Responses are collected from adult smartphone users in the target country via our mobile-first sampling network. Respondents are intercepted across a rotating set of partner mobile applications, weighted toward broad demographic reach rather than a single platform. Each respondent supplies age band, gender, and occupation alongside their answer.
Questions are written in the local language by native speakers, then translated to English for the published dataset. Both the original and English versions are shown on every poll page. Vote-intent questions follow standard ballot-style wording with candidate names and party affiliations included.
Margin of error is calculated at 95% confidence assuming worst-case variance (p = 0.5). For a sample of n = 25,000, this is approximately ±0.62%; for n = 1,500, approximately ±2.5%. The exact MoE for each poll is shown on its detail page.
Data is published unweighted by default. Raw demographic crosstabs (age, gender, occupation) are included so buyers can apply their own weighting frameworks. Where census-based benchmarks are available for a country, we provide notes on observed sample skew on the relevant poll page.
If you have a methodology question on a specific poll, post it on our feedback board.