Pinion Polls
Show shoppers your listing options, get a vote plus a written reason from each shopper, and act on the explanations, not just the pie chart.
Pinion Polls
A Pinion Poll shows shoppers your options, asks them to pick one, and makes them explain why. It's the test you'll run most.
Setting one up takes four steps: Internal Setup, Options, Audience, Review and Submit.
Before you start
Have your options ready as finished files. Shoppers judge what you show them, so a mocked-up option competing against a polished one will lose for the wrong reason.
Change exactly one variable between options. If you want to test a tilted photo and a new title, that's two polls.
Pick a poll type
Six types, depending on what you're comparing:
| Type | What it does | Limit |
|---|---|---|
| Image Split Test | Show image options, shoppers pick one | Up to 8 |
| Video Split Test | Show video options, shoppers pick one | Up to 3 |
| Text Split Test | Show text options: titles, bullets, brand names | Up to 6, 1000 characters each |
| Search Simulation | A realistic Amazon search results page you control entirely | Up to 8 products |
| Stacked Image Test | Compare whole image galleries rather than single images | 2-3 galleries of 3-10 images |
| Ranked Test | Shoppers order options best to worst | 3-8 options |
Search Simulation is the one to understand. Every other type asks shoppers to compare your options side by side, which is not how anyone shops. Search Simulation drops your image into a realistic search results page next to real competitors, so shoppers choose the way they actually would. For anything about click-through rate, including main images, titles, and price, this is the honest test.
If you'd rather not start from scratch, press START WITH A TEMPLATE and pick from the test library. The template pre-fills the poll type, the shopper instruction and the goal.
Write your internal goal
Two fields here, and neither is shown to shoppers:
Your internal goal: why are you running this test? Write the decision you're trying to make, not the mechanics. "Deciding whether to reshoot the main image before Q4" beats "testing images". In three months this is the only record of what you were thinking.
Product or category name keeps things findable once you have a few hundred tests.
Add your options
This is where you give shoppers the context and the things to choose between.
The instruction matters more than people expect. It's the framing every shopper reads before they choose, and it decides what they're judging. The default is generic; replace it.
For click-through questions, put them in the shopping mindset:
Pretend you're browsing on Amazon. Which product would you be most likely to buy? Why?
For a specific attribute, ask about that attribute:
Which infographic image do you prefer? Why?
Don't reveal which option is yours, and don't hint at the answer. "Which of these looks more premium?" will get you votes for whichever looks more expensive, even when the cheap-looking one sells better.
Products can be imported by ASIN. Paste them separated by spaces or commas, or create them by hand.
Choose your audience
ProductPinion's shoppers are best for first impressions. Your own customers already know your brand, so don't use them to judge whether a new shopper would understand your listing.
Who's audience to use? ProductPinion's shoppers, or your own customers via a public link. Your own audience is free on Brand and Agency plans and useful as a cross-check, but they already know your brand, so don't use them to judge first impressions.
How many shoppers? At least 50 per option. With two options, 100. The recommendation shown next to the slider scales with your option count.
Standard Demographics defaults to Fastest Available Shoppers, which returns results quickest. Narrow it only when the demographic genuinely changes the answer. For a product only bought by one group, a general sample will mislead you. Every constraint you add slows the test down.
Advanced Targeting lets you pick up to two further facets: ethnicity, relationship status, education, employment status, and interests.
You can also exclude shoppers from previous tests, up to three, so a follow-up poll reaches fresh eyes. Worth doing whenever you're re-testing a variation of something you've already shown.
Review and submit
The review screen recaps all four steps and flags problems: an instruction you left at the default, or too few shoppers for the number of options. Both are worth going back for.
While it runs
The poll moves to In Progress and fills up over hours rather than days. You don't need to wait for completion to read it. If the split is 80/20 after sixty responses, it will very likely still be 80/20 at the end.
Reading the result
Full detail in Reading your results, but the short version:
Start with the written explanations, not the pie chart. The vote tells you which option won; the explanations tell you why, and the why is what you can act on. A 60/40 win where forty people all say "I couldn't tell what size it was" has handed you a bigger fix than the one you were testing.
Then check the certainty figure. Above 85%, act. Between 66% and 85%, treat it as a lean. Below 66%, your options were too similar to separate. Add shoppers or make the difference bolder.
Common mistakes
- Testing two changes at once. You'll get a winner and learn nothing.
- Too few shoppers. Under 50 per option, noise wins.
- A leading instruction. You'll get the answer you implied.
- Comparing options side by side when the real question is click-through. Use Search Simulation.
- Stopping at the vote. The explanations are the product.