Market Definition, Sizing and Demand Forecasting

Introduction

Entrepreneurs make decisions about products, pricing, investment, staffing and distribution on the basis of expected demand. Yet “the market” is not a single, naturally occurring number. It is an analytical construct defined by a particular customer problem, product or service, geography, time period, channel and competitive context. A poorly defined market can make a small opportunity appear enormous or cause an attractive niche to be overlooked.

Market sizing estimates the economic opportunity associated with a defined market. Demand forecasting estimates how much customers are likely to buy, when they will buy it and under which conditions. These activities are related but distinct: market sizing describes opportunity, whereas forecasting supports decisions about future demand. Both require explicit assumptions, credible evidence and systematic treatment of uncertainty.

Learning Objectives

By the end of this chapter, you should be able to:

1. Define a relevant market using customer, need, product, geography, channel and time boundaries.
2. Distinguish total addressable market (TAM), serviceable available market (SAM) and serviceable obtainable market (SOM).
3. Apply top-down, bottom-up and value-theory approaches to market sizing.
4. Identify demand drivers, adoption barriers, seasonality and substitution effects.
5. Construct forecasts using growth rates, adoption models, moving averages and scenarios.
6. Evaluate sensitivity, uncertainty, data quality and ethical risks in market analysis.
7. Translate market evidence into entrepreneurial decisions.

1. What Is a Relevant Market?

A relevant market is the group of buyers and suppliers considered together for a specific decision. It is bounded by:

• Customer group: households, students, firms, hospitals or another segment.
• Job or need: the problem customers seek to solve, such as commuting, accounting or food delivery.
• Product category: the solutions that can satisfy that need.
• Geography: local, national, regional or global coverage.
• Channel: direct sales, retail, marketplaces, distributors or mobile applications.
• Time: a month, year, season or forecast horizon.
• Transaction definition: units sold, subscriptions, users, orders or revenue.

A market definition should be narrow enough to support measurement and broad enough to include meaningful substitutes. For example, a firm selling bicycle-sharing services should not automatically define its market as “bicycle-sharing.” For many customers, buses, taxis, walking and private bicycles may be substitutes. The relevant market may instead be “short-distance urban journeys under five kilometres,” subject to evidence about substitution.

Market boundaries and the customer job

The jobs-to-be-done perspective asks what customers are trying to accomplish rather than what they currently purchase. This prevents incumbents from defining the market too narrowly. However, market expansion must remain empirically defensible. A theoretical substitute is not necessarily an effective substitute if it differs greatly in price, accessibility, performance or regulation.

A useful market statement has the form:

In [geography] during [period], [customer segment] purchases or could purchase [solution category] to accomplish [job], through [channels], measured in [units or revenue].

The statement should be recorded before collecting market-size figures. Otherwise, analysts may unconsciously adjust the definition to fit available data.

2. TAM, SAM and SOM

The three-level framework separates theoretical opportunity from realistically reachable opportunity.

TAM is not a sales forecast. It is a ceiling under stated assumptions. SAM incorporates restrictions such as language, regulations, distribution and product compatibility. SOM incorporates competition, capacity, financing, awareness, sales productivity and execution.

A simple relationship is:

This relationship is conceptual rather than a guarantee that all estimates use identical definitions. A common error is to calculate TAM in global revenue, SAM in potential users and SOM in monthly orders. The measures must use compatible units and periods.

Revenue and unit measures

Market size can be stated as customers, units, transactions or revenue. For a homogeneous product:

The variables must refer to the same period and segment. If buyers differ substantially in frequency or price, segment-level calculations are preferable:

Here, i indexes customer segments; Nᵢ is the number of buyers in segment i; fᵢ is purchases per period; qᵢ is units per purchase; and pᵢ is average price per unit. The formula assumes that the variables are measured consistently and that double counting is avoided.

3. Market-Sizing Methods

3.1 Top-down sizing

Top-down analysis begins with an external market estimate and filters it through relevant proportions. For example:

Top-down methods are efficient when reputable industry, government or trade data exist. They are useful for strategic context and investor communication. Their weakness is that the original category may not match the venture’s definition, and successive percentages can conceal weak assumptions.

Illustrative example. Suppose a national digital-learning market is valued at $800 million. A venture targets adult learners, estimated at 30% of spending; its region represents 20% of adult-learning spending; and its product addresses 25% of regional spending.

This is a scenario estimate, not a verified fact. The analyst should document the sources and test alternative percentages.

3.2 Bottom-up sizing

Bottom-up sizing builds the estimate from observable operating quantities:

For a subscription service, the calculation may use eligible accounts, expected adoption and annual subscription price. For a marketplace, gross transaction value should be distinguished from the platform’s revenue. If the platform takes a commission:

Bottom-up analysis is often more actionable because it connects directly to sales capacity and operating plans. Its principal risks are overestimating the addressable customer count, confusing leads with paying customers and assuming that observed behaviour will immediately transfer to a new product.

Illustrative example. A venture identifies 40,000 eligible small businesses. It estimates that 12% will subscribe within the relevant year, at an annual price of $240.

The result is a forecast under stated adoption assumptions, not the entire TAM.

3.3 Value-theory sizing

Value theory estimates willingness to pay from the economic value created for customers. It is particularly useful for novel products for which historical category data are weak. A buyer’s economic value may be represented as:

Reference value is the value of the next-best alternative. Differentiation value is the measurable improvement or saving created by the new offer. A firm cannot normally capture all created value: price is constrained by customer willingness to pay, alternatives and bargaining power.

Example. A scheduling tool saves a business 10 labour hours per month. If the relevant labour cost is $22 per hour, monthly time value is $220. It also reduces avoidable errors valued at $50 per month. The maximum economic value, before considering switching costs and uncertainty, is:

A price of $100 may be attractive if the evidence supports the savings, but the calculation does not prove that every customer will pay $100. Value claims require customer research, experiments or observed outcomes.

Comparing the methods

Strong analyses triangulate at least two methods. Convergence increases confidence; divergence identifies assumptions requiring research.

4. Demand Drivers and Adoption

Demand is influenced by economic, demographic, technological, social, regulatory and competitive factors. Important drivers include:

• disposable income and business budgets;
• population, household or firm growth;
• prices and customer switching costs;
• complementary products, infrastructure and skills;
• awareness, trust and perceived risk;
• regulation, taxation and public procurement;
• product performance and availability;
• competitor actions and substitute prices.

Diffusion and adoption

Adoption is not the same as awareness. A customer may know about a product but lack willingness, ability or occasion to purchase. Analysts should distinguish:

1. Eligible population: can legally and practically use the product.
2. Aware population: knows the product or category exists.
3. Interested population: expresses a relevant need.
4. Trial users: have used the product at least once.
5. Repeat users: continue purchasing.
6. Paying customers: generate revenue under the defined accounting rule.

A simple adoption forecast is:

The adoption rate should be defined precisely, and t denotes the period. For a mature category, adoption may be estimated from historical penetration. For a new category, comparable markets, experiments and customer interviews are more appropriate than unsupported “hockey-stick” curves.

A logistic curve can represent adoption constrained by a maximum market:

Here, A(t) is cumulative adoption at time t, K is the carrying capacity or maximum adoption, r is the growth-rate parameter, t₀ is the midpoint, and e is the base of natural logarithms. The model assumes a relatively smooth S-shaped process and is unsuitable when adoption is driven by abrupt regulation, supply shortages or discontinuous technology.

5. Seasonality and Time Patterns

Many markets vary systematically by month, week or day. Examples include tourism, education, clothing, agriculture and heating. Seasonality must not be confused with long-term growth.

A seasonal index for period j can be calculated as:

An index above 1 indicates above-average demand; an index below 1 indicates below-average demand. If the overall monthly average is 1,000 units and the average in December is 1,400 units, the December index is 1.4. A forecast based on a deseasonalized trend can then be multiplied by the relevant index, provided the seasonal pattern is reasonably stable.

Analysts should investigate changing seasonality. Climate change, school calendars, promotions, platform algorithms and income timing can alter historical patterns. Seasonal indices based on too few observations may mistake random fluctuations for regular demand.

6. Forecasting Approaches

Naïve and growth-rate forecasts

A naïve forecast uses the latest observed value as the next forecast:

where Fₜ₊₁ is the forecast for the next period and Yₜ is the latest observation. It is a useful benchmark, not necessarily a sophisticated method.

A constant-growth forecast is:

where Y₀ is the base demand, g is the assumed growth rate per period, and t is the number of periods. The formula assumes a constant compounded growth rate. It should not be used mechanically when market capacity, prices or conditions change.

Moving averages

A simple moving average of k observations is:

It smooths random variation but responds slowly to turning points. A weighted moving average assigns weights wᵢ that sum to 1:

The weights must be non-negative when the method is intended as a weighted average.

Example. Monthly orders were 90, 100 and 110. A three-month moving-average forecast is:

Causal forecasting

Causal models relate demand to explanatory variables such as price, income, advertising, population or temperature. A simple linear model is:

where Q is demand, P is price, I is income, A is advertising or another driver, a is an intercept, b, c and d are estimated coefficients, and ε represents unobserved influences and measurement error. The model requires appropriate data, careful interpretation and attention to correlation versus causation. For example, advertising may increase when managers expect high demand, creating reverse causality.

Forecast evaluation

Forecasts should be compared with actual outcomes. Mean absolute error is:

where n is the number of forecast periods, Yₜ is actual demand and Fₜ is forecast demand. MAE is expressed in the same units as demand. Mean absolute percentage error is:

MAPE is problematic when actual demand is zero or close to zero, and percentage errors can be misleading when volumes differ greatly. Forecast accuracy should therefore be assessed with operational consequences in mind, not a single metric alone.

7. Scenarios, Sensitivity and Uncertainty

A point estimate hides uncertainty. Scenario analysis describes internally consistent combinations of assumptions. A useful set is:

• Downside: slower adoption, lower conversion, price pressure or supply constraints.
• Base: most defensible assumptions from current evidence.
• Upside: favourable but plausible conditions.

Scenarios should not be labelled merely “low, medium and high” without explaining what changes. For example, an upside scenario may combine faster awareness growth, stronger retention and improved distribution; combining every optimistic assumption may produce an implausible result.

Sensitivity analysis changes one assumption while holding others constant. If annual revenue is:

where N is target customers, c is conversion rate and p is annual price, the analyst can examine how revenue responds to each variable. A 10% relative increase in conversion increases revenue by 10% if all other variables remain fixed. In practice, variables may interact: a higher price could reduce conversion, so a multi-variable scenario is also necessary.

Uncertainty has several sources:

• Measurement uncertainty: imperfect or inconsistent data.
• Sampling uncertainty: estimates based on a sample rather than the whole population.
• Model uncertainty: an inappropriate forecasting structure.
• Parameter uncertainty: uncertain values such as conversion or growth.
• Structural uncertainty: unexpected changes in regulation, technology or competition.

Forecasts should report assumptions, ranges, dates, sources and confidence limitations. False precision—such as reporting a market as $12,483,271 without reliable evidence—reduces credibility.

8. Applied Case: GreenRoute Shared E-Bikes

This is an illustrative fictional case.

Background

GreenRoute plans a shared e-bike service for three university districts in a city. The proposed service targets students and staff making journeys of two to six kilometres. Revenue would come from pay-per-ride fees and monthly passes.

Problem

The founders claim that the national urban-mobility market is the TAM and that capturing 1% would generate substantial revenue. The research team must develop a defensible local estimate and a first-year forecast.

Analysis and market definition

The relevant market is defined as trips by university students and staff within the three districts, during the first operating year, for which e-bikes are an acceptable substitute. The team excludes long-distance commuting, users without smartphone access and trips prohibited by local regulations. It also investigates buses, walking, private bicycles and ride-hailing as substitutes.

The team identifies 30,000 eligible people. Survey and pilot evidence suggest that 40% make at least one suitable trip weekly, and 25% of those are willing to consider e-bikes. Thus the potentially interested population is:

Suppose the expected first-year paying adoption is 20% of this group, with an average of 18 rides per user per year and net revenue of $2.50 per ride:

This is a bottom-up obtainable-demand estimate, not the whole market. The team separately estimates SAM by considering all 3,000 interested potential users and a theoretical annual frequency of 30 rides:

A top-down estimate based on city transport spending produces $350,000, but the category includes car parking and bus fares that GreenRoute cannot serve. The divergence indicates category mismatch rather than proof that the bottom-up figure is wrong.

Scenarios

The corresponding annual revenues are approximately $9,274, $27,000 and $62,899. The range is wide because adoption and usage compound. The founders should therefore delay fleet expansion until pilot data validate repeat usage, not merely registrations.

Approach and solution

GreenRoute conducts a four-week pilot, records rides rather than app downloads, monitors usage by district and time, and tests price messages. It uses a conservative capacity plan based on peak demand, repairs and weather. The company also obtains municipal permissions, communicates data practices clearly and establishes safety procedures.

Outcome and learning points

The case demonstrates that market definition precedes arithmetic; local bottom-up evidence may be more useful than a large national category; and obtainable demand depends on behaviour, capacity and execution. A professional recommendation would present the assumptions, scenario range and validation plan rather than a single impressive number.

9. Technology, AI and Governance

Digital platforms generate detailed demand data, including searches, transactions, location patterns and retention. Forecasting software can automate data cleaning, seasonality decomposition and model comparison. Machine-learning models may improve predictive performance when large, relevant and stable datasets exist. However, complexity does not guarantee accuracy or causal insight.

Entrepreneurs using AI should:

• document data sources, definitions and transformations;
• test forecasts on periods not used for model fitting;
• check for bias across customer groups and locations;
• protect personal and location data through minimisation and access controls;
• distinguish prediction from explanation;
• retain human accountability for consequential decisions;
• disclose material uncertainty to investors and partners.

Privacy, consumer-protection, competition and sector-specific regulations may apply. Demand data should not be collected or used beyond legitimate purposes. Manipulative personalisation, discriminatory pricing and fabricated market evidence are ethical failures even when they improve short-term conversion.

10. Professional Best Practices and Limitations

A defensible market study should:

1. Freeze and state the market definition.
2. Use multiple independent sources where feasible.
3. Separate observed facts, estimates and assumptions.
4. Reconcile units, prices, time periods and customer counts.
5. Segment customers when behaviour or economics differ.
6. Triangulate top-down, bottom-up and value-based estimates.
7. Validate willingness to pay and repeat behaviour, not just interest.
8. Present scenarios and key sensitivities.
9. Benchmark forecast accuracy and revise assumptions.
10. Record the research date because markets change.

Limitations remain. Surveys may overstate intended behaviour; industry reports may use commercial definitions; historical data may not represent a new category; and competitive responses can invalidate a forecast. Market sizing is therefore a decision aid, not a guarantee of sales.

Chapter Summary

Market definition establishes the customer problem, product category, geography, channel, time period and measurement unit. TAM expresses the broad relevant opportunity, SAM applies practical service constraints, and SOM estimates realistically obtainable demand. Top-down methods provide broad context, bottom-up methods connect estimates to operating realities, and value theory links willingness to pay to customer benefits. Reliable analysis triangulates these methods rather than treating any one as authoritative.

Demand forecasting requires attention to adoption stages, economic and behavioural drivers, substitutes, seasonality and capacity. Growth formulas, moving averages, causal models and adoption curves are useful only when their assumptions fit the situation. Scenarios and sensitivity analysis reveal how outcomes change when uncertain assumptions change. Forecast quality depends not only on statistical accuracy but also on data definitions, ethical governance, transparent communication and managerial usefulness.

Key Terms

Market definition: Explicit boundaries describing customers, needs, products, geography, channels and time.

Relevant market: The set of buyers and suppliers considered for a particular decision.

TAM: Total addressable market under broad stated assumptions.

SAM: Serviceable available market that fits the venture’s practical scope.

SOM: Serviceable obtainable market that the venture can realistically capture.

Top-down sizing: Estimating a market by filtering a broad external estimate.

Bottom-up sizing: Building an estimate from customers, transactions, usage and prices.

Value theory: Estimating opportunity from the economic value created for customers.

Substitute: An alternative solution that may satisfy a similar customer need.

Adoption rate: The proportion of an eligible population using or purchasing a product under a defined rule.

Penetration: The share of a defined population that has adopted a product or category.

Seasonality: Systematic variation associated with recurring calendar periods.

Forecast: A structured estimate of future demand or another variable.

Scenario analysis: Evaluation of outcomes under internally consistent assumptions.

Sensitivity analysis: Testing how an outcome changes when an assumption changes.

MAE: Mean absolute error, the average absolute difference between actual and forecast values.

Carrying capacity: The maximum adoption level represented by a diffusion model.

Willingness to pay: The maximum price a customer is prepared to pay under stated conditions.

Triangulation: Combining independent methods or sources to improve confidence.

Forecast uncertainty: The range of possible outcomes arising from data, model and environmental limitations.

Review Questions

Short Answer Questions

1. What distinguishes SAM from SOM?
2. Why can a national industry report be unsuitable for a local venture’s TAM?
3. Define seasonality and give one example.
4. What is the principal advantage of bottom-up market sizing?
5. Why should awareness and paying adoption be measured separately?

Long Answer Questions

1. Explain how customer jobs and substitutes influence the definition of a relevant market.
2. Compare top-down, bottom-up and value-theory market sizing, including a strength and limitation of each.
3. Explain how a venture can combine adoption analysis, seasonality and capacity constraints in a demand forecast.
4. Discuss the difference between forecast accuracy and forecast usefulness in entrepreneurial decision-making.
5. Explain the ethical and governance responsibilities associated with AI-enabled demand forecasting.

Analytical / Application Questions

1. A market contains 18,000 eligible customers. Thirty percent are expected to adopt, each purchasing four times per year at $15 per purchase. Estimate annual revenue and state two assumptions that should be tested.
2. A founder reports a global TAM of $2 billion but serves only one region, one customer segment and one regulated channel. Design a procedure for converting the claim into a defensible SAM and SOM.
3. Monthly demand for the last three months was 240, 270 and 300 units. Calculate a three-month moving-average forecast for the next month, then explain one reason why the forecast may understate demand if the trend continues.

Multiple Choice Questions with Answers

Question 1

Which statement best describes SOM?

A. The total global value of every possible substitute

B. The portion of the market the venture can realistically obtain

C. The number of people who have heard of the product

D. The value created for customers before pricing

Correct Answer: B. The portion of the market the venture can realistically obtain

Question 2

A market has 40,000 eligible businesses, a 12% expected adoption rate and a $240 annual subscription. What is the estimated annual revenue?

A. $96,000

B. $480,000

C. $1,152,000

D. $9,600,000

Correct Answer: C. $1,152,000

Question 3

Which method begins with a broad industry estimate and applies relevant filters?

A. Bottom-up sizing

B. Value-theory sizing

C. Top-down sizing

D. Experimental pricing

Correct Answer: C. Top-down sizing

Question 4

In the formula Market revenue = buyers × frequency × units per purchase × price, what must be consistent across variables?

A. Only the currency symbol

B. The time period and segment definition

C. The competitor’s advertising budget

D. The survey sample size

Correct Answer: B. The time period and segment definition

Question 5

A three-month moving average is calculated from demand of 90, 100 and 110 units. What is the forecast?

A. 90 units

B. 95 units

C. 100 units

D. 110 units

Correct Answer: C. 100 units

Question 6

What does a seasonal index of 1.4 indicate?

A. Demand is 40% below average in that period

B. Demand is 1.4 units in every period

C. Demand is approximately 40% above the overall average in that period

D. The forecast has a 1.4% error

Correct Answer: C. Demand is approximately 40% above the overall average in that period

Question 7

Which is a major limitation of MAPE?

A. It cannot compare any forecasts

B. It is unsuitable when actual demand is zero or near zero

C. It is always measured in dollars

D. It ignores forecast errors entirely

Correct Answer: B. It is unsuitable when actual demand is zero or near zero

Question 8

Value-theory sizing primarily estimates opportunity from:

A. The number of competitors alone

B. The economic value created for customers

C. The largest available industry report

D. The venture’s staffing level

Correct Answer: B. The economic value created for customers

Question 9

What is the main purpose of sensitivity analysis?

A. To change all assumptions to optimistic values

B. To test how an outcome responds to changes in an assumption

C. To eliminate all uncertainty from a forecast

D. To replace customer research

Correct Answer: B. To test how an outcome responds to changes in an assumption

Question 10

Which practice improves governance of AI-enabled demand forecasting?

A. Presenting unverified predictions as facts

B. Ignoring performance differences across customer groups

C. Documenting data sources and testing forecasts on unused periods

D. Collecting all available personal data without limitation

Correct Answer: C. Documenting data sources and testing forecasts on unused periods