Build a decision-useful market range when published market reports use opaque definitions or incompatible totals.
A market size is a model with a boundary, not a discovered fact. Census County Business Patterns provides establishment, employment and payroll data by industry and geography [1]. The Census quarterly e-commerce program explains its sample, coverage, weighting, imputation and exclusions [2]. BLS documents QCEW classifications, revisions and imputation resources [3]. These sources are valuable because their universes and limitations are inspectable. They still do not define a product's addressable market. The analyst must translate observable establishments, workers or transactions into eligible buyers and annual spend, then reconcile that bottom-up result with a separately bounded top-down pool.
Write the boundary before the number
Specify buyer, product, geography, channel, price basis, time period and exclusions. Distinguish total economic activity, serviceable available market and realistically obtainable market. Decide whether the unit is revenue received by vendors, customer spend, gross transaction value or budget. A marketplace's gross value and its fee revenue cannot share one market total.
Choose observables that align with the buyer. CBP counts employer establishments and excludes several industries and most government employees [1]. Its establishment is a location, not necessarily a buying account. QCEW may support employment and wage intensity, but its revisions and classification rules must be preserved [3]. E-commerce sales include specified internet-negotiated retail transactions and exclude categories such as online travel and financial brokers [2]. Every source therefore requires a coverage adjustment rather than blind use.
- Top-down base: a measured spending pool with explicit category coverage.
- Bottom-up base: eligible accounts or units times adoption and annual spend.
- Bridge: exclusions, overlap, channel share, nonemployers and price basis.
- Range: low, base and high values tied to named assumptions.
Build two independent estimates
For top-down, begin with an official or filed total and subtract categories outside the definition. Do not apply a market-share percentage from a source with a different universe. For bottom-up, segment accounts by size or need, assign an eligible share, adoption probability and annual spend. Use observed customer prices or dated procurement values when available. Keep assumptions visible and avoid false precision.
Worked example, explicitly hypothetical: a workflow tool serves 20,000 eligible establishments. The bottom-up case assumes 35% adoption and $6,000 annual spend, producing $42 million. A top-down observable spending pool is $120 million, but 40% covers excluded services and 20% of the remainder belongs to non-target buyers, producing $57.6 million. The $15.6 million gap is not averaged away. Investigation shows the bottom-up model counts headquarters while the top-down pool includes branch purchases. The decision range becomes $42–58 million, with branch-level buying authority named as the resolution question.
Reconcile through definitions
Create a bridge table where every difference has a sign, amount or range, evidence and owner. Test double counting among multi-location firms, reseller revenue and bundled products. Confirm reference years: CBP is annual and arrives with a lag, while other sources may be quarterly or preliminary. Do not combine a current price with an old account count without showing the mixed dates.
Use sensitivity analysis on the assumptions that drive the result, usually eligibility, adoption and spend—not on every input equally. A market is decision-ready when the range changes a product, sales-capacity or research decision and the next observation can narrow it. This is planning analysis, not investment advice. If the top-down and bottom-up estimates still differ materially after boundary reconciliation, publish the disagreement and fund the missing measurement.
Take it into the meeting
- Define buyer, unit and exclusions before sizing.
- Build top-down and bottom-up estimates independently.
- Investigate discrepancies instead of averaging them.
Sources & boundaries
Source statements are attributed; the decision process is Signal Atlas analysis. Examples marked hypothetical are teaching inputs, not observed outcomes.
- The worked market and all numbers in it are hypothetical.
- Official datasets have coverage, lag, revision and disclosure-protection limits.
- This framework is not investment advice.
- County Business Patterns: 2022U.S. Census Bureau · Source publication: 2024-06-27 · Retrieved 2026-09-19
Establishment, employment and payroll observables Industry and geography segmentation Disclosure-avoidance and reference-year caveats
- Quarterly E-Commerce Report — About the SurveyU.S. Census Bureau · Source publication: not established · Retrieved 2026-09-19
E-commerce survey universe and exclusions Sampling, weighting and imputation method Preliminary-estimate and revision caveat
- QCEW Methods OverviewU.S. Bureau of Labor Statistics · Source publication: not established · Retrieved 2026-09-19
Classification and data-documentation resources Revision tracking Imputation and reporting-rate considerations