How product photos affect conversion rate: the data

How product photos affect conversion rate, and which images actually move the number

Retailers expected buyers to send back 16.9 percent of their 2024 sales. The National Retail Federation and Happy Returns put the value of that at 890 billion dollars across retail. A large share of it is avoidable. The item was fine. The photo failed to tell the buyer what was coming, so the product arrived smaller or shinier or flatter than the picture implied.

That is the frame worth keeping for the whole question. How product photos affect conversion rate is a question about what an image tells the buyer. Every image on a product page either settles a specific doubt or it does not. The ones that settle doubt move the number. The rest get skipped. What follows is the evidence, the five questions your gallery has to answer, and how to measure the effect on your own store.

What the research says about how product photos affect conversion rate

Start with where attention goes. Baymard Institute’s product page usability study found that 56 percent of test subjects explore the product images first. They do that before they read the title or the description. The gallery runs the first round of persuasion whether you design it that way or not.

Attention is selective. Nielsen Norman Group’s eyetracking work settles it. Users pay attention to information-carrying images that show content relevant to the task at hand. They ignore decorative images. In one eCommerce test the same study clocked 0.9 fixations per product thumbnail against 4.4 for the description. Text wins that comparison, and the reason matters. Generic mood-setting thumbnails lose attention. Images carrying real product information win it.

When the images carry information the sales effect shows up in listing data. eBay tells sellers that listings meeting its photo standards are 4.5 percent more likely to sell. The bar is low: at least 500 pixels on the longest side, no added text or graphics, and a direct upload. Cornell Tech researchers studying secondhand marketplace listings found shoes with better photos 1.17 times more likely to sell than shoes with worse ones. Handbags ran at 1.25 times. Salsify’s 2025 consumer research covers the other side of the trade, where 71 percent of shoppers report returning an item because it did not match the listing.

These are separate studies with separate definitions of product image quality. They still point one direction. Better information in the image, more conversions out.

Why some photos convert and others get skipped

What separates them is the job each frame does. A hero shot on a white background does one job well. It confirms the product exists and looks like the buyer expects. After that first frame its persuasive value drops to near zero, because it stops answering new questions. A second white-background shot from another angle adds little.

The photo of the bag on a shoulder answers a new question. So does the mug beside a hand for scale, and the macro of the fabric weave. Those frames earn the 4.4 fixations. Those frames turn a maybe into a checkout.

The usual advice says add more photos and make them prettier. The research says something narrower. Add photos that answer more questions in a set order. Stop when the questions run out. A five-image gallery where every shot settles a different doubt beats a twelve-image gallery of near-identical glamour shots.

Our guide to AI product photography covers how to generate those variants without a studio. The strategy question comes first, and it is which doubts your gallery leaves open.

The five questions a buyer asks a photo

Most conversion loss on a product page traces to one of five unanswered questions. Walk your own gallery against the list.

How big is it. Baymard found 42 percent of users try to work out product size from the images. Its 2026 product page benchmark puts 37 percent of sites at no single in scale image. Written dimensions do not fix this, because most people cannot picture 24 centimeters. A photo of the product in a hand or on a desk or beside a common object does. Missing scale is one of the quiet reasons an item comes back.

How does it sit on a real body. The same benchmark puts 23 percent of sites at no human model image for wearable products. Baymard’s testing found shoppers lose confidence without that view. They read mannequins as cheap. Shoppers also reject cosmetics when no image shows a model with a skin tone close to their own, so each variation needs models across a range of tones. On-model photography answers the fit question that a flat lay leaves open.

What is it made of. Texture and finish and sheen do not survive a small thumbnail. A macro shot of stitching or grain carries information no caption matches. Give this question a frame of its own on anything where the material drives the price.

How will it look in my life. A lifestyle shot puts the product in the context the buyer is already imagining. Staging earns its keep here, and it is the reason localized and seasonal scenes test well. Our post on why white backgrounds are costing you sales covers the ad-spend side of that.

Can I trust this. Shoppers treat customer-submitted photos as more objective than official ones. Baymard’s benchmark finds 63 percent of sites give users no way to navigate across reviewer-submitted photos. A gallery that mixes clean studio frames with credible real-world ones answers the trust question that polished photography alone cannot.

Answer those five and you cover the doubt behind most abandonment. Leave one open on a high-traffic listing and you pay for the gap twice, in lost carts and in returns.

Which questions matter most for your category

The five questions carry different weight depending on what you sell. Ranking them for your catalog tells you which frame to shoot first.

Apparel and footwear lean on fit. The on-model frame does most of the work, and it belongs second in the gallery behind the hero. Buyers want the garment on a body close to their own, so a size range across models beats one model in more poses. Scale matters less here, because a size chart already does part of that job.

Jewelry, accessories and small homeware invert that ranking. Scale is the top leak. A ring or a candle or a desk organizer photographs the same at every size, and buyers guess wrong until a hand or a coin or a shelf appears in the frame. Put the in scale image second and the material macro third.

Furniture and large homeware need context above everything else. A sofa on white tells the buyer nothing about whether it fits the room they have. A staged room shot answers that. A second staged shot in another style widens the set of buyers who see themselves in it.

Beauty and consumables run on trust and material. Skin tone coverage decides whether a shopper can read the shade at all. Reviewer photos carry more weight here than in any other category, because buyers discount brand imagery of a product that touches their face.

Tools and electronics sit closest to a spec sheet, which makes the port and the connector and the finish the frames that matter. Buyers arrive with a compatibility question. Shoot the detail that answers it rather than another angle of the whole unit.

How to test product photos on your own store

Industry averages start the conversation. They do not answer it. A 1.25 times lift on handbags in a Cornell study predicts nothing about your listing. The only number that settles it is the one you measure.

Treat the gallery as a testable variable. Pick your highest-traffic products and change one thing, an added on-model shot or a scale reference. Split test it against the current version. Watch conversion rate and add-to-cart. Watch the return rate on that SKU over the following weeks too, because a photo that lifts conversion while raising returns sets an expectation the product cannot meet. Salsify’s 71 percent matters as much as any conversion figure.

Run the test long enough to mean something. A listing with 200 sessions a week needs about a month before the split reads as signal, and swapping the image mid-run resets the clock. Hold everything else steady. A price change or a new title in the same window makes the result unreadable. Log the start date and the exact frame you added, because six months later the only record of what worked is the one you wrote down.

Audit before you shoot. Open your five best-selling listings and score each gallery against the five questions. The gaps you find are your test backlog, ranked by traffic. Make scale and on-model frames non-negotiable for the categories that need them. Tag return reasons in your helpdesk so you can see when an image oversells.

The reason most sellers skip this test is production cost. A studio session runs 500 to 2,000 dollars and takes weeks, so five gallery variations across a catalog never reach the top of the list. Generating variants changes that math. Vision produces a scale reference or an on-model version for a few dollars in minutes, which turns eCommerce product images into something you test rather than something you commission once a year. Running that across a full catalog carries its own failure modes, and our write-up on an AI photoshoot at catalog scale covers them. Marketplace sellers should read the Amazon image guide for the compliance side.

Teams argue photography budgets as a brand question. The evidence makes it an information question. Every frame either closes one of the five doubts or takes up space in the gallery. Sort your own frames into those two piles and you know what to shoot next, and your analytics tell you what it was worth.

References

Where attention goes

Baymard Institute, product page research on shoppers exploring images before text. https://baymard.com/research/product-page
Baymard Institute, “Product Page UX Best Practices 2026,” source of the in scale, human model and reviewer image figures. https://baymard.com/blog/current-state-ecommerce-product-page-ux
Baymard Institute, “Provide at Least One ‘In Scale’ Image.” https://baymard.com/blog/in-scale-product-images
Baymard Institute, on human model imagery, mannequins and skin tone. https://baymard.com/blog/human-model
Nielsen Norman Group, “Photos as Web Content,” on decorative versus information-carrying images. https://www.nngroup.com/articles/photos-as-web-content/

What sells