This guide explains how to run maxdiff by connecting the business decision, evidence design, sample, quality controls, analysis and action rules.
Seven-step workflow
Write the decision owner, alternatives, timing and consequence of a wrong answer.
Separate what must be estimated, understood, compared, predicted or tested.
Review internal data, prior research, market sources and assumptions before commissioning new fieldwork.
Select methods because they distinguish the competing explanations, not because they are fashionable.
Define population, recruitment, field checks, exclusions, missing-data rules and audit trail.
State how evidence will be combined and what each plausible result will trigger.
Lead with finding, implication, recommendation, uncertainty and next action; keep detail behind the executive answer.
Design choices that should be explicit
Who or what the result is intended to describe, compare or predict.
The respondent, household, account, outlet, event, transaction or market being analysed.
How precise, stable or convincing the result must be to change the decision.
The action attached to each credible range of outcomes.
Quality gates
Pause the work if recruitment cannot verify the intended population, if the instrument does not map to the decision, if fieldwork quality cannot be audited, or if the analysis will be unable to distinguish the main competing explanations. Speed is useful only while the evidence remains decision-safe.
What the final output should contain
- A one-page executive answer.
- Decision context and assumptions.
- Method, population, sample and quality summary.
- Findings with effect size, uncertainty and counter-evidence.
- Implications, recommendation, owners and next decision gates.
Common mistakes
- Beginning with a questionnaire or discussion guide before the decision is clear.
- Collecting every interesting variable and creating a long but unfocused instrument.
- Hiding uncertainty because the audience asked for certainty.
- Ending with observations instead of implications and actions.
Practical checklist
This reference is written by Ninth Atlas as a decision-oriented explainer. It separates definition, design, analysis and limitations so a method is not mistaken for an answer. Final study specifications should be reviewed against the actual population, evidence, risk and regulatory context.