Reading Your Results
Understand votes, shopper explanations, certainty figures, and audience breakdowns, then filter noise, use AI tools, and share results.
Reading Your Results
Read the explanations first
The pie chart is the headline; the explanations are the story. The vote tells you which option won. The written reasons tell you why, and the why is the only part you can act on.
Skim every response once without counting anything. You're looking for the sentence that repeats, the thing several shoppers mention unprompted. That recurring sentence is usually worth more than the result you were testing for.
Then the certainty figure
ProductPinion reports how confident you can be that the winner is really the winner:
| Certainty | What it means | What to do |
|---|---|---|
| 85% and above | A real difference | Act on it |
| 75-84% | Probably real | Act if the change is cheap; add shoppers if it's expensive |
| 66-74% | Too close to call | Add shoppers, or make the options more distinct |
| Below 66% | No signal | Your options aren't different enough to separate |
A close result is information too. If shoppers genuinely can't tell your options apart, neither can your customers, and you've just learned the change you were agonising over doesn't matter. Spend the effort elsewhere.
Check the demographic split
Open the demographic breakdown before you act on anything close. A 50/50 result is sometimes two groups disagreeing strongly, younger shoppers picking A, older shoppers picking B, which is a completely different finding from genuine indifference, and a reason to look at who you're actually selling to.
Filter out the noise
Not every response is useful. You can hide low-effort answers, report ones that look like bad-faith responses (reported responses can be replaced), and favourite the ones you'll quote to your team. Filter by option, demographic, source, or keyword to test a hunch about who said what.
Let AI do a pass
Analyse for me returns key points, sentiment split into positives and negatives, and suggested actions. PinionGPT lets you ask questions of the response set directly, such as “what did people who chose B say about the price?”
Use them after your own read, not instead of it. The summary is good at volume and bad at noticing the one oddly specific comment that changes your mind.
Share and continue
Share Results creates a public link, useful for clients and suppliers. Add more results tops up a finished test with more shoppers at the same demographics. Launch Similar clones the setup for your next variation, which is how a single test becomes a testing habit.