Methodology & Data Transparency

Credible research depends on openness about how it was done. For every published poll and report we disclose the information recommended by the American Association for Public Opinion Research (AAPOR) Transparency Initiative.

Two kinds of polls

Open online polls are open to anyone. They are a quick, engaging way to gauge sentiment among people who choose to take part, but they are not statistically representative of all Kenyans and we never attach a margin of error to them.

Scientific surveys use probability or quota sampling designed to represent a defined population – for example, registered voters in a county. Interviews may be conducted face-to-face (CAPI), by telephone (CATI) or online with a recruited panel. Results are weighted to known population characteristics and reported with a margin of error and confidence level.

What we disclose

• Who sponsored and who conducted the research
• The population surveyed and how respondents were selected
• Sample size and fieldwork dates
• Mode(s) of data collection and languages used
• Weighting procedures and the variables used
• Margin of error and design effect (probability samples only)
• Exact question wording and order
• Known limitations

Quality assurance

We screen for duplicate and fraudulent responses (device, browser and network checks; bot detection; proof-of-work challenges), monitor interview durations, verify field interviews with GPS where respondents consent, back-check a share of interviews, and review flagged responses before results are finalised.

Weighting

Where appropriate we apply raking (iterative proportional fitting) so that the sample matches population benchmarks for gender, age group and region from the Kenya Population and Housing Census and IEBC voter register. We report both weighted and unweighted results internally and publish the effective sample size.