Product Image A/B Testing 2026: How to Test Main Images on Amazon, Shopify, Shopee and Lazada

Most sellers choose their main product image the same way: they shoot a few options, pick the one they like best and move on. The trouble is that your taste is not your customers' taste. A main image that looks elegant to you might be the one shoppers scroll straight past. The only reliable way to find out is to test it.

Product image A/B testing means showing two versions of an image to comparable groups of shoppers and measuring which one gets more clicks, add-to-carts or sales. Done properly, it turns guesswork into evidence. Done badly, it produces confident-looking results that are really just noise. This guide covers how to do it properly, whatever platform you sell on.

Why the main image is worth testing first

On a marketplace search page, shoppers see your main image, title, price and star rating, and almost nothing else. Of those, the image takes up the most space and is processed first. That makes it the single biggest lever on click-through rate, and click-through rate feeds directly into how much traffic a listing earns over time.

Secondary images matter too, but they only influence shoppers who have already clicked. If you have limited time and traffic, start with the main image, then move on to the second and third images in the gallery, which tend to get the most views after the hero shot.

Where you can run image tests

Not every platform gives you a built-in testing tool, so your method depends on where you sell.

Amazon Manage Your Experiments

If you are brand registered, Manage Your Experiments is the gold standard because Amazon splits real shoppers between the two versions at the same time. That removes the biggest problem with home-made tests: changes in demand from week to week. Amazon only makes the tool available on ASINs with enough traffic, so you may find low-volume listings cannot be selected. Most sellers set the maximum duration for a cleaner result, and you should make sure both images meet Amazon's main image rules (pure white background, no added text or graphics) before you launch.

Paid ads as a fast testing ground

If your platform has no testing tool, ads are the quickest way to compare images. Run two ad variants that are identical except for the product image, with the same budget, audience and placement, and compare click-through rate. Ads give you thousands of impressions in days rather than weeks. The caveat is that ad audiences behave a little differently from marketplace searchers, so treat the winner as a strong hypothesis and confirm it on the listing itself.

Sequential testing when nothing else is available

Sequential testing means running version A for a period, then version B for an equal period, and comparing. It is the weakest method because anything else that changes between the two periods (a payday, a public holiday, a competitor's sale, a mega campaign like 11.11) contaminates the result. To make it as fair as possible:

• Use periods of at least two full weeks each, so both include every day of the week.
• Avoid campaign periods and public holidays entirely.
• Keep price, title, vouchers and ad spend unchanged throughout.
• Compare click-through rate (clicks divided by impressions) rather than raw sales, since it is less affected by overall traffic swings.
• If you can, repeat the test by switching back to A and checking the effect reverses.

What to test: ideas that tend to move the needle

Small tweaks rarely produce a measurable difference. Change something a shopper would notice in a thumbnail the size of a postage stamp. Good candidates include:

• Angle: a three-quarter view against a straight-on front view. Three-quarter angles often show depth and shape better.
• Frame fill: the product filling roughly 85 per cent of the frame against a smaller product with more white space. Bigger usually reads better on mobile.
• Single item against bundle: showing exactly what is in the box can lift clicks for multi-packs and sets.
• Packaging in or out: for gifts and premium items, the box can signal quality; for commodity items it can just add clutter.
• Colour variant shown first: leading with your best-selling colour, or the most eye-catching one, can change click-through noticeably.
• Background on platforms that allow it: on Shopify, Shopee and Lazada you have more freedom than on Amazon, so a lifestyle or tinted background against plain white is a valid test.

Whatever you test, change one thing at a time. If version B has a new angle, a new background and a bigger crop, you will never know which change caused the result.

How long to run a test, and how to read the results

The most common mistake in image testing is stopping too early. Early results swing wildly. A variant that looks 30 per cent better after three days can easily end up level after three weeks.

Rules of thumb for sample size

• Run for at least two full weeks, and longer on lower-traffic products.
• Aim for several hundred clicks, or ideally 1,000 or more views, per variant before drawing conclusions.
• The smaller the difference you are trying to detect, the more traffic you need. Spotting a 2 per cent lift takes far more data than spotting a 20 per cent lift.
• Decide the test length in advance and stick to it. Checking daily and stopping the moment one version pulls ahead inflates your false-positive rate.

Measuring the right thing

For a main image, the primary metric is click-through rate, because that is what the image most directly controls. Also watch conversion rate on the product page. An image that wins clicks by overselling the product can increase clicks but reduce sales and increase returns. The best result is a higher click-through rate with stable or improved conversion.

Use a free online significance calculator to check whether a difference is likely real. A common threshold is 95 per cent confidence. If your result falls short of that, the honest conclusion is "no clear winner", which is still useful: it tells you to test a bolder change next time.

Keep your test images technically identical

A test is only fair if the only difference is the thing you are testing. If version A is a crisp 2000 x 2000 pixel JPEG and version B is a soft 800 x 800 pixel export, you are testing image quality rather than angle or composition. Before uploading variants, make sure they share:

• The same pixel dimensions and aspect ratio (for example 2000 x 2000 for Amazon, 1:1 at 1000 x 1000 or above for Shopee and Lazada)
• The same file format and similar compression, ideally under 1 MB for marketplaces that recompress heavily
• The same white balance and exposure, unless that is what you are testing
• The same padding around the product

If you sell on several marketplaces, a batch tool such as PixelPrep can export each variant at every marketplace's required size in one pass, so you are not accidentally comparing a sharp image against a badly resized one.

Roll out winners carefully

When you find a winner, apply it to the listing you tested first and confirm the improvement holds over the following weeks. Then consider applying the same principle, not necessarily the same image, to similar products. If a three-quarter angle beat a front view on one pair of trainers, it is a sensible default for the rest of your footwear range, and you can retest later.

Keep a simple log of every test: the product, what you changed, dates, traffic, result and confidence level. Over a year, that log becomes a style guide based on how your customers actually behave, which is far more valuable than any single test. When you reshoot or refresh a range, a tool like PixelPrep makes it quick to reformat the winning style across your whole catalogue.

Product image A/B testing checklist

1. Start with the main image on a product that gets steady traffic.
2. Choose your testing method: a built-in experiment tool, paid ads, or a careful sequential test.
3. Change one visible element only, and make the change bold enough to notice in a thumbnail.
4. Export both variants at identical dimensions, format and compression.
5. Check both images meet the marketplace's image rules before launching.
6. Fix the test length in advance: at least two weeks, longer for low-traffic listings.
7. Avoid sale campaigns, holidays and price or voucher changes during the test.
8. Judge on click-through rate first, then confirm conversion rate has not dropped.
9. Check statistical confidence before declaring a winner.
10. Log the result and use it to guide your next shoot.

Image testing is not glamorous, but it compounds. Each test you run teaches you something about your buyers, and every winning image keeps working for as long as the listing is live.