Market Model
Why size a market — and how
Before it is a number in a deck, a market size is a decision: whether to carry on at all — and, if you do, where to point everything you have.
Is there a market at all?
The first question is not how big. It is whether. A product can be good, wanted by everyone who sees it, and still be built for a few hundred enthusiasts. Counting is the cheap way to find that out — before three years of assuming.
If not yet, could there be?
Where nothing comparable is bought, the question changes shape rather than going away: how big could this become, and what would have to be true to get there? A size you cannot build at all is itself a finding.
Where does the effort go?
A number on its own settles nothing. The point of the exercise is the next move — which customers, through which channel, at what rate. That is what a plan is actually made of.
The three layers do three different jobs, and only one of them is about scale.
That is the shape of the whole exercise, not just the last layer. You size to find out what you do not know, and the answer tells you what to go and check — segmentation, pricing work, the first real customers. Each pass starts from firmer ground than the one before it.
A market size is a claim. Every number in one is somebody's judgement about what counts, what is in scope, and what a customer will pay — and the arithmetic that joins them is the easy part. What makes a size defensible is not the precision of the answer but whether you can say where each figure came from.
Which is why this tool holds sixteen ways to do it rather than one. They are not sixteen degrees of accuracy; they are four genuinely different places to start, and the point of having them is to reach the same market from as many of them as your market allows.
15 methods, four ideas. The four families differ in what they start FROM — a total somebody published, a count of customers, what the customer gains, or how fast the thing gets adopted. Every read you add from a new starting point earns its place: where two disagree is somewhere you can go and check, and where they agree, they agree for a reason worth trusting. Reads that start from the same quantity are the exception — they will agree for reasons that have nothing to do with being right, however many of them you build.
1. Top-down5 methods
Start from a total somebody else has already measured, and cut it down to you.
Reach for it when the category has a name an analyst, a trade body or a statistics office would recognise — somebody has already counted something close to what you sell.
The trap is that the figure you start from decides the answer, and it was measured for somebody else's purpose. Every step after it is a cut, and a cut you cannot name a filter for is a guess wearing a percentage sign.
Show the 5 methodsHide the methods
Existing references
a list you build yourselfPublished market sizes, used as they are — or narrowed to your slice with a filter you can name.
Watch out for Each row is only as trustworthy as its source — a stale or vendor-inflated number silently anchors everything downstream. Where you've narrowed a figure, the % is the weakest link and the first thing a reviewer will push on: it must be a scope cut you can name a filter for (an application, a geography, a customer type), never a guess at the share you'll win. The published figure must genuinely contain your market, too — a broader category that only overlaps yours isn't a parent.
Top players & share
a list you build yourselfAdd up the competitors you can name, then gross up by the share they hold between them.
Watch out for This grosses up via combined revenue ÷ combined share% — the fewer players you can name, the more that math leans on your share guess. Most reliable with 3–5 dominant players; weakest in fragmented markets.
Volume × price
Market volume (units) × Price per unitHow many units the whole market moves in a year, times what one costs.
Watch out for Volume and price must be measured at the same point in the value chain — mixing retail volume with wholesale price silently distorts the total.
Analogous market
Analogous market × Analogous market size × Analogous market volume × Read-across factor (×)Take a neighbouring market somebody has measured, and scale it to yours.
Watch out for The most subjective top-down method — this is the one a skeptical reviewer will push on hardest, so the read-across factor needs the clearest documented reasoning.
From the value chain
Total chain volume (units) × Your layer's share of volume × Value content per unit × Yield / conversion (%) × Price captured per unit of contentFollow the volume through the chain and keep only the value your layer captures.
Watch out for Five inputs multiply together — an error in any one (yield is the easiest to overstate) compounds through all the others. Best used when you genuinely know the chain's economics, not as a first guess.
2. Bottom-up4 methods
Start from the customers and what each one spends, and add it up.
Reach for it when you can count the buyers, or count something that stands in for them — outlets, seats, vehicles, plants — even where nobody has published a total.
The trap is that each step is a share of the one before it, so an error in the first number survives every step after it. Four multiplied estimates is thinner ground than one measured figure, however precise the result looks.
Show the 4 methodsHide the methods
Customers × price
Customers × Price per customer / yrCustomers times what each pays a year. The bluntest read — best as a ceiling to check others against.
Watch out for The crudest bottom-up method — no funnel narrowing, no adoption curve. It gives you the whole market at that spend, so it is a ceiling check against Funnel or Segment build-up rather than a standalone estimate. And with no quantity between the two figures, a per-seat or per-unit price understates it by however much each customer gets through in a year.
Segment build-up
a list you build yourselfSize each customer segment on its own terms, then add them together.
Watch out for The price field here shows no currency symbol like other methods do — double-check what you're entering. Segments should be mutually exclusive; make sure none of your real customers are missing from every row.
Units × usage
Customers × Units per customer × Price per unitCustomers times how much each one gets through, times the price per unit.
Watch out for Models usage intensity, not just customer count — best for consumption-based or usage-based pricing where one customer buys many units. Don't double-count if this customer figure already reflects funnel narrowing done elsewhere.
Build it yourself
a list you build yourselfWrite the multiplication your market actually needs, one named factor per row.
Watch out for The factor carrying the most uncertainty is usually a rate in the middle, and it is invisible once it has been multiplied — write down where each one came from in its own note. Check the units read through: a build with no price in it multiplies out to a quantity rather than a market, and every factor after a division has to be on the same side of it as you meant.
3. Value-theory3 methods
Start from what the customer gains and how much of it you can hold — the way to price something with no market price to read.
Reach for it when nothing comparable is sold yet, so there is no volume to count and no going rate to read. This is how you price a thing that has no price.
The trap is that a researched number and a guessed one look identical once they are in the box. The value you create is an argument, not an observation — so write the argument down, because that is the part a reviewer can actually check.
Show the 3 methodsHide the methods
Value × capture
Value created per customer × Your capture (%) × Target customersWhat you are worth to a customer, times the share of that you can charge for.
Watch out for The most assumption-heavy value-theory method — value created is rarely a hard number. Write down the ROI logic behind it explicitly so a reviewer can check your reasoning, not just your conclusion.
Willingness-to-pay
Willingness-to-pay per customer × Target customersWhat a customer says they would pay, asked directly — with no value logic underneath it.
Watch out for Skips vt_cap's value-creation logic and asks for the price ceiling directly — easy to enter, but only as credible as the research behind it. A guessed number looks identical to a researched one in this tool.
Cost-of-alternative
Cost of the alternative per customer × Conversion from the alternative (%) × Target customersWhat they already spend on the thing you would replace, and how much of it you can take.
Watch out for “Conversion” here is a capture-rate assumption, not a literal % of customers switching — a common misread. Be explicit about which alternative you're benchmarking against.
4. Diffusion3 methods
Start from how fast this gets taken up, when the market is still being adopted rather than already served.
Reach for it when the question is not how big the market is but how big it becomes, and when — a market still being taken up rather than already served.
The trap is that the ceiling and the curve are two different claims, and only one of these three methods makes the second one. Knowing where adoption ends says nothing about the year you get there.
Show the 3 methodsHide the methods
Analog penetration
Addressable population × Peak penetration (%) × Price per adopter / yrHow far this eventually spreads through the population. A ceiling, with no date on it.
Watch out for A static end-state estimate — no timing. It tells you the eventual ceiling, not when you'll get there; pair with Bass diffusion if timing matters.
Bass diffusion
Ceiling population × Innovation coefficient (p) × Imitation coefficient (q) × Price per adopter / yr × Target yearAn adoption curve over time — some buyers arrive from awareness, the rest from word of mouth.
Watch out for The only method with real time dynamics — p drives early adoption from outside influence (advertising, awareness), q drives word-of-mouth adoption among people who've already seen it in use. Small changes to q swing the curve a lot; don't guess these without at least a rough literature comparison.
Look-alike ramp
Comparable product's users × Relative fit (%) × Price per adopter / yrBorrow a comparable product's real user count and discount it to how well it fits yours.
Watch out for Only as strong as the comparable you pick — a superficially similar but structurally different product (different price point, different buyer) will mislead. Document why the comparable is a fair one.
How many to build, and which
- 1
Build as many as your market supports
Keep going until another one stops changing your mind. Every method you tick joins the same reconcile — the reconciled TAM is the mean of it, or a figure you pick inside its range — so the answer is the set, not any one member of it.
- 2
A new starting point, not another number
The test is what you would go and find first. Two methods that need the same first figure are one read entered twice, and their agreement means nothing.
- 3
Start with the pair that has least in common
One top-down, one bottom-up, wherever both are possible: they can be wrong in completely different directions. Two is the floor, not the target — one method leaves the reconcile nothing to compare. When the sizing plan hands you a pair, it is handing you a place to start.
- 4
Stop when the spread stops moving
When a new method lands inside the range you already have and you cannot name a starting point you have not used, you are finished. A wide spread is not a failure — it tells you which assumption to go and settle next.
What your market’s stage rules out
Not every method can be built for every market, and the tool greys out the ones that cannot rather than letting you fill them in. One question decides it: does this market exist yet?
| Family | Establisheddemand realised and served today | Still being adoptedreal today, far from its ceiling | A new categorynothing bought under this description yet |
|---|---|---|---|
| Top-down | 5 of 5 | 5 of 5 | 2 of 5 usable |
| Bottom-up | 4 of 4 | 4 of 4 | 4 of 4 |
| Value-theory | 3 of 3 | 3 of 3 | 3 of 3 |
| Diffusion | none available | 3 of 3 | 3 of 3 |
A category nobody has named has no published total, no players to add up and no volume to count. What survives are the methods that need no history — read across from a neighbouring market, build up from customers you can identify, argue from the value created, or model the adoption curve.
What a finished sizing looks like
If these three are true, the number holds up under questioning:
- Every method is named
Not “market research says” — the specific build, and which family it started from.
- The reads started in different places
And landed together. Agreement between two versions of the same read is worth nothing.
- Every figure has a source beside it
A published report, a count you did, or a stated assumption. All three are acceptable; silence is not.
That is what turns a number into something a board, an investor or your own team can interrogate, challenge on the one assumption they disagree with, and then plan against — rather than one they can only take or leave.
← Back to the Market Model