Start with the source
Election seasons can bring invented numbers dressed as research. A polished graphic is easy to circulate and may have no identifiable firm behind it. Pause before forwarding it. A genuine release should let you work backwards from the claim to a responsible organisation, a method and a permanent source.
1. Is a real, named firm responsible?
Check the organisation’s name and its own contact details. A logo on a graphic is not enough. Look for the release on the firm’s website and use that site’s contact channel if the attribution is unclear.
2. Is the sponsor or funder stated?
Who commissioned and paid for the poll is part of its context. Disclosure does not by itself settle whether a study is good, but its absence removes information readers need.
3. Is there a methodology statement?
Find the population, sampling method, achieved sample size, fieldwork dates, interview mode and margin of error or other appropriate uncertainty statement. Without these, it is difficult to judge who the numbers describe or how the evidence was collected.
4. Do the numbers survive arithmetic?
For a single-choice question, check whether the shares make sense together, allowing for rounding. Multiple-choice questions can legitimately total more than 100%, so read the exact wording. Also check whether a county or other subgroup contains enough interviews for the precision being claimed. A large national sample does not make every subgroup large.
5. Does it have a permanent home?
Check the firm’s own registry or release page. Every genuine ResearchCannon poll has a permanent page on this site with its methodology. If a graphic attributed to ResearchCannon has no page here, it is not ours. Use Verify a Poll to check the source and report a questionable attribution.
A poll you cannot interrogate is a rumour with decimals.