Product photography gets discussed as taste. Which background, which angle, does it match the brand.
It is not mainly a taste question. Images are the largest thing on almost every product page, which makes them the main reason the page is slow. They are the only substitute a customer has for holding the object, which makes them the main reason people do or do not buy. And they are increasingly read rather than looked at — by search engines, by assistants, and by every customer using a screen reader.
Three jobs, one asset, and it is usually the line item that gets cut when the budget tightens.
What a product photo is for
Online, the photograph is doing the work that picking something up does in a shop. That means the standard is not "attractive". It is "answers the question that would otherwise stop the sale".
Those questions are boringly consistent across categories:
- How big is it, really?
- What is the material, up close?
- What does it look like from the side and the back?
- What does it look like in use, at the scale a person uses it?
- What exactly is in the box?
- Is the color on my screen the color of the object?
A single well-lit hero shot on white answers none of these. It is a thumbnail, and treating it as the product page is why so many stores answer these questions in the returns process instead.
The shot list worth paying for
For most physical products, six images:
- Hero on plain background — the catalog image, the one that appears in listings and search results.
- Scale reference — next to a hand, a familiar object, or on a person. The most under-shot image in e-commerce and the one that prevents the most returns.
- Detail at close range — texture, stitching, finish, the part the price is justified by.
- The other sides — back and profile. Absence here reads as concealment.
- In context — being used, where it is used.
- What arrives — everything in the box, laid out.
If budget forces a cut, keep 1, 2 and 3. Scale and texture are what the screen takes away.
The performance half
Here is the part that is not about photography at all.
On a typical product page, the main image is the largest thing the browser has to paint. That makes it the element that decides how fast the page feels, and page speed is a revenue variable rather than a technical score — we made that case in full in speed is a pricing decision.
Which means an uncompressed 4 MB photograph is not a beautiful asset. It is a discount you are giving to everyone on a phone.
Four rules, and none of them costs quality if done properly:
Serve modern formats. WebP or AVIF at sensible quality, with a fallback. The savings against JPEG are large and visually indistinguishable at the sizes anyone actually views.
Serve the size being displayed. A 3000-pixel-wide original delivered to a 390-pixel phone is wasting 90% of the bytes. Multiple sizes, chosen by the browser.
Load the first image eagerly, everything else lazily. The hero should start downloading immediately. The gallery images below should not compete with it.
Reserve the space. Width and height on every image so the page does not jump as pictures arrive. Content that moves under a finger is a real usability failure and it is entirely preventable.
One illustration of how much this matters, from our own work: on Finkonkurs we replaced a Material Symbols icon webfont of 3.8 MB with inlined SVG. That was icons, not photographs, and it was still multiple megabytes of the page. Images are usually larger.
The part a machine reads
Every image should have alt text, and it is worth understanding that alt text now does three jobs rather than one.
It is the image for a blind customer. That is the original purpose and the sufficient reason. In the EU and UK this also sits inside a legal obligation rather than a best practice — see the accessibility bill comes due.
It is the image for a search engine. Images are matched on their description.
It is increasingly the image for an AI assistant summarizing your product for a buyer who never opens your page. Where a model is reading your catalog, the alt text is a large part of what it knows the product looks like.
How to write it
Describe the thing, not the file. "Blue ceramic mug, 350 ml, matte finish, seen from the side" — not "IMG_4471" and not "mug".
Do not start with "image of". It is already an image.
Do not stuff keywords. It reads as spam to a search engine and it is genuinely unpleasant to listen to.
Leave it empty when the image is decorative. An empty alt="" tells a screen reader to skip it, which is correct for a background flourish and much better than making someone listen to a description of a gradient.
Two principles we hold on this come from our own accessibility and extraction work, and they generalize. On Legible, which produces alt text and reconstructed tables for PDFs and images, unreadable values stay null and are flagged rather than being guessed — because a confident wrong value is worse than no value at all. On RecipeReel, which reads on-screen text from cooking videos, a quantity that cannot be determined is left blank rather than invented.
Apply that to your catalog: an alt text that describes a color you are not certain of is worse than a shorter, true one. Especially since the customer using it cannot check.
The consistency that makes a catalog feel like a shop
Individually good photos still produce a bad catalog if they were shot at different times by different people.
One background, one lighting setup, one crop ratio across everything in a category. A listing grid where the images are inconsistent looks like a marketplace of unrelated sellers, and it is one of the strongest signals people use to judge whether a store is real.
One aspect ratio, enforced. This is the single easiest win. Mixed ratios produce a ragged grid or distorted images, and neither is fixable in CSS without cropping something badly.
A file naming convention, so nobody has to guess which of eleven files is the hero.
This is also a catalog data question, not only a photography one. Bad or missing catalog attributes change what people buy — pushing them toward the one attribute that is always populated, which is price. We wrote that up separately in your catalog is the product, and images are the most visible attribute of all.
Budgeting it honestly
The uncomfortable arithmetic: for a store with a few hundred products, photography frequently costs more than the website. Our e-commerce builds start at $499 over one to five weeks. Professional photography of 200 products, done properly with six shots each, will exceed that.
This is exactly why it belongs in the brief rather than being discovered in week five. Three honest options:
Shoot it properly. Right for products where appearance is the purchase decision.
Shoot it yourself, systematically. A window, a white backdrop, a phone on a tripod, one setup used identically for everything. Consistency matters more than equipment, and a disciplined amateur set beats a mixed professional one.
Use supplier images, and know the cost. Fast and free, and your page now looks like every other retailer selling the same item. Acceptable for commodity goods, fatal where you are trying to justify a higher price than the competitor showing the same photograph.
Whichever you choose, put it in the scope with a date. A store build waiting on photography is the most common reason a fixed-price project sits still, which is why we list it explicitly in the brief that gets you a real price.
The ten-minute audit
On your own store, on a phone, on mobile data:
- Time how long the main product image takes to appear.
- Check its file size and format. Anything over a few hundred kilobytes for a product shot needs work.
- Count the images per product. Is scale answered? Is texture?
- Scroll the listing grid. Do the images share a ratio, a background, a crop?
- Read the alt text on five products. Would it let someone buy without seeing?
- Check whether the page jumps as images load.
Every one of those is fixable without a redesign, and together they move both the speed number and the sales number.
Tell us what you need. You get a fixed price and a delivery date back — free, and with no obligation to proceed.
Frequently asked questions
How many images per product?
Five or six for most physical goods. Fewer than three and you are leaving questions unanswered; more than eight and people stop looking. The number matters less than the coverage: scale, texture, other sides, in use.
Are lifestyle photos or plain-background photos better?
Both, for different jobs. Plain background is the catalog and comparison image — it is what belongs in the grid and in search results. Lifestyle answers "what is this like to own". A store with only lifestyle images is hard to shop; one with only plain shots is hard to want.
Is video worth it?
For products where movement or scale is the question — clothing, furniture, anything mechanical — a short clip answers more than another photo. Keep it short, do not autoplay with sound, and never let it block the page: a heavy video on a product page undoes the speed work above. Give it a poster image so the space is not an empty rectangle while it loads.
Should I use AI-generated product images?
For your actual products, no. The customer is buying the object in the picture, and any difference between them is a return, a dispute or a misrepresentation claim. For backgrounds, scene composites and decorative context, it is a tool like any other — as long as the product itself is real and accurately represented.
What about images that contain text?
Avoid them for anything important. Text in an image cannot be read by a screen reader, cannot be translated, cannot be searched, and does not reflow on a phone. If a claim matters enough to put on the page, put it in the page as text next to the picture.
