How to size a market: TAM, SAM and SOM without fooling yourself

Market sizing is where a lot of early-stage thinking quietly goes wrong. A number is produced from a published industry report, a small percentage is applied to it, and the result becomes a slide rather than a decision input. The purpose of sizing a market during Evaluate is narrower and more useful than that: to establish whether the opportunity is large enough to be worth the cost and risk of building a company, and to expose what you are assuming about customers and price.

The three definitions

The relationship matters. TAM sets the ceiling, SAM sets the field of play, and SOM sets the plan. Investors discount TAM heavily because it is easy to inflate. SOM is where credibility is won or lost, because it forces you to state a route to customers and a rate of acquisition.

Bottom-up beats top-down

Top-down sizing starts with a published market figure and cuts it down. Bottom-up sizing starts with a customer, a price and a count. Bottom-up is better for three reasons: the assumptions are visible, each one can be tested during customer discovery, and the calculation doubles as the first version of your revenue model.

A worked example

Suppose a venture sells compliance software to regulated mid-sized firms in the UK. The illustrative arithmetic below is deliberately simple — the discipline is in defending each input, not in the multiplication.

The SOM line is the one that changes decisions. It is derived from acquisition capability rather than ambition, which means it can be argued with — and anything that can be argued with can be tested.

Five mistakes that make sizing useless

• Sizing the industry rather than the spend on the specific problem you solve.
• Using a headline market report figure without checking what it includes.
• Applying an arbitrary "one per cent of the market" capture rate.
• Pricing at what you would like to charge rather than what the problem currently costs the buyer.
• Sizing once, then never revising it as customer discovery changes the picture.

What to do with the answer

Sizing produces one of three outcomes, and all three are valuable. The market is clearly large enough and the venture proceeds to Create. The market is clearly too small for the model, and the venture stops — cheaply, at the right stage. Or the market is large enough only under a different segment, price point or model, in which case the sizing has just told you what to change.

That third outcome is the most common and the most useful. It is also why we treat evaluation as a decision-making stage rather than a documentation stage, as set out in validating an opportunity before you build.

Sizing a market that does not exist yet

The hardest sizing problems are the ones where the category is new, because there is no published figure to argue with and no incumbent revenue to divide up. The instinct is to size an adjacent market instead, which produces a large number and no information. A better approach is to size the substitute: whatever the buyer does today in place of the thing you intend to sell. Every unsolved problem is currently being handled by somebody — usually by people, spreadsheets and time — and that cost is measurable, defensible and directly comparable to your price.

Where the opportunity comes from regulatory change, the substitute may not exist yet either. In that case the sizing input is the cost of non-compliance and the number of organisations exposed to it, both of which are usually a matter of public record. This is one of the practical advantages of the opportunity sources described in where startup opportunities actually come from: each one implies a different, checkable sizing input.

Sizing as a live model, not a slide

We treat the sizing model as a working document that is revised every time customer discovery changes an assumption. Contract values move when buyers tell you what they actually pay. Segment counts move when you discover which firms are genuinely in scope. Capture rates move when the first acquisition channel produces real conversion data during Launch. A sizing model that has not changed between Evaluate and Launch has almost certainly not been used.