Build market size from a buying unit

'The market is worth billions' is usually a category fact with no purchase path attached. A founder needs a smaller model: who can buy, how many such units exist, what triggers a purchase, what one unit can spend, and how many units the chosen channel can plausibly reach.

This complements the company-type discussion in company-design chapter, where the natural ceiling matters more than startup fashion.

Define the buying unit

The buying unit might be a firm, location, clinic, team, household, project, device, or transaction. Pick the level at which a budget and decision recur. A chain with 200 locations may be one enterprise contract, 200 site purchases, or a hybrid. The model changes accordingly.

Write the unit as a sentence: 'A US employer establishment in NAICS X, with 20–99 employees, operating in these states, where this role owns the workflow.' NAICS is an official production-oriented classification for business establishments, not a customer taxonomy (Census NAICS). Use it as a counting scaffold, then narrow with actual buying conditions.

Count, then subtract

County Business Patterns provides establishment counts, employment, payroll, geography, industry, and size bands for US employer businesses (Census CBP datasets). It excludes some industries and nonemployers and uses disclosure protections; those limits travel with any estimate. BLS's QCEW offers quarterly employment and wage counts by industry and geography and covers more than 95% of US jobs (BLS QCEW). Neither dataset tells you who has the pain, budget, technical readiness, or permission to buy.

Build three filters: eligible by observable structure, addressable now by workflow and constraints, and reachable by channel and capacity. Do not hide those reductions inside one adoption percentage.

Worked hypothetical model

Hypothetical, not a real TAM: A startup sells exception-triage software to regional distributors. Its research finds 2,400 establishments in a chosen industry-and-size slice. The founder estimates 55% run the relevant workflow, 60% of those use compatible systems, and the initial sales territory can reach 25% within three years.

Reachable units = 2,400 × 0.55 × 0.60 × 0.25 = 198. At a modeled annual contract value of $12,000, reachable annual contract capacity is $2.376 million. This is not forecast revenue. It is the revenue ceiling if every reachable eligible unit buys at the assumed value.

Sensitivity matters more:

The large spread is the finding. It identifies which assumptions need field evidence.

Market-model artifact

Keep one auditable sheet with:

• buying-unit definition and exclusions;
• source year, geography, industry code, and size band;
• establishment-versus-enterprise treatment;
• workflow-fit evidence and sample;
• technical or regulatory eligibility;
• purchase frequency and annual value;
• reachable accounts by named channel and founder capacity;
• low, base, and high values for every uncertain input;
• owner and next test for the two most sensitive assumptions.

Make a decision, not a poster

Use the model to decide whether the market supports the intended company shape, whether a narrower segment is commercially sufficient, and which assumption could collapse the case. Do not add adjacent segments merely to improve the total. Model each with its own buyer and economics.

Limits

Public datasets lag, classifications blur workflows, establishments can double-count enterprises, and private or international markets require different sources. The model also omits competitive response and price erosion. It is a structured estimate, not a valuation, investment conclusion, or promise of obtainable revenue.