Skip to main content
Weighting is appropriate when the achieved sample differs from the population the study describes and a reliable target distribution is available. It cannot correct coverage, measurement, or questionnaire problems by itself.

​
Before you start

Bring the achieved sample distribution, a relevant population definition, and the proposed benchmark sources.

​
Evaluate the need

Compare the achieved sample with the intended population on variables that are plausibly related to the outcome. Consider weighting when the difference is material to the analysis and the benchmark is relevant to the population and time period. An unweighted result may be preferable when:
  • No reliable target distribution exists.
  • The study is exploratory and relies mainly on themes or individual responses.
  • The base is too small to support stable reweighting.
  • Weighting would require narrow cells with few or no respondents.
  • The selected variables are not related to the measures being interpreted.

​
Select target sources

Potential sources include official population statistics, an appropriate syndicated dataset, a verified customer database, or a documented study-specific benchmark. Record the source, field dates, population definition, variable mapping, and any category grouping. Use broad, interpretable categories where possible. Every additional variable or nested target creates more constraints and can reduce effective sample size.

​
Review the effect

Create the scheme and compare weighted with unweighted results before selecting it for a report. Review:
  1. Weighted N and weighting efficiency.
  2. Minimum and maximum respondent weights.
  3. Missing cells and excluded respondents.
  4. Whether the scheme converged.
  5. Measures whose conclusion changes after weighting.
If one conclusion changes materially, identify which sample cells drive that change and whether their bases support the interpretation.

​
Report the method

Document:
  • The variables and target distributions.
  • The source of each target.
  • The method used, including nesting where applicable.
  • The unweighted and weighted bases.
  • The weight range and efficiency.
  • Any exclusions caused by missing weighting variables.
  • Whether reported conclusions differ from the unweighted read.
Significance tests in crosstabs use unweighted respondent bases even when the table displays weighted values. Carry the raw base with statistical claims.

​
Apply weighting before report production

Choose and review the scheme in the analysis plan before generating outputs. Duplicate a locked scheme to revise its targets, then select the copy in the plan. Regenerating with different weighting updates matching charts without archiving analysis conversations. Recheck saved slides and figures before delivery.

​
Check the result

The target sources and category mappings are documented, diagnostics are acceptable for the intended use, and material changes in conclusions have been explained before report generation.