An election graphic says how a group voted, while another reports how many ballots a candidate received. Both may appear during the same broadcast, but they are not two views of an identical dataset. One may come from an exit poll; the other from election administration.
The American Association for Public Opinion Research explains the distinction in its guide to exit polls. Exit polls survey voters and can help describe voting patterns. They do not replace the process of counting ballots.
Ask what the number represents
Before sharing a percentage, look for the population, question and collection method. Is the figure about all voters, one subgroup, people interviewed at selected polling places or a wider design that also reaches early and postal voters? The headline alone may not tell you.
Small subgroup estimates deserve particular care. Splitting a survey into narrower categories leaves fewer respondents behind each comparison. A dramatic difference on a chart can look more decisive than the underlying evidence supports.
Exit-poll estimates can also be revised as further information becomes available. AAPOR describes adjustments using election results. An early screenshot and a final table may therefore differ without either being a ballot-count record.
Keep the time attached to the claim
For a useful note, save the publisher, release time, table title and methodology link. Write “this survey estimated” rather than “the ballots proved” when describing a polling result. If a later version changes the figure, compare the same population and question before drawing conclusions.
Imagine an early graphic about first-time voters being compared with a later table about all younger voters. Even before considering revisions, those labels describe different groups. A careful reader checks that mismatch before debating the percentages.
For the official result, consult the relevant election authority and its stated stage of counting or certification. For explanations of voter behaviour, read the survey’s methods and limitations. Keeping those two purposes separate makes election coverage easier to follow without asking a poll to do a ballot count’s job.