Start with the population
Every poll starts by defining the people its results claim to describe. For a national poll, that might be adults aged 18 or over resident in Kenya. A survey of registered voters, customers of one business, or residents of one county answers a different question. Before looking at a percentage, check who is included and who is outside its scope.
Find a way to reach them
A sampling frame is the list or process used to reach that population. It could be a list of households, a telephone frame, or a mapped set of areas. No frame is perfect. People without access to a particular phone network or address list may be harder to include. A useful methodology statement names the frame and explains its limitations.
Give each part of the population a place
Stratification divides the population into groups before selection, often by county or region and by urban and rural areas. It helps make sure that coverage is planned rather than left to chance. Within each stratum, selection should be random or systematic according to a documented procedure. Interviewers should not replace that procedure with a choice of friendly faces.
Control the fieldwork
Trained interviewers use scripted introductions and consistent questionnaires. Callbacks help reach people who were unavailable at first contact. GPS records or call records can help supervisors check that interviews followed the plan, while back-checks revisit a share of interviews. These controls support the method; they do not replace careful sampling.
Report what actually happened
The target sample is the number planned. The achieved sample is the number of interviews actually completed and accepted after checks. Honest reporting distinguishes the two and explains meaningful gaps. Readers need to know both the intended design and the evidence that was ultimately collected.
When you read any poll, ours included, ask first: who exactly does this claim to describe, and how were those people chosen?