# Market sizing
Use this method to create a transparent estimate that helps a real decision. The goal is not the largest number; it is a model whose inputs, boundaries, and limits a reader can inspect.
Read `references/sizing-methods.md` before calculating. It contains the definitions, formulas, source hierarchy, calibration rules, failure checks, and confidence rubric. Read `references/host-operations.md` only when current external evidence must be collected.
Choose the decision lens
Choose one lens before selecting a denominator, and keep it consistent:
- **Product:** vendor revenue pool; the buyer purchases the product. - **Services:** addressable buyers × engagements per buyer per period × average engagement value; the buyer purchases one or more service engagements. - **Operator:** acquirable revenue or EBITDA; the decision is whether to build, buy, or operate a business.
Do not substitute a software-market total for a services opportunity or an industry revenue total for an operator opportunity. If the request does not establish the lens and different choices would materially change the answer, ask once before calculating.
Method
1. State the decision, lens, unit of sale, buyer population, geography, period, and relevant capacity or distribution constraints. Define TAM, SAM, and SOM for this decision before calculating them. 2. Choose one or more appropriate models. Treat top-down estimates as ceilings or cross-checks, not sufficient answers. Use bottom-up sizing when direct coverage is missing or suspicious. Use value-theory sizing for a genuinely new category; comparable pricing is an input to willingness to pay, not a separate sizing method. Prefer two independent methods for a decision-critical estimate when evidence permits. 3. Gather dated primary inputs first: official population data, public filings, published pricing, attributable transaction or operating evidence, and supplied first-party data. Use analyst estimates as cross-checks. 4. Show each calculation, unit, conversion, and assumption. Do not multiply incompatible populations, prices, or time periods. 5. Separate the theoretical market from the reachable market. Explain filters for buyer fit, geography, readiness, route to market, sales cycle, and capacity. 6. Reconcile materially different methods. Keep the range when the evidence cannot justify a point estimate. 7. Assign every reported figure a 1–5 confidence grade using the rubric in the reference. Name the limiting factor and the inputs that would most change the answer. 8. Recommend a decision or next validation action proportionate to confidence.
Calculation and evidence discipline
For services models, state the engagement frequency and period explicitly; do not silently assume one engagement per buyer. For each model, show the formula, every unit and conversion, source date, geography, period, and assumption. Keep a calculation table that makes ranges reproducible from stated inputs. Use low/base/high scenarios where an input is uncertain, and name the sensitivity driver. Classify inputs as observed, reported, modeled, or unknown; identify proxies and vendor claims rather than presenting them as measured facts.
Reconcile conflicts by checking market boundary, segment, timing, and unit. Do not average incompatible estimates to create false precision. Reject the “1% of a large market” shortcut unless share follows from stated acquisition and capacity assumptions. A result is not a forecast, valuation, or commitment to enter a market.
When a supplied customer or organization anchors the question, size the buyer category one or two levels above that example. The example may appear as a sub-band inside SAM; it must not define the entire denominator. Do not widen the category so far that its buyers no longer share a concrete job or purchase.
Capture source URLs while researching. Cite each non-trivial external input at the point of use, cite it again when the claim is restated, use one canonical source per claim where possible, reproduce all sources at the end, and verify each URL-to-claim pairing before finalizing. If evidence is thin, identify the missing input and provide a bounded scenario or order-of-magnitude estimate with the appropriate low confidence grade; do not silently fill the gap.
Output
```markdown # [Market] sizing ## Decision, lens, and definitions ## Inputs and sources ## Method 1 ## Method 2 or reason unavailable ## Reachable opportunity ## Per-figure confidence and sensitivity ## Family confidence and limiting factor ## Recommendation and next validation ## Sources ```
Quality rules
- Label all modeled values as estimates, ranges, or scenarios. - Keep product, services, and operator denominators separate. - Tag every source as vendor-published, analyst-independent, analyst-sponsored, industry-press, academic, customer-reported, primary-interview, or another clearly defined type. - Do not use an empty search result as proof that a market or competitor does not exist. - Use supplied sources only when current research access is unavailable, and state the limitation.