A/B Testing Product Listings on Amazon: What to Test and How
A/B Testing Product Listings on Amazon: What to Test and How
Stop guessing which image converts. Test it, measure it, ship the winner.
Why most sellers test wrong
Almost every seller has an opinion about which photo is better. Almost none have data. A/B testing on Amazon is now built into Seller Central, but the tool only helps if you set it up properly: one variable at a time, enough traffic to reach significance, and a clear hypothesis before you start. This article covers exactly what to test, in what order, and what numbers to expect.
What you can test on Amazon, and what you cannot
Amazon offers Manage Your Experiments (MYE) inside Seller Central, free for Brand Registered sellers. It runs server-side, splits traffic evenly, and reports conversion lift with confidence intervals.
Currently testable elements
- Main image (the hero shot, the one that shows in search).
- Product title.
- A+ Content (the rich enhanced content below the bullets).
- Bullet points (rolling out in some categories).
- Product description.
Not directly testable inside MYE
- Secondary images and infographics (slots 2-7). You can rotate them manually and watch sessions data, but it is not a controlled experiment.
- Pricing (handled in Pricing Dashboard, not MYE).
- Backend keywords (not visible to buyers).
Eligibility
You need Brand Registry, the product must have enough traffic (Amazon estimates roughly 1000+ sessions a week for a usable result), and the test runs 4-10 weeks depending on volume.
The order to test in: highest impact first
Not every element moves the needle equally. Test in this order, do not skip ahead.
1. Main image
The main image is the single biggest lever in your listing. It controls click-through from search, which feeds into the conversion rate that Amazon uses to rank you. Realistic lifts from a winning main image test: 5% to 25% in conversion rate, with outliers up to 40% on poorly-shot original listings.
What to vary: background tone (pure white versus near-white), product angle, fill percentage, packaging in or out, color of the product on display (lead with the bestseller variant).
2. Product title
The title affects both relevance ranking and click-through. Test order of keywords, presence of brand name at the start, inclusion or exclusion of size and color in the title.
Realistic lifts: 3% to 12% in conversion, sometimes more from search impressions changing.
3. A+ Content
A+ matters for conversion among visitors who scroll. Test modular layout (image-text rows versus comparison tables), the hero banner, and the headline.
Realistic lifts: 2% to 10% in conversion. Lower than main image because fewer visitors see it.
4. Bullet points and description
Smallest single-element impact, but cumulative. Test feature-first versus benefit-first phrasing, presence of dimensions in bullets, and bullet length.
Realistic lifts: 1% to 5%.
Sample size, duration, and reading the results
The number one mistake is calling a test too early. A 12% lift after 200 sessions per arm is noise, not signal.
How much traffic you need
Rule of thumb: each variant needs at least 1500-2500 sessions and 100+ unit sales before you trust the result. Amazon MYE will tell you when the test reaches significance (usually labeled high or very high confidence). Wait for it.
How long to run
Minimum 14 days even on high-traffic listings to absorb day-of-week patterns. For mid-volume listings (5000-15000 sessions per week), plan on 4-6 weeks. For low-volume listings (under 2000 sessions per week), MYE will either reject the test or run it 8+ weeks.
Reading the dashboard
MYE shows two numbers that matter: conversion rate and units sold per session. Watch both. A variant can win on conversion rate but lose on units sold if it draws in lower-intent traffic, this happens with overly clickable but vague titles.
One variable at a time
If you change the main image and the title in the same test, you have no idea which one drove the result. Even if MYE technically allows multi-variable changes, do not do it. Run sequential tests.
- Testing in low season. Q1 traffic patterns differ from Q4. Test mainly during stable months, not during Prime Day, Black Friday, or major category seasonalities.
- Running ads differently across variants. Amazon assigns ad traffic to whichever listing version a buyer sees. Keep your PPC unchanged during the test, no new campaigns, no big bid changes.
- Testing on a new product. New products have erratic traffic and conversion baselines. Wait until you have at least 90 days of stable sales.
- Calling a winner at low confidence. If the dashboard says low or medium confidence, the result is not reliable, no matter how big the percentage lift looks.
- Testing a tiny change. Two main images that differ only by 5% in background brightness rarely show a measurable difference. Test meaningfully different versions.
- Ignoring negative results. A test that says no significant difference is still useful, it tells you to stop spending money on that decision.
Etsy does not offer native A/B testing. Sellers rotate listing photos manually and watch shop stats over 2-4 week windows, accepting that the signal is dirtier. Shopify has built-in A/B testing in some plans, plus third-party apps like Intelligems, Neat A/B, and Shogun A/B Testing. eBay has no native testing tool. The discipline is the same: one variable, enough traffic, enough time.
Send us your current main image and listing URL. We will produce a challenger version designed against your category, ready for Manage Your Experiments. If it loses, you keep your original. If it wins, you keep the lift.