Nano Banana 2.1 ist da | Präzise Bearbeitung, konsistente Charaktere.

Run a market sizing workflow

Market sizing is useful only when another person can trace every number back to a source or an assumption. An AI agent can shorten the research and calculation work, but it should never turn an unsupported guess into a confident market figure. This workflow shows how to calculate TAM, SAM and SOM, document uncertainty and produce an output that a founder, investor or product team can audit.

Quick answer

TAM is the full annual demand if every relevant customer bought the offering. SAM narrows that demand to the customers your product and operating footprint can actually serve. SOM is the share you can plausibly win within a defined time horizon. Use an AI agent to collect evidence, normalize units, test formulas and expose missing assumptions—not to invent a market total.

TAM, SAM and SOM are three different decisions

MeasureQuestion it answersTypical calculationWhat must be stated
TAMHow large is the full relevant market?Potential customers × annual spendCustomer definition, geography, year and currency
SAMWhich part can this product serve?TAM filtered by product, region, channel and constraintsEvery inclusion and exclusion rule
SOMWhat can the business realistically capture?Reachable accounts × conversion × annual revenueTime horizon, capacity, channel and conversion assumptions

A common mistake is to present an industry report's headline number as TAM, then apply arbitrary percentages for SAM and SOM. The result may look tidy, but it does not show who buys, what they pay or why the business can reach them.

Before the agent starts, write a one-sentence market definition

Use a sentence that specifies the buyer, problem, offering, geography and period. For example: “Independent online retailers in the United States buying AI product-photo tools during 2027.” This prevents the agent from mixing global creator software, enterprise design suites and e-commerce photo services into one number.

Also define the unit of demand. It might be businesses, seats, monthly active creators, transactions, images or annual contracts. Choose one primary unit and convert every source into that unit before calculating.

A six-step AI market sizing workflow

1. Lock the scope and decision

State whether the estimate supports a product launch, budget, fundraising narrative or sales plan. Record the target year, geography, customer segment, currency and price basis. A market can change dramatically when one of these fields changes.

2. Build an evidence table before calculating

Ask the agent to create columns for claim, value, unit, geography, period, source URL, publication date, extraction note and confidence. The U.S. Small Business Administration recommends checking demand, market size, location, saturation and pricing; it also distinguishes broad existing sources from direct customer research. Use that logic to keep demographic, economic, competitor and interview evidence separate.

3. Calculate bottom-up first

Bottom-up sizing begins with countable customers and a defensible annual spend. If the business sells seats, use eligible companies × average seats × annual price. If it sells usage, use eligible customers × annual transactions × revenue per transaction. Keep the price assumption separate from the customer count so each can be tested independently.

4. Run a top-down cross-check

Use an industry total only as a second method. Filter it by the exact segment, geography and product category, then compare it with the bottom-up estimate. A large gap is not something the agent should average away; it is a signal to inspect definitions, years, currencies or double counting.

5. Convert SAM into a capacity-based SOM

Build SOM from the sales and delivery system: reachable accounts, qualified rate, win rate, onboarding capacity, retention and the selected time horizon. If the business can contact only 20,000 accounts in three years, a SOM based on 10% of a million-account SAM is not operationally credible.

6. Make the agent challenge the result

Require a second pass that identifies unsupported claims, conflicting sources, hidden percentages and stale data. The final output should include a base case plus conservative and upside scenarios, not one false-precision number.

Source reliability ladder

LevelUseful sourcesBest useRisk to flag
Primary public dataGovernment statistics, regulators, filingsPopulation, business counts, revenue and trade dataDefinitions may not match your segment
First-party market evidencePricing pages, customer data, interviews, surveysSpend, conversion, behavior and constraintsSample bias or unpublished methodology
Credible secondary researchNamed research firms, trade bodies, academic workCategory trends and cross-checksPaywalled methods or reused estimates
Discovery-only materialSearch snippets, unsourced summaries, AI answersFinding leads and terminologyDo not use as final evidence

For U.S. work, useful starting points include SBA market research guidance, Census data and Bureau of Labor Statistics data. The source still has to match the market definition; an authoritative source can be irrelevant to the exact segment.

Worked example with explicit assumptions

Assume a fictional AI product-photo service targets 1,200,000 eligible online retailers at an annual spend of $240. Its TAM is 1,200,000 × $240 = $288 million. The current product supports only 240,000 retailers in its selected countries and channels, so SAM is 240,000 × $240 = $57.6 million.

Over three years, the team can realistically reach 80,000 accounts. If 20% qualify and 30% of qualified accounts buy, the reachable customer count is 4,800. SOM is 4,800 × $240 = $1.152 million in annual revenue. These numbers are examples, not market facts; the value of the model comes from replacing each assumption with evidence.

TAM = total eligible customers × annual spend
SAM = serviceable eligible customers × annual spend
SOM = reachable accounts × qualification rate × win rate × annual revenue per customer

Copyable AI agent task brief

You are a market research analyst. Estimate TAM, SAM and SOM for [offering].

Market definition: [buyer + problem + offering + geography + year]
Primary demand unit: [businesses / seats / transactions]
Price basis: [annual price or revenue per unit]
Decision supported: [launch / budget / fundraising / sales plan]

Requirements:
1. Build an evidence table with claim, value, unit, geography, period, source URL, publication date, extraction note and confidence.
2. Prefer primary public data and first-party evidence. Treat search snippets and AI summaries as leads only.
3. Calculate a bottom-up base case and show every formula.
4. Use a top-down estimate only as a cross-check and explain definition gaps.
5. Build SOM from reachable accounts, conversion, capacity and a stated time horizon.
6. Produce conservative, base and upside scenarios.
7. Flag unsupported assumptions, conflicting sources, stale data and double counting.
8. Never invent a source or fill a missing value without labeling it as an assumption.

What the final deliverable should contain

A useful market sizing brief includes the one-sentence market definition, source table, calculation sheet, scenario table, assumption log and a short explanation of the largest uncertainty. Separate observed values from calculated values and assumptions. Add the retrieval date to web sources so the work can be refreshed later.

Common errors to catch

Mixing years or currencies: normalize them before multiplying. Counting users and businesses together: choose one demand unit. Using a percentage without a mechanism: explain the product, geography, channel or capacity filter. Double counting segments: make categories mutually exclusive. Treating AI output as a citation: open and inspect the original source.

FAQ

Can AI calculate market size accurately?

AI can organize evidence and calculations, but accuracy depends on the market definition, source quality and reviewed assumptions.

What is the simplest TAM formula?

For a customer-based model, TAM equals the number of eligible customers multiplied by annual revenue per customer.

How is SAM different from TAM?

SAM removes customers the current product, geography, channel or operating model cannot serve.

How should SOM be calculated?

Use reachable accounts, qualification, conversion, capacity and a stated time period instead of an arbitrary share of SAM.

Should I use top-down or bottom-up market sizing?

Use bottom-up for the core estimate and top-down as a cross-check; investigate rather than hide large differences.

Can an AI answer be used as a source?

No. Use it to find possible sources, then cite and verify the original publication or dataset.

How often should a market sizing model be updated?

Update it when pricing, product scope, geography or a major source changes, and record the new retrieval date.

Run a market sizing workflow

Tool used in this workflow

iMini Agent Market Sizing

Organize sources, assumptions and calculations into a reviewable market sizing brief.

Open tool · Run market sizing