A credible release makes the population, fieldwork, questions and limits easy to find.
Start with the population and timing
Who was the survey intended to represent, and when were interviews completed? A national adult sample and a local electorate are different populations. The publication date is useful, but the field dates tell you when the responses were collected.
Look for recruitment and the questionnaire
How were respondents selected and interviewed? Read the exact wording, response options and relevant question order. A headline paraphrase is not enough to assess whether a question measured the concept being discussed.
Check the base and uncertainty
Find the total sample and the unweighted size of any subgroup being reported. Ask how weighting was applied and how uncertainty was calculated. Statistics Canada cautions that conventional sampling-error claims require a probability sample. A large number of responses alone does not resolve selection bias.
Identify who commissioned the work
A release should disclose the sponsor or commissioning organization, explain what was measured and acknowledge the limits. Our editorial standard is simple: a reader should be able to inspect the evidence without needing to contact the author for the basic method. A poll is a measured snapshot, with uncertainty, rather than a promise about a future outcome.
Evidence note. This is an editorial research guide or perspective, first published on 25 October 2026. It does not report original Northbound polling. Read our publication policy.