Amazon Product Image Testing: What to Change First, and How to Read the Result

Search for Amazon's main image requirements and you'll find a dozen guides that contradict each other. One puts the floor at 500 pixels on the longest side. Another says 1,000. A third insists on 2,000. They can't all be right, and it barely matters, because pixel count is not the reason your detail page isn't converting.

Specs are table stakes. Your competitor cleared them too. The real lever is which of your gallery slots answers the question that's stopping the purchase, and in what order you find that out. Most brands never find out, because they change six things at once.

Across our client portfolio the pattern repeats: a brand refreshes its whole image stack, sales move a little, and nobody can say which asset did it. Next quarter, six more changes. That isn't a testing program, it's redecorating. If you'd rather hand the production and the test design to an operator, that's what our Amazon creative services team does, and you can book a call to walk through your catalog before you commit to a reshoot.

Why the spec guides don't move your conversion rate

Because clearing a requirement is not the same as answering an objection. The requirements exist so Amazon can serve a consistent search grid: white background on the main image, product filling most of the frame, no added text or props, a file large enough to zoom. Clear them and you're eligible to compete. That's all you've bought.

The contradictions in those guides are worth noticing for a different reason. Third-party blogs disagree on minimum pixel dimensions, on how many slots you get, and on what apparel main images are allowed to show. The authoritative list lives in Seller Central and it changes. Read it there, confirm your category's exceptions, then stop thinking about it. Specs are a compliance task for whoever uploads, not a strategy question for whoever owns the P&L.

What to change first

Order matters more than volume. Each slot has one job, and the early slots carry far more weight because plenty of shoppers never scroll the gallery at all. We work them in this sequence.

Asset	What it has to answer	The usual failure
Main image	Is this the product I just searched for?	Styled like a brand campaign, unreadable at thumbnail size
Second image	What exactly do I get, and how much of it?	Repeats the main image from a different angle
Third image	How big is it, or how does it work?	Scale left to the shopper's imagination
Remaining gallery slots	Which objection is still standing?	Generic lifestyle photography that answers nothing
Video	Does it actually do the thing it claims?	A long brand film that needs sound to make sense
A+ content	Why this brand instead of the cheaper one below?	Template modules, stock imagery, no comparison

If you only have budget for two assets this quarter, spend it on the main image and the second slot. We've found that's where the measurable movement sits, and it's also the cheapest work to redo. Creative production with us starts around $250 per asset, so a two-asset test isn't a capital decision.

Can you A/B test an Amazon main image?

Yes, through Manage Your Experiments in Seller Central, provided you're enrolled in Brand Registry and the ASIN carries enough traffic to reach significance. Amazon runs the split itself, serving two versions to comparable audiences over the same period, which is the part you can't replicate with a before-and-after comparison.

Three constraints decide whether the tool is open to you at all:

• Brand Registry and the right role. Experiments sit behind brand enrollment and a brand representative role on the account, not an ordinary seller login.
• Published content only. A+ variants have to be live and approved before they can be tested. Drafts don't qualify.
• Traffic. This is the constraint that disqualifies most catalogs. An ASIN with a few hundred page views a week will not separate a small lift from noise inside a sensible window, and Amazon will either tell you so or simply never declare a winner.

Published guidance on which elements are testable, and on how long a test should run, is inconsistent across third-party sources, and Amazon has changed both before. Check what your own account can test under Brands, then Manage Experiments, rather than planning a quarter of tests from a blog post. Including this one.

If your traffic can't support a test

Most catalogs have a handful of ASINs with real volume and a long tail without. On the tail you don't get a clean experiment, so stop pretending you do. Two things still work. Apply what the tested ASINs taught you, because the objection that stopped buyers on your hero product usually stops them on its siblings. And if you read a change sequentially, use a long window in a stable period, call it directional rather than proven, and accept that Q4 will lie to you.

The thumbnail is the real creative brief

Your main image gets judged twice, and the first judgment happens at roughly the size of a postage stamp. A shopper meets your product as one tile in a search grid, beside a dozen competitors, usually on a phone. That's where the click is won, and it's a different design problem from the one your photographer solved.

So evaluate every candidate at thumbnail scale before you look at it full size. Shrink it on your own phone, in a grid with the competitors ranking above you. Plenty of assets that look expensive at full resolution turn into a grey smudge at 150 pixels, and the cheaper-looking competitor with one high-contrast silhouette takes the click.

Video follows the same logic. If the first two seconds don't establish the product without sound, assume the sound is off, because a good share of the time it will be.

What a flat result actually tells you

A test that comes back flat is not a wasted test, and this is where brands quit too early. Flat means the thing you changed wasn't the thing holding the purchase back. That's a finding, and it narrows the field for the next one.

Be honest about the arithmetic, though. Say a page converts at 12%. Moving it to 13% is an eight percent relative gain, and proving a gain that size takes a lot of sessions. Brands routinely run two weeks on a low-volume ASIN, see nothing, and conclude that images don't matter. What they've actually learned is that the sample was too small to answer the question they asked.

When image testing isn't your problem

Sometimes the creative is fine and the listing still doesn't convert. Before you commission a reshoot, rule out the cheaper explanations: you're going out of stock intermittently, your price has drifted above the comparison set, your rating or review count sits below the category's floor, or your ads are buying traffic for search terms the product doesn't serve. Any one of those flattens a conversion rate no image can rescue. The full diagnostic is in why Amazon listings don't convert, and the copy and keyword half of the same job sits in Amazon listing optimization.

Creative also can't fix a positioning problem. If the honest answer to "why this instead of the cheaper one below" is "no reason", then better photography just delivers that news faster.

What this looks like sequenced properly

When we launched Gabb on Amazon, the listing work and the conversion testing ran as part of the launch rather than as a cleanup afterwards, alongside the ad architecture. That channel did over $1M in its first 90 days with 96% sales growth, in an electronics category where competitors had years of catalog advantage. The full Gabb breakdown is here, including how the launch was sequenced.

That sequencing is the point. Testing creative on a page whose price, availability and ad targeting are a mess produces results nobody can read.

Where to start

Take your three highest-revenue ASINs. Put their main images in a grid with the competitors ranking above them, at thumbnail size, on a phone. If yours is the one you'd scroll past, you have your first test and you didn't need a spec sheet to find it. If you'd rather the production and the test design were handled by a team doing this across a portfolio, bring us the catalog.