's page notes, often leave the obtainable market smaller than the headline. The table takes the causes in the order a reviewer usually finds them.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| The headline total is far larger than anything the business's own data suggests, and its source describes a different category | The publisher's category definition, in its own words, against the business's definition of the job, unit, geography and period | A published category figure used without checking the publisher's definition against the business's own | Restate the source on the business's definition or log it as context only, and rebuild the total from a definition that holds |
| The serviceable figure is the total reduced by several round percentages, each unexplained | Whether any second route reaches a comparable figure without sharing those filters | A single top-down chain of percentages never tested against a bottom-up count | Build an independent bottom-up estimate from counts and purchase evidence, and read the gap |
| The growth assumption in the plan traces back to a vendor's or trade body's forecast | The source log: producer, interest, method, dates and definition for that forecast | A vendor or trade forecast logged without its interest and method, so its assumption became the plan's | Log and judge the source, mark the forecast as a forecast, and carry it as a range with a second independent source |
| A key input, such as replacement cycle or share of buyers with the problem, rests on one conversation | Whose view it was, what basis they gave, and whether published data or customers agree | One expert's view recorded as a figure, never weighed against published data or customers | Record the view as a coded assumption with a confidence and weigh it against other evidence |
| The obtainable share is a tidy percentage with no capacity or competitor evidence behind it | Marketing's reach, the conversations sales can hold, conversion rates and incumbent strength | An obtainable share chosen for how it sounds rather than from sales capacity and competitor strength | Derive the share from capacity, conversion and competitor evidence, and check it against the five forces |
| The paper gives one number and the reviewer's first question is "how sure are you?" | Which assumptions move the answer most, and whether a low and high case exist for each | A point estimate presented where the evidence only supports a range | Present a range with the driving assumption named, on a dated worksheet |
Fix, verify and prevent each cause
A published category figure used without checking its definition
The market sizing guide opens with this failure: a huge figure from an analyst report and a claim that a small slice of it would be enough, with nobody having asked how the report drew the market. A publisher's category reflects its own choices about product, buyer and territory, which seldom match the business's, and the market and competitor research hub asks of every source, among its five questions, whether its definition of the market matches ours.
Fix: write the business's definition first and paste the publisher's description of its category beside it in the source log; where the two reconcile, restate the total on the business's definition line by line, and where they cannot, log the figure as context only, with the reason, and build the total from a count that fits.
Verify: a colleague who has read neither the report nor the plan applies the written definition to the source and reaches the same restated figure, or the same verdict that it cannot be restated.
Prevent: open every sizing paper with the definition rather than the number, and admit no total to the worksheet until its publisher's definition sits in the log beside the business's.
A single top-down chain never tested bottom-up
Top-down sizing narrows a broad figure filter by filter, and the sizing guide is blunt about the cost: each filter is an assumption, the assumptions multiply, and reasonable-looking percentages can be made to give nearly any total, so with one route an error in the source flows straight into the plan. The guide asks for at least two routes built independently, preferably by different people, and read for why they differ rather than averaged; a top-down result far above the bottom-up one usually points to a published category wider than the business's, or to a segment the count left out.
Fix: build the bottom-up route from the unit of purchase: count the buyers who fit the serviceable definition from official registers, trade directories or customer lists, estimate from the business's sales, competitors' disclosures or primary research how many buy, how often and for how much, and multiply, with each line open to inspection; then set the two routes side by side and write down why they differ.
Verify: each bottom-up line cites its own assumption, the two routes share none, and the triangulation section explains the gap instead of splitting it.
Prevent: give the two routes to two people from the outset, and treat a worksheet with one estimate as unfinished.
A vendor or trade forecast logged without its interest and method
The desk research tutorial exists to stop this: a number copied without its definition, or from a source nobody can trace later, is how a vendor's forecast becomes the plan's growth assumption. It logs every source before any figure leaves it, with the producer, the method, the dates, the market definition and the producer's stake in the answer, and weighs each against the hub's five questions. The hub's tiers explain why: official figures have documented methods but may lag, trade bodies want their sector to look healthy, and vendor research exists to generate leads. A forecast is labelled a forecast, never a measurement, and the desk research glossary entry sums up the procedure.
Fix: enter the forecast in the log with every column filled, ask how the figure might look if the producer's stake pointed the other way, find a second independent source and reconcile the two by definition, period and geography, and where the gap will not close, carry both as a range and say so.
Verify: the plan's growth line traces to a worksheet assumption with a source, a weight and a reconciliation note, and nothing in the paper comes from a producer or method the log cannot name.
Prevent: open the source log at the first search of every sizing pass, so that "where did this number come from" is answered by a cell rather than a memory.
One expert's view recorded as a figure
Distributors, installers, consultants and trade journalists see many purchases at once, so the guide to interviewing industry experts uses them to map a market's structure, test bottom-up assumptions and explain why two published figures disagree. But a view is evidence of what one person believes from where they sit, and every vantage point hides something: a distributor can miss the buyers who buy direct. The guide asks for the basis of every claim and for a range rather than a single number, and records each view as a worksheet assumption under the expert's code, with a confidence reflecting their position and whether experts elsewhere agreed; a view nobody else supports becomes something to research next, not a planning input.
Fix: return to the notes, separate what the expert saw from hearsay and opinion, re-enter the input as a coded assumption with a low and a high value and a confidence, and weigh it against published figures, the business's own data and an expert who sees the market from another side.
Verify: the assumptions table shows the code, basis and confidence beside the figure, and the triangulation section records whether other evidence agreed.
Prevent: recruit experts from different positions so that each view has a cross-check, ask "how do you know?" in every conversation, and never let a view reach a number without passing through the worksheet.
An obtainable share chosen for how it sounds
The sizing guide treats the obtainable layer as the planning figure and the one most often made up, and builds it from what the business can actually sell: the prospects marketing reaches, the conversations sales can hold, how often they convert and how fast customers leave incumbents; a share that needs more conversations than the team has is not obtainable. Competitor strength sets the other limit, and the competitive intelligence guide gathers it from public sources: positioning, pricing and packaging changes, search visibility and share of voiceYour share of market visibility., hiring, filings and what competitors say in deals. It keeps that evidence clean too: competition authorities, the UK's Competition and Markets Authority among them, publish guidance on the risks when rivals share future prices, margins or plans, so competitor strength is read from public material and the business's own customers, never from talking to a rival about prices; this is general direction, not legal advice.
Fix: rebuild the share from the funnel the business has, reach, conversations, conversion and switching, then rate the five forces for the defined market with evidence for each rating and cut the share where buyer power, entrants or substitutes require it.
Verify: the obtainable figure connects to the customers it implies, the capacity needed to win them and the KPIA metric tied directly to an objective. the plan will track, and the competitor rating uses the set buyers compare rather than the one leadership worries about.
Prevent: keep the layers apart on every page, derive the share before anyone sees the serviceable figure it will multiply, and refresh the competitor baseline each quarter.
A point estimate where the evidence supports a range
The market sizing worksheet is the record that makes a size auditable, and its answer is written as ranges: low and high ends for the serviceable and obtainable markets, the assumption that widens them most, the date and what should prompt a refresh. The sizing guide calls a single figure where the evidence only supports a range false precision, because readers then take an assumption for a measurement. Both pages build the range the same way: find the handful of assumptions that move the answer most, run each at a low and then a high value, and report both ends with their evidence and the assumption that drives most of the width; if the range is too wide for the decision, the next research spend goes into that assumption rather than into a bigger report.
Fix: move every assumption into the worksheet's single assumptions table with a low and a high value, a source and a confidence, run both estimates at the ends, and replace the single number with the range, its driver and what would narrow it; where a planning figure is still needed, take it from the end the better-evidenced route supports and say so.
Verify: a reviewer can change one assumption and watch both estimates and the answer move, and the summary tab carries the date and the refresh condition.
Prevent: present the range as the answer from the first draft, and when a survey narrows the driving assumption, give it its own frame and checks, because the diagnosis for survey results that are not representative shows how a measured share inherits whoever happened to answer.
Prevention: two independent routes, a range with its driving assumption and a dated, logged worksheet
Every cause above is a step the worksheet would have exposed, so the routine is the worksheet filled in order whenever a size will inform a decision. It begins with the decision, which sets how precise the size must be, and the definition, which sets what counts. Two estimates follow by routes that share no assumptions, by two people where possible, each line of arithmetic on its own tab and each assumption entered once in a single table with its low and high value, source and confidence. Proxies go in as checks, each with a sentence on why it should track the market: category statistics, audience reachUnique people who saw an ad. for the publications a segment reads, and query search volumeThe estimated number of times a query is searched in a period, usually per month. for the problem, the category and the substitutes, which shows how much demand is expressed and where. The SEO traffic estimator turns rankings into likely clicks, showing how far total searches sit above the visits a business could earn, so the shop window is not mistaken for the market. The triangulation section then reads the agreement and the gap, the obtainable section derives the share, and the answer goes down as ranges with the driver, the date and the condition for a refresh.
The source log stays open beside the worksheet for the whole pass, and desk researchResearch from existing sources, the business's own and published ones, done first so that interviews and surveys cover only what nobody has answered. run that way gives the figures their weight: every source logged and judged before use, each figure taken with its definition and date, and disagreements resolved by definition rather than averaged. The professional codes of the Market Research Society, as the site's research guides present them, ask for findings to be reported with their limits; for a market size that means naming each figure's source, date and gaps, and giving a range rather than a point. A size built this way gets challenged on its assumptions, a conversation worth having, rather than on its credibility.
Keep the obtainable layer honest with a standing check: re-read the quarterly competitor baseline and the capacity numbers the plan tracks before any size is reused, and date every estimate and name what should prompt its refresh, because a size made for one decision is easily carried into the next without anyone asking whether its definitions still fit. A business comparing countries fills one worksheet per market, and the guide to market entry and prioritisation combines each size with demand, competition, regulation and cost-to-serve signals to rank them. And a size shows how many could buy, not whether they will want this product; the demand-testing tutorial puts that question to buyers with an honest page and a real commitment before anything is built.
At the readout, lead with the definition, show both routes and their gap, give the range with the method behind each end, name the driving assumption and what would narrow it, and tie the obtainable figure to the customer targets it implies. If a rebuilt size is accepted and the plan still does not move, the fault lies past the research, and the diagnosis for research findings that are not acted on starts where this page stops. The marketing research troubleshooting hub routes the section's other symptoms, from a survey nobody answers to interviews that produced only compliments.
A refurbished laptop reseller rebuilds a figure the board rejected (illustrative)
Pimlothorpe Computing
Pimlothorpe Computing, an invented reseller of refurbished laptops to schools, takes a market size to its board to support hiring 3 more sales staff. The slide starts from an analyst report putting business computing spend in the country at £9.4 billion, applies 12% for education, 30% for schools, 25% for devices and 20% for refurbished, reaches a serviceable market of £17 million and claims an obtainable 5% because it sounds modest. The finance director asks what "business computing" includes, why each percentage is round, and how sure the team is. Nobody can answer, and the board sends the paper back.
The marketing lead rebuilds it on the worksheet. The market is defined by the job, replacing school devices within a fixed budget, in the 2 regions the firm delivers to, in annual spend. The analyst report, copied into the source log in its own words, covers servers, software and services as well as devices, so it is logged as context only. Bottom-up, the official register lists 4,180 schools in the 2 regions; the firm's own orders put devices in use at between 160 and 200 per school, and 2 trade interviews, with a distributor and a former school IT manager, each logged with its basis and a confidence, put the replacement cycle at 4 to 6 years, which becomes the assumption carried as a range. Dividing devices in use by the cycle and pricing each replacement at the firm's average of £310 a device, the serviceable market comes out at about £35 million to £65 million a year. Search demand for refurbished school laptops, read as a relative signal, has grown in both regions over the last 2 school years, which supports the category's direction but is kept out of the size itself.
The obtainable share is derived rather than chosen: with 5 sales staff holding about 400 school conversations a year and the firm's conversion rate of 18%, the team can win around 70 schools a year, and the five forces review rates buyer power high because 3 purchasing consortia control a large share of orders. The paper presents that serviceable range, names the replacement cycle as the driver of the spread, proposes a short survey of 300 school IT leads to narrow it, and sets an obtainable figure of about £1.2 million to £2.2 million a year from the 140 or so schools the team can win within 2 years. The board approves 2 hires rather than 3, with the survey as the condition for the third.
Common mistakes
- Defending the number instead of the definitions. The challenge is usually about what was counted; answer it with a definition someone else could apply.
- Quoting the publisher's total as if it were yours. A category drawn by an analyst rarely matches the job, unit, geography and period the business sells into.
- One route, several round percentages. Filters compound and nothing tested the chain; build a second route from counts.
- A forecast carried as a fact. A vendor or trade figure without its interest, method and date in the log becomes the plan's assumption by default.
- One expert as the market. A single confident view, entered without its basis or a confidence and never weighed against published data or customers.
- A share that sounds modest. An obtainable percentage with no capacity, conversion or competitor evidence behind it is the slide reviewers distrust most.
- Search volume as market size. Queries are not buyers, and everyone who buys without searching is missing.
- A point estimate, undated. False precision that hides the driving assumption, then reused for a later decision.
