What our data does — and doesn't — represent.
A poll creator publishes a question with a set of options and shares a link, QR code, or embeddable widget. Anyone who receives that link can open it and vote — there's no pre-screened panel, no random sampling of the population, and no guarantee that respondents reflect Kenya's actual demographic makeup. This is a strength for speed and reach, and a real limitation for statistical representativeness. We say so plainly rather than let a result imply more precision than it has.
Depending on how a poll creator configures it, a poll can enforce any combination of:
No system fully eliminates determined abuse — a person with multiple phone numbers and devices can still work around per-device/per-phone limits. What we can honestly claim is that casual, single-click ballot-stuffing is meaningfully harder here than on a poll with no verification at all.
When a results page shows a candidate or option "leading," that means they currently have the most votes among people who chose to participate in that specific poll — not that they are leading in Kenya as a whole, and not a projection of what would happen in an actual election. Two different polls on the same topic, shared through different channels to different audiences, can legitimately show different results. That's expected, not a sign either one is broken.
Where a voter optionally shares their county, gender, or age group at the time of voting, we can show a breakdown by those fields. This reflects the self-reported, opt-in characteristics of that poll's actual respondents — again, not a weighted estimate of the true population in that county or age group.
Our Rankings pages aggregate standing across polls a candidate appears in, and record periodic snapshots so we can show genuine historical movement over time rather than a number computed fresh (and potentially inconsistently) on every page load. The same self-selection caveat above applies to every underlying data point that feeds a ranking.
Reports published through our Report Library include a written summary from our team alongside the underlying data and charts. The summary reflects our team's reading of that specific dataset — it is not an independently peer-reviewed academic publication.