When this framework fits
Use it when
- Capital, product, entry, or underwriting depends on the size of demand.
- The category boundary or unit of demand is contestable.
- Internal operating data can anchor at least one construction.
Do not overbuild it when
- A narrow customer test can answer the decision faster.
- The market is known and the binding question is ability to win.
- The product or buyer remains too undefined to count consistently.
Define what is being sized
Write the buyer, use case, geography, segment, demand unit, price basis, and time horizon before multiplying anything. Include substitutes. A market for "analytics" is not countable; a market for annual compliance workflow software purchased by regulated industrial sites in a named region is.
Keep three concepts distinct. TAM is the total demand under the defined problem and unit, assuming broad reach. SAM is the portion the current offer, geography, channel, and operating constraints can serve. SOM is the portion plausibly obtainable over a stated period, constrained by sales capacity, adoption, competition, implementation, and retention. SOM is not "one percent because that seems conservative."
Three independent constructions
1. Top-down, with visible filters
Start from a credible aggregate and decompose it through explicit filters for geography, buyer type, use case, price tier, and applicability. Each filter needs a reason and a source or stated assumption. The value of the path is not authority borrowed from a large published number. It is the ability to see exactly where the broad category becomes the market relevant to the decision.
2. Bottom-up, from the unit of demand
Count qualified buyers, locations, seats, transactions, devices, or workloads. Estimate adoption, frequency, and realized price. A Fermi decomposition exposes every factor before multiplication. Use internal pipeline, usage, conversion, and price data where available, but do not assume the current customer mix represents the full market.
3. Demand-side, from displaced behavior
Identify the money, time, risk, or existing tool spend the offer replaces. Demand-side sizing tests whether the proposed category corresponds to a real budget or costly behavior. It can also reveal that a large theoretical problem has little willingness to pay, or that value is concentrated in a narrower segment than buyer counts imply.
4. Reconcile before reporting
The estimates should disagree. Interrogate the difference: one path may include adjacent buyers, another may assume an unrealistic price, and a third may omit informal labor or a substitute. Do not average away the conflict. Adjust only where the evidence identifies a scope, data, or assumption problem. The reconciliation memo often contains more strategic value than the terminal number.
Sizing compliance workflow software
A top-down category estimate produces $310 million after broad regional and industry filters. A bottom-up build counts 12,000 qualified sites at an expected annual price of $17,000, or $204 million. A demand-side build based on current software, contractor, and administrative spend produces $174 million.
Reconciliation shows that the top-down path includes low-regulation sites with little need, while the demand-side path omits internal labor that buyers may value but will not fully convert into budget. Pricing interviews support a $14,000 to $18,000 range. The team reports a SAM of $170 million to $220 million, with site qualification and realized price as the dominant drivers. It does not report the average of the three original estimates.
Evidence requirements
| Input | Useful evidence | Quality question |
|---|---|---|
| Buyer count | Registries, filings, licensing data, internal account lists | Are duplicates, inactive entities, and ineligible buyers removed? |
| Adoption | Observed penetration, analog cohorts, conversion history | Do analogs share switching costs and purchase process? |
| Price | Transactions, quotes, budgets, and willingness-to-pay research | Is this realized price rather than list price? |
| Frequency or usage | Product events, transaction records, workflow volumes | Are seasonal and one-time effects separated? |
| Ability to capture | Sales capacity, implementation throughput, channel reach | Does SOM reconcile to an operating plan? |
Report a range with drivers
Use low, base, and high assumptions or a simple uncertainty simulation where the decision warrants it. Show the two or three inputs that dominate the range. Keep market size and adoption forecast separate: a market can be large while near-term penetration remains low. Attach confidence to the evidence, not to the number's visual precision.
Failure modes
Percent-of-a-large-number logic: the estimate says nothing about demand for the offer. Unstated filters: readers cannot reconstruct how the broad category narrowed. Analogy without structure: adoption is borrowed from a market with different switching costs or procurement. List-price sizing: discounts, mix, and purchasing power disappear. Ambition as SOM: target share is substituted for sales and implementation capacity. Two-sided blindness: a marketplace is sized as if only one side must change behavior.
Diligence-ready sizing checklist
- The buyer, use case, geography, unit, price basis, and substitutes are defined.
- TAM, SAM, and SOM answer different questions and use stated constraints.
- Three sizing paths were built independently before reconciliation.
- Every filter and multiplier is sourced or labelled as an assumption.
- Differences between paths are explained, not averaged.
- The range identifies its dominant adoption, price, or buyer-count drivers.
- SOM ties to channel, capacity, implementation, and time.
Limitations
A market size is an estimate, not a fact about the future. New categories may lack stable definitions, public datasets can lag, and internal demand may overrepresent early adopters. Triangulation does not guarantee truth if every path shares the same hidden assumption. The aim is inspectable reasoning and decision-relevant bounds, followed by monitoring as real demand arrives.
Related services, cases, and Insights
Build the number someone else can rebuild
We can turn a headline estimate into a sourced model, range, and decision-ready explanation.