Market sizing is the most abused step in geographic market entry. It is easy to produce a “big number” that makes a market look exciting and just as easy to produce a “small number” that justifies staying home. Neither helps you make a good decision. What you need is a sizing approach that is explicit about definitions, uses multiple methods, and produces ranges tied to confidence levels. In other words, you need a sizing system that prevents self-deception.
This chapter shows you how to size and forecast demand in a way that supports real choices: which segments to target first, what entry mode can be justified, how fast you can plausibly ramp, and what must be true for returns to clear your hurdle rate. We will keep the math simple and the logic rigorous. A good market model is not complex; it is auditable. If an executive cannot trace the path from a handful of drivers to the final number, you should assume the model is hiding fragility.
4.1 Market Definition: Boundaries, Substitutes, And Segmentation Rules
The first fight in market sizing is never the spreadsheet. It is the definition. If you define the market too broadly, you inflate TAM and tell yourself you can win shares from substitutes that customers do not actually consider. If you define it too narrowly, you miss adjacent profit pools and under-invest. The answer is not “pick a definition and move on.” The answer is to define the market in a way that matches how customers buy and how you would compete.
Start with the customer’s problem, not your product category. Ask: what job is the customer trying to get done, under what constraints, and what alternatives do they consider? This pushes you to include true substitutes and exclude false ones. It also helps when products are bundled differently across countries or when local competitors sell a lower-feature “good enough” version that still competes for the same budget.
Set three explicit boundary decisions and write them into your model so they cannot drift as the analysis evolves. Customer boundary: which customer types are included (and excluded), and why. Use-case boundary: which applications you will count as in-scope, and how you treat multi-use products. Economic boundary: whether you are sizing end-customer spend, manufacturer revenue, or your reachable economics net of channel margins. The last boundary is often ignored, and it creates the most dangerous illusion: a large “market” that is not monetizable at your required margin.
Make one more definition choice that prevents endless debate later: value versus volume. Value basis: dollars of spend or revenue. Volume basis: units, transactions, tons, users, or hours. If you can, build both. Volume is often more stable for triangulation (especially when prices vary by channel or contract), while value is what you need for the business case. When you convert volume to value, be explicit about whether prices include taxes, freight, installation, financing, or service. Those items swing results and often vary country to country.
Next, make substitutes explicit. Not every alternative is a substitute worth sizing. Use a simple test: would a rational buyer switch to that alternative if your offer were unavailable or 20% more expensive? If yes, it is a substitute in the decision set and should be reflected in your SOM assumptions (because it limits obtainable share). If not, treat it as a different budget category and keep it out of TAM, even if it looks related on paper.
Segmentation is the second place where models fail. Teams segment by what is easy to label (industry codes, income brackets, company size) rather than what predicts adoption and economics. Your segments should be designed for decision-making: they should tell you where to start, how to sell, and what it will cost to serve.
Use three segmentation rules as non-negotiables. Mutually exclusive: each customer or transaction belongs to one segment. Collectively exhaustive: the segments cover the full in-scope market, including an “other” bucket if needed. Economically meaningful: segments differ in at least one key driver such as willingness-to-pay, buying process, channel access, frequency of purchase, or service intensity.
In geographic entry, a practical segmentation stack often looks like this: customer type (who buys), use case (what they need), and route-to-market (how you reach them). Add a fourth layer only if it materially changes economics, such as regulated versus unregulated buyers, urban versus rural distribution, or enterprise versus SMB procurement. Keep segments limited. If you need more than eight to explain the market, your model will become unmanageable and your go-to-market will lose focus.
Finally, define how you will treat cross-border and informal activity. In many countries, a meaningful portion of demand is served through gray channels, parallel imports, or informal providers. Ignoring this can overstate the attractiveness of “official” channels; including it can overstate what you can capture if you will not operate in that space. Make the choice explicit: Reportable market: the official, measurable portion. Economic market: the true consumption including informal channels. Use the economic market to understand the competitive reality, but base your SAM and SOM on what you can legally and practically serve.
4.2 Sizing Methods: Top-down, Bottom-up, And “consumption-based” Triangulation
A single sizing method is a single point of failure. Top-down models can look clean and be wildly wrong because the proxies do not hold locally. Bottom-up models can feel “real” and still miss parts of the market because the customer count or usage assumptions are incomplete. Consumption-based models can capture reality but depend on behavior inputs that are hard to validate early. The solution is triangulation: use all three methods, reconcile the differences, and produce a range that reflects uncertainty honestly.
Top-down sizing starts with an external aggregate (a published market size, national spend category, import value, installed base) and refines it to your definition using ratios and filters. It is fast and comparable across countries, but definitional mismatch is common. Keep the driver chain short and auditable: anchor, scope adjustment, reachability adjustment, and economic adjustment.
A practical top-down sequence looks like this. Anchor: choose a number that is closest to your category, even if imperfect. Scope adjustment: remove out-of-scope segments, substitutes, or export volume if the anchor includes them. Reachability adjustment: remove customers you cannot serve due to regulation, channel access, or your guardrails. Economic adjustment: convert end-customer spend to your revenue basis (net of channel margins, taxes, or pass-through costs) so the number is usable for your business case.
Bottom-up sizing builds the market from the ground up using counts and unit economics. It is the best method for a decision-ready SAM and SOM because it forces you to tie demand to identifiable buyers and a plausible selling motion. The key is to avoid “hero assumptions” that cannot be validated. The bottom-up should be driven by observable denominators: number of target accounts, number of locations, number of relevant assets, number of procedures, or number of transactions.
Build bottom-up in five moves. Define the unit: what you will count (accounts, sites, users, devices, tons, shipments). Count the units: using registries, industry lists, or credible proxies. Apply eligibility: which portion of units match your target segment and constraints. Apply adoption: expected penetration over time, with a ramp curve, not a single percentage. Apply value: price or annual spend per adopted unit, reflecting local willingness-to-pay and channel economics.
Consumption-based sizing is a specific flavor of bottom-up that uses usage intensity rather than customer counts. It is especially useful when the market is fragmented, when customers are hard to enumerate, or when consumption follows physical or digital activity. It asks: how much of the product or service is consumed per unit of activity, and what drives that activity? For example, consumption might scale with miles driven, square footage built, or transactions processed.
A consumption-based model typically uses one of three forms. Installed base × replacement: units in operation times replacement rate times price. Activity × intensity: activity volume times usage per activity times price. Population × penetration × frequency: target population times adoption rate times purchase frequency times price. Use whichever matches the physics of your category. Then calibrate the coefficients using a comparable market where you have better data, and adjust only when you have a clear reason to believe behavior differs locally.
Triangulation is where discipline shows. You should expect the three methods to differ. The goal is to explain the gaps, not to average them away. Often, top-down provides the sanity-check on total demand, consumption-based provides the reality-check on volume, and bottom-up provides the most decision-relevant SAM and SOM because it links directly to segments and go-to-market.
Reconcile in three steps. First, align definitions: ensure all methods are sizing the same boundary (value vs volume, end spend vs supplier revenue, gross vs net). Second, identify the swing factors: which assumptions explain most of the gap (penetration, price, unit count, usage coefficient). Third, choose a range and a confidence rating, and turn the remaining gaps into a short validation plan for the next sprint.
Do not let currency and price basis become a hidden source of error. State the basis once: Currency: local currency for planning; Time basis: nominal or real; Price point: ex-factory, distributor sell-in, retailer shelf, or customer total cost. Note whether VAT is included, and sanity-check anchors (trade statistics, published prices) against what customers actually pay.
Finally, convert TAM to SAM and SOM with explicit logic. TAM: total in-scope demand under your definition, regardless of your ability to serve it. SAM: the portion you can serve given product fit, channel access, and constraints. SOM: the portion you can realistically win over a defined horizon given competition and your capacity. Build SOM from drivers—coverage, conversion, retention, and ramp—not from a generic “percent of TAM.” In early years, SOM is often constrained by execution capacity (hiring, delivery, service), so make that constraint explicit.
4.3 Growth Drivers And Demand Shifts: Macro, Category, And Customer Behavior
Market size is a snapshot; entry decisions depend on the movie. Growth forecasting is where teams either anchor on generic GDP growth or build fragile five-year projections. The right approach is to separate what you believe is structurally true from what you are betting on, and to connect growth to a small set of drivers you can explain.
Start with three layers of drivers. Macro drivers: factors that change the overall activity level, such as population, income, industrial output, construction activity, or digital connectivity. Macro drivers matter most when your category scales with broad activity. Category drivers: penetration, substitution, and regulation that shift spend into or out of your category. Examples include new standards, reimbursement rules, import restrictions, electrification, or platform shifts that make a new solution viable. Customer behavior drivers: changes in buying preferences, channel mix, procurement centralization, and switching costs that change how fast demand moves.
Quantify growth using a “base plus shifts” structure. Base growth: the growth you would expect if the market kept its current penetration and mix, driven primarily by macro activity. Penetration shift: the change in adoption of the category versus substitutes. Mix shift: the change in segment mix toward higher- or lower-value use cases. Price and value shift: inflation and willingness-to-pay changes net of discounting pressure. This structure forces you to state what is driving the forecast rather than hiding behind a single CAGR.
Avoid using one straight-line CAGR for every segment. Growth in early-adopter segments often decelerates as they saturate, while laggard segments can accelerate when a channel becomes reliable or regulation shifts. Use an S-curve when adoption is behavior-driven, and a replacement cycle when demand is asset-based. Then stress-test with a downside case where adoption stalls.
Then explicitly consider demand discontinuities. Geographic markets often experience step-changes: a regulation enables a category, a government procurement program accelerates adoption, a supply shock constrains volume, or a dominant platform changes distribution economics. The mistake is to ignore discontinuities because they are hard to model; the better practice is to model them as scenarios with triggers and probabilities. A good forecast rarely needs more than three scenarios: Base: most likely trajectory, Upside: one or two catalysts arrive earlier or stronger than expected, and Downside: adoption slows due to regulation, channel friction, or pricing pressure.
Finally, translate growth into what matters for your go-to-market. Growth in the wrong segments does not help you. If your early beachhead is enterprise customers, you should forecast enterprise segment growth, not the whole market. If your route-to-market is through modern retail, growth in informal channels may not be reachable. Tie your forecast to your SAM, not your TAM, and be explicit when the market is growing but your reachable portion may not be.
4.4 Sensitivities And Confidence Levels: What Must Be True For The Plan To Work
The output of sizing should not be a single number; it should be a decision-ready range plus an explanation of what drives it. The practical way to do that is to identify the few variables that dominate outcomes, run sensitivities, and attach confidence levels that match your evidence. This is how you avoid the most expensive market entry mistake: scaling a plan built on untested, high-leverage assumptions.
Start by linking your market model to your business case model. If market sizing says the SAM is $200M but your business case assumes you can win $80M in three years, the implied share is 40%. That may be possible in a winner-take-most category, but in most markets it is a red flag. Do this implied-share check early and repeatedly. It is the fastest way to spot self-deception.
Next, run three sensitivity tests that are easy to explain and hard to game. Tornado test: vary each high-leverage assumption up and down (price, penetration, unit count, channel margin, usage coefficient) and see what moves the result most. Threshold test: compute the breakpoints that make the entry unattractive, such as the minimum adoption rate or price required to hit hurdle returns. Time-to-proof test: estimate how long it will take to validate the key assumptions in-market, because slow learning is a hidden cost.
Attach a confidence rating to each major driver, not just to the final number. Use a simple scale: High confidence: supported by multiple independent sources or direct customer evidence. Medium confidence: supported by one good source plus reasonable comparables. Low confidence: largely inferred from proxies or a small number of interviews. Then focus your next sprint on upgrading confidence on the highest-leverage, lowest-confidence drivers. This turns market sizing into a learning plan rather than a debate.
When you present ranges, show what is inside and outside the range. A range is not a hedge; it is a statement of uncertainty. If the range is wide because price is unknown, say that, and show the price range explicitly. If the range is wide because the customer count is uncertain, show the denominators and the source quality. Leaders are generally comfortable with uncertainty when it is transparent. They become uncomfortable when uncertainty appears only after a commitment has been made.
Finally, convert sensitivities into “must be true” statements that connect directly to your entry thesis. For example: “Net price after channel margins must be at least $X,” “Regulatory approval must be achievable in Y months,” or “Beachhead segment penetration must reach Z% by month 18.” These statements become your gating criteria and stop-loss triggers. If you cannot articulate the must-be-true conditions, your model is not yet decision-ready.
4.5 Market Sizing Template + Red Flag Checklist
A market model should be reusable, auditable, and readable. Use this template to structure the work so leaders can see what you did, what you assumed, and what you still need to learn.
- Definition: market boundaries (customer, use case, economics), value/volume basis, and treatment of substitutes and informal channels.
- Segmentation: 4–8 segments with rules, key drivers per segment, and a reconciliation to the total.
- TAM: range from at least two methods, with sources and confidence notes.
- SAM: reachable portion with explicit constraints (regulatory, channel, product fit) and time horizon.
- SOM: obtainable share built from capacity and competitive reality (coverage, conversion, retention, ramp).
- Growth: base plus shifts forecast for SAM, with scenarios and triggers.
- Sensitivities: top swing factors, thresholds, and must-be-true statements linked to validation actions.
Red flag checklist: If you see any of these, stop and fix the model before you let it drive a decision.
- Definition drift: TAM is end-customer spend but SOM is expressed as supplier revenue, or value and volume are mixed without conversion logic.
- Double counting: the same demand appears in multiple segments, channels, or use cases.
- Unbounded penetration: adoption assumptions exceed what comparable markets have achieved without a clear structural reason.
- Share without mechanism: SOM is a percentage of TAM with no coverage, conversion, or capacity story.
- Ignoring channel economics: gross market spend is treated as your revenue, without accounting for distributor margins, taxes, or required service cost.
- Single-source certainty: one report or one expert interview drives the entire model without triangulation.
- Flat ramp fantasy: the model assumes immediate steady-state penetration rather than a ramp that reflects sales cycles and trust-building.
- No downside case: sensitivities exist, but no realistic downside scenario is modeled for the key uncertainties.
