Market reports describe their methods using two broad labels. Primary research means collecting new information directly, usually by talking to people in the industry. Secondary research means using information someone else has already collected and published. Almost every report uses both, and the way they are combined determines how much the final numbers can be trusted.
The two kinds of research compared
| Primary research | Secondary research | |
|---|---|---|
| What it is | New data gathered for this study | Existing data gathered by others |
| Typical forms | Interviews with suppliers, distributors and buyers; surveys; site visits | Official statistics, trade data, company filings, association data, academic work, earlier reports |
| Strengths | Current, specific to the question, can capture pricing and intentions | Cheaper, often larger samples, sometimes audited or official |
| Weaknesses | Small samples, respondent bias, expensive | Definitions may not match, can be out of date, may itself be estimated |
| What to ask | How many interviews, with whom, and when? | Which sources, which years, and how were definitions aligned? |
How estimates are typically built
Publishers rarely measure a market directly, because there is rarely a single source that records every sale. Instead they build estimates from pieces. Two approaches dominate.
Top-down
Start from a large, known quantity and narrow it down. For example, take total spending in a broad category from official statistics, then apply estimated shares to reach the specific product. Top-down estimates are quick and tie the result to a credible total, but every share applied is an assumption, and errors multiply.
Bottom-up
Start from the parts and add them up. That might mean estimating each major supplier's sales in the category, or estimating the number of buyers and multiplying by how much each spends.
Market value = Units sold × Average selling price
or, on the demand side: Number of buyers × Adoption rate × Annual spend per adopter.
Hypothetical illustration: if an imaginary market has 40,000 potential business buyers, 25% have adopted a product, and each adopter spends 1,500 a year, the bottom-up estimate is 40,000 × 0.25 × 1,500 = 15,000,000 per year. Change the adoption rate to 20% and the estimate falls to 12,000,000. Small changes in one assumption move the whole result.
Triangulation
Careful researchers build the estimate more than one way and compare. If a top-down and a bottom-up estimate land close together, confidence rises. If they are far apart, the researcher has to find out why before publishing. Many reports claim to triangulate; fewer explain what the separate estimates were or how gaps were resolved.
From a size to a forecast
- Fix the base year. Establish the best estimate for the most recent year with reasonable data.
- Identify drivers. Decide which factors move the market, such as prices, the number of buyers or replacement cycles.
- Project the drivers. Forecast each driver, often drawing on interviews for near-term expectations and on economic projections for the longer term.
- Combine and test. Run the model, compare with history and with other sources, and adjust.
- Present ranges or scenarios. Better reports show how results change under different assumptions.
Where each approach tends to go wrong
Primary research is vulnerable to who gets interviewed. Suppliers may talk up their own segment, and a handful of conversations can be dominated by one region or customer type. Secondary research is vulnerable to definitions: an official statistic may group the product with related goods, or measure it at a different point in the value chain. A methodology section that names both the interview mix and the main secondary sources lets you judge which of these risks applies. One that names neither leaves you guessing.
What this means for readers
A market size is a modelled estimate, not a measurement. The useful questions are about the inputs: which secondary sources anchor the model, how many primary interviews tested it, and which assumptions matter most. If a report will not say, its numbers are hard to judge. Our red-flags checklist covers the warning signs of reports that skip these steps, and our segmentation explainer covers how the same methods produce the splits within the total.
Unfamiliar term? See the glossary.