You ran one survey. Two thousand people online, already collected, raw answers in hand. One step is left, the step every online study needs: make the sample representative. There is more than one honest way to do it. Try a few.
The question asked: would you switch to the new pack? Now weight the raw sample to match the country, however you think best.
Among under-30s: 64%
Change the recipe below and each new answer lands here.
Every switch here is a choice a professional routinely defends. None is the one "right" answer, which is the whole point.
Illustrative dataset, not recorded model output. The swing is patterned on real demonstrations where the same raw data moved roughly five to eight points from weighting choices alone (Cohn, 2016; Clinton, 2026). The point is a general property of weighting, not one survey's number. Built 12 August 2026.