Market sizing studies often need more than completes-only reporting. If a respondent is screened out because they do not qualify for a market, that response is still useful evidence. It tells you something about incidence, eligibility, or the size of the reachable audience.
The challenge is deciding which partial completes belong in each estimate. Including every partial can overstate the base for later questions. Including only completes can understate the screening population. The right approach is to anchor each estimate to the question that defines the market frame.
When partial completes matter
Use partial completes when a question is part of the sizing logic, not just part of the final analysis.
For example, imagine a study that asks:
- Age
- Gender
- Category usage
- Recent purchase
- Brand consideration
- Purchase intent
Respondents may terminate at Age, Gender, Category usage, or Recent purchase. For a concept test, you may only want people who completed the full study. For market sizing, the terminations are part of the story:
- People who answer Age but terminate before Gender can still contribute to an age-based market size.
- People who answer Category usage but fail Recent purchase can still contribute to category incidence.
- People who never reach Category usage should not be counted in a category-usage denominator.
That is the core rule: include partials only when they reached and answered the question that defines the base for the estimate.
Worked example
Suppose 1000 respondents enter a survey.
| Stage | Respondents who reached stage | Respondents screened out at stage | Respondents continuing |
|---|---|---|---|
| Age | 1000 | 100 | 900 |
| Gender | 900 | 150 | 750 |
| Category usage | 750 | 300 | 450 |
| Recent purchase | 450 | 200 | 250 |
| Full survey complete | 250 | 0 | 250 |
If you report only on completes, the base is 250. That is correct for questions asked after the full qualification path, but it is too narrow for estimating category incidence in the recruited population.
If you include every partial, the base is 1000. That is correct for age-level source composition, but it is too broad for recent purchase, because many respondents never reached that question.
Instead, choose the base question for the estimate:
| Estimate | Recommended base |
|---|---|
| Age distribution of incoming respondents | Age |
| Gender distribution among respondents who reached Gender | Gender |
| Category incidence by Age | Category usage, crossed by Age |
| Recent-purchase incidence by Age | Recent purchase, crossed by Age |
| Brand consideration among qualified buyers | Completes or the first post-qualification question |
Using cross-tabs to include the right partials
A cross-tab naturally limits each cell to respondents who answered both the row question and the column question.
For example, if you cross Category usage by Age, the age columns include respondents who:
- answered Category usage, and
- answered Age.
Respondents who answered Age but never reached Category usage are not included in the Category usage counts. Respondents who answered Category usage but do not have an Age response do not appear in an Age column.
This is useful for market sizing because it ties the estimate to a concrete question pair. If Age is the balancing variable and Category usage is the incidence question, the cross-tab gives you incidence by Age among people who reached the incidence question.
Using denominator questions
Some report views also let you choose a percentage base. The most useful setting for this workflow is a question-based denominator: calculate percentages using respondents who answered a specific base question.
Use this when the displayed percentage should answer:
Out of everyone who answered this sizing question, what share selected this response?
For example:
- Use Age as the base when you are sizing against the age-balanced intake frame.
- Use Category usage as the base when you are measuring share among category respondents.
- Use Recent purchase as the base when you are measuring outcomes among recent purchasers.
This changes the denominator used for percentages. It does not rewrite the entire report universe. Counts still come from the row and column questions in the table.
Common patterns
Age-balanced market sizing
Use this when fieldwork is balanced or click-balanced on Age, and the client wants age-cell incidence.
- Include the relevant respondent statuses for analysis, usually completes plus the terminated or screened-out statuses that represent valid partials.
- Build a cross-tab with the incidence question as the row and Age as the column.
- Use the incidence question or Age as the denominator, depending on the exact estimate:
- Age denominator: share of each age cell selecting an incidence response.
- Incidence-question denominator: distribution among respondents who reached the incidence question.
- Read the weighted percentages and effective sample sizes before making population claims.
Screening-funnel sizing
Use this when each screener narrows the market.
- Treat each screener as its own denominator.
- Report the pass rate at each stage.
- Multiply stage rates only when each rate is conditional on the previous stage.
For example:
| Step | Rate |
|---|---|
| Category user among Age respondents | 60% |
| Recent purchaser among category users | 40% |
| Brand considerer among recent purchasers | 50% |
The estimated brand-considerer share of the age-qualified population is:
So the estimate is 12% of the age-qualified population, before applying any external population total.
Completes-only outcome analysis
Use this when the question is asked only after qualification, or when the analysis is about opinions among fully qualified respondents.
In that case, use completes only. Partial completes should inform the incidence funnel, not the post-qualification attitudinal readout.
What not to do
Do not include all partial completes by default. A respondent who terminates at Gender has no information about Category usage or Recent purchase.
Do not use completes-only bases for incidence unless the incidence question is asked only of completes. Completes-only reporting can hide screenout rates and make niche markets look larger or smaller than they are.
Do not mix unconditional and conditional rates in the same calculation. If one percentage is "among all Age respondents" and another is "among category users", label that clearly before multiplying or comparing them.
Quality checks before sharing results
Before publishing a market size from partial completes, check:
- Question reach: Did every respondent in the denominator have a real chance to answer the sizing question?
- Status inclusion: Are the included terminated or screened-out respondents valid research observations rather than fraud, duplicates, or technical failures?
- Weighting: Are weights calibrated to the population frame you want to size?
- Effective sample size: Are small weighted cells being carried by too few respondents?
- Field report evidence: Do termination points match the intended screener logic?
The field report is the best place to validate the termination pattern. See Field reports for how to inspect terminations by question, quota, and device.
Recommended workflow
- Identify the market-sizing question.
- Decide which respondent statuses are valid for that estimate.
- Cross the sizing question by the balancing or reporting variable, such as Age.
- Use a question-based denominator when the base should be tied to a specific screener or sizing question.
- Check weighted percentages, raw base, and effective sample size.
- Document the denominator in the chart title, table note, or methodology.
A good table label is explicit:
Category usage by Age, among respondents who reached Category usage.
That is much clearer than:
Category usage by Age.
The first label tells readers exactly how partial completes were used, and it prevents the most common mistake in market sizing: treating every partial complete as if it answered every screener.