How to define the population for a discrete-choice experiment

Edited and reviewed by Avi Yashchin, CEO and Co-Founder · August 13, 2026 · Updated October 2, 2026

On this page

The decision this step makes

Once a research or insights lead has a question, define who the study is intended to represent. Conjoint experiments estimate preferences from the choices presented to respondents (Sawtooth Software). A result from one buyer group does not automatically describe another: eligibility, coverage, recruitment or generation, and weighting all affect the interpretation.

Define the population and sample separately

The target population is the group the decision concerns. The sample is the set of respondents or generated profiles actually studied. Segment analysis can reveal different preferences in a conjoint model (Qualtrics technical overview), but neither a large sample nor a detailed profile proves that the sample covers the target population.

What traits define the population?

Define eligibility first: buying authority, product use, geography, and decision period where relevant. Add traits needed for planned subgroup comparisons or population matching. The number of traits does not determine validity; one decisive eligibility rule may matter more than a long demographic description.

Traits should follow from the decision, not from a list of available fields. Three questions narrow the choice:

Who actually makes or influences this decision? If the study is about a purchase, the relevant trait might be role or buying authority, not just demographics.
Which traits need separate analysis or population matching? Use prior evidence or a stated hypothesis to nominate possible preference differences. Include eligibility variables even when they are not expected to change the answer.
What evidence describes this population? Check a dated customer or account roster, recruitment frame, or other relevant source for coverage and joint trait distributions. A domain expert can check plausibility, but recognition alone cannot establish representativeness.

Reviewing traits before committing the design

For example, a hypothetical renewal study might target US small-business accounts renewing in a specified quarter and the people authorized to choose a plan. Use the current account roster to identify eligible accounts, remove duplicates, and check which buyer roles are missing. If account-size groups are sampled unequally, derive target shares and weights from verified roster counts; do not invent market proportions from the personas.

Define eligibility, verify the population source and coverage, plan subgroup sampling and weights, then review the realized sample. — Population coverage needs evidence beyond a trait list.

What this step doesn't decide

In a generated study, traits guide how profiles are constructed; they do not establish that the resulting profiles reproduce the population's behavior or joint characteristics. In a human study, the same eligibility brief can guide recruitment, but the realized sample still needs coverage and quality checks. Align the population and instrument when comparing generated and human results, and record remaining differences.

Generated profiles require distribution and response validation; recruited respondents require eligibility, coverage, and response-quality checks. — A shared population brief does not make generated and recruited samples equivalent.

Where to go from here

Before scoping a study with Subconscious, bring the eligibility rules, dated population source, subgroup questions, and any missing coverage. Agree on the profile or recruitment approach, weighting assumptions, and human comparison needed for the decision. The public research record provides aggregate evidence about the method.