Shopify Variant Image Coverage Checker

What variant image coverage actually measures

Every Shopify variant has one optional image slot. In the product feed it appears as featured_image, and it is either an image object or null. When it holds an image, themes that support variant image switching can swap the gallery, the product card, and the cart line thumbnail to that photo the moment the shopper selects that variant. When it is null, the theme falls back to whatever the product's first image happens to be.

Coverage is the share of variants that hold an image. If a store has 3,400 variants and 1,980 of them have a featured image, coverage is 58.2 percent. That is the headline number this tool computes from your actual catalog rather than from a rule of thumb.

The number is useful as a trend line. Run it monthly, watch it move, and you will notice the day someone bulk imported 200 products without touching image assignment. As a pass or fail grade, though, the headline percentage is a bad judge, and the next section explains why.

One more mechanical detail worth knowing. The featured image is set per variant, not per option value. Assigning a photo to Blue does not exist as an operation in Shopify. You assign it to Blue Small, Blue Medium and Blue Large individually, or you assign it to one of them and leave the rest empty. Plenty of catalogs are in that half assigned state without anyone realizing, which is exactly the pattern the group analysis below is built to surface.

Why a null image is not automatically a bug

This is where most variant audits get it wrong, including some paid ones. They count nulls, call every null a problem, and hand you a scary number that you cannot act on.

Consider a t-shirt in one color and five sizes. Five variants, one photo, four nulls. Should you shoot four more photos of the identical shirt? Obviously not. The shirt looks the same in every size, the shopper knows that, and adding size specific photography would waste money and slow the page down for no benefit.

Now consider the same t-shirt in eight colors and five sizes. Forty variants. If the store assigned one photo per color to a single size and left the other four empty, thirty two variants show as null. Only some of that is a real problem. The eight colors are all photographed, so the catalog is genuinely covered at the level that matters, but shoppers who land on Rust Medium instead of Rust Small may see the wrong photo depending on how the theme resolves the fallback.

Finally, consider that same shirt where two of the eight colors were never photographed at all. Those two colors are the actual failure. A shopper picks Olive, sees Navy, and either leaves or orders while unsure. Two unphotographed colors will cost you more money than thirty unphotographed sizes ever will.

So the tool reports all three states separately, and it never counts a Size null in the problem totals. Being straight with you about this is the whole credibility of the page. A tool that inflates its own findings to look more useful is just a lead magnet with a progress bar.

The coverage states this tool reports

For each product, the scanner looks at the options that carry visual meaning and groups the variants by their value. Then it labels each group. Only the first two states below are counted as problems.

Single variant products deserve a note of their own. Shopify gives them a hidden option called Title with the value Default Title, and their one variant almost always has a null featured image. That is completely normal behaviour, the product simply shows its gallery, and the scanner excludes those from the problem counts and tells you how many it excluded.

How to read the per option breakdown

The option table is the part you should screenshot and send to whoever owns your product data. Each row is one option name found in your catalog, with the number of products using it, the number of variants sitting under it, and then the columns that actually decide anything: how many distinct values of that option exist, how many of those values have no photo at all, and what share of them are photographed. Products that come in only one value of an option sit outside those three columns, in line with the state table above, because a value the shopper cannot switch away from cannot show them the wrong thing.

Read the value columns, not the variant columns. A raw null count cannot tell a Color option apart from a Size option, because on a product with both, every variant sits under both, so both rows carry the same number. Counting values fixes that. Eight colors with two of them unphotographed is 75 percent value coverage. Five sizes sharing the one photo assigned to Small is 20 percent, and that 20 percent is fine.

Each row also carries a type label. Visual means the option changes what the product looks like, so a value with no photo there is a real gap. Size like means the option changes a measurement or a quantity, so sharing a photo is the expected and correct setup. Unclassified means the option name did not match either list, and you should apply your own judgment.

The classification is a heuristic based on the option name, nothing more. It reads names like Color, Colour, Pattern, Style, Finish, Material, Print and Fabric as visual, in several languages, so Farbe and Couleur are recognised too. It reads Size, Length, Width, Weight, Capacity, Quantity, Pack and Voltage as size like. When a name contains a word from both lists, the one further to the right wins, because in an option name the last word is the one doing the work. Print Size, Frame Size, Strap Length and Fabric Weight are all measurements, so they are read as size like, while Frame Colour and Colorway stay visual. Anything it does not recognise lands in the unclassified bucket.

Unclassified options are still analysed for missing photos, on the assumption that a color option in a language we did not anticipate should not slip through silently. What they are not is counted as the real problem. They get their own line in the summary, outside the headline figure, and every value they produce is tagged as unclassified in the worst products table so you can go through them one at a time. Nobody's catalog uses vocabulary a keyword list can fully predict, so the tool shows its working rather than pretending to be certain.

One caveat on the arithmetic. A variant that belongs to both a Color option and a Size option is counted in the variant column of both rows, because it genuinely sits under both options. That means the variant columns do not sum to your total variant count, and the tool says so directly under the table rather than hoping you do not notice. The value columns do not have that problem, since a value belongs to exactly one option.

So compare value coverage across the rows. If your Color row shows 90 percent of its values photographed and your Size row shows 20 percent, your catalog is in good shape and the headline percentage was lying to you. A Color row down at 40 percent is the one to act on, whatever the Size row says.

Products with fewer images than variants

This list answers a question the coverage percentage cannot. Coverage tells you what was assigned. Image count tells you what exists to assign.

A product with 30 variants and 4 photos cannot reach full per variant coverage no matter how carefully you click through the admin. The photos are not there. You either shoot more, or you accept that variants will share, and if they share you want them sharing sensibly, which usually means one photo per color value rather than one photo for the entire product.

The list only includes products that vary by an option the scan reads as visual, or one it could not classify, and within those it is sorted by the size of the shortfall. A shoe listed in one color and thirteen sizes with five photos is not short of anything, so it stays out of the table and gets counted in a footnote instead. The products at the top are the ones where the gap between what shoppers can select and what they can actually see is widest, which makes them the products where a shopper is most likely to be looking at a photo of something they did not pick.

Use this list to plan a photo shoot, not to plan an app install. We build variant image software and we will still tell you that no software conjures a photograph that was never taken. Assignment tooling solves assignment problems. Missing photos are a photography problem.

Worth pairing this with the Variant Image Calculator, which works out how many photos a given option structure actually needs before you book the studio.

How to fix the gaps once you have found them

Start with the unphotographed values, because those are the ones where the shopper is looking at the wrong product. Work down the worst products list in order. It is already ranked by how many variants sit under a value that has no photo.

The manual route. Open the product in Shopify admin, scroll to Variants, click a variant, and use the image picker to attach the right photo. Repeat for each variant. On a product with a handful of variants this takes a minute. It works, it costs nothing, and for a small catalog it is genuinely the right answer.

Where manual stops working. The arithmetic turns against you quickly. A catalog of 300 products with 8 colors and 4 sizes each is 9,600 variant assignments, and every new colorway you add reopens the job. That is the point where merchants start looking for tooling, and it is the honest threshold rather than a scare tactic.

The app route. Rubik Variant Images, built by Craftshift, assigns a whole image set to each variant instead of Shopify's single featured image, so picking Rust filters the gallery down to the Rust photos rather than swapping one hero shot. It loads through Shopify metafields with no external API calls. It also has two different bulk paths that people mix up constantly, so here is the distinction in plain terms: AI auto assign works on one product at a time and reads four data points, the product name, the variant name, the image filename and the image alt text, to decide which photo belongs to which variant. Bulk assign is a different mechanism entirely, grouping by the order of images in your Shopify gallery using featured image boundaries, with no AI involved. On the App Store it sits at 5.0 stars across 420 ratings as of August 2026, and it is free to install.

You can see variant filtering running on a live storefront in the Rubik demo store, or read the getting started guide before you install anything.

The case where no variant tooling helps. If your colors are separate products rather than variants of one product, variant image assignment is the wrong lever, because there are no variants to assign to. That is a product linking problem, and the Separate Products vs Variants comparison walks through which structure fits which catalog.

What the scan can and cannot see

The public product feed is generous but not complete, and pretending otherwise would make every number on this page suspect. Here is the honest boundary.

• It sees published products only. Drafts, archived products and anything hidden from the Online Store sales channel are absent from the feed, so they are absent from the audit too.
• It sees the default market. The feed returns products as your primary storefront serves them. Region specific catalog differences will not show up.
• It cannot see a password protected store. If the storefront is behind a password, the feed is unreachable and the scan fails with an error rather than a fake result.
• It cannot always reach headless storefronts. A custom frontend may not expose the endpoint on the domain you typed, though the underlying myshopify.com domain often still does.
• It stops at the page limit. Eight pages of 250 products is the ceiling. Larger catalogs get scanned up to that point and clearly labelled as partial.
• It does not judge photo quality. A variant with a blurry, badly cropped or mislabelled photo counts as covered. Coverage is about presence, not quality. For quality checks use the Product Image Audit.

Also note that fetched pages are cached briefly on our side to stay polite to the store being scanned, so if you fix something and rerun the check within a few minutes you may see the previous result. Wait a little and run it again.

Common mistakes when auditing variant images

Six patterns come up again and again in the support questions we get about variant imagery. Most of them are audit mistakes rather than store mistakes.

• Treating the headline percentage as a grade. A store selling shoes in one color and fourteen sizes will always score low, and there is nothing wrong with it. Read the option table, not the big number.
• Assigning one photo to a whole color and calling it done. If Blue Small has the photo and Blue Medium does not, the theme may fall back to the product's first image for Medium. Assign across the full size run of each color.
• Fixing the newest products first. Your traffic is concentrated in a handful of products. Fix the ones people actually land on, then work backwards. Cross reference the worst products list with your analytics before you start clicking.
• Assuming the theme does variant switching at all. Some themes ignore the variant featured image entirely, or only apply it to the main gallery and not to the product card. Assignment without theme support changes nothing on the page. Check with the Theme Compatibility Checker first.
• Shooting new photography before checking image counts. Many catalogs already have the photos uploaded and simply never assigned them. Look at the fewer images than variants list before you book anything.
• Auditing once and never again. Coverage decays. Every import, every new colorway, every product duplicated from a template reintroduces gaps. Put it on a monthly calendar reminder and the number stays healthy.

Who this tool is for

Merchants use it as a pre season check before the catalog doubles in size. Run it in September, fix what it finds, and the peak season traffic lands on product pages that show the right thing.

Agencies use it during onboarding. Running a new client's store through the scanner before the discovery call turns a vague conversation about product presentation into a specific list of products with names and numbers attached. It is a fast way to show you did the homework.

Developers and app builders use it to reproduce merchant reports. When someone says the wrong image shows on their product page, the first question is whether the variant even has an image assigned. Often it does not, and the theme is behaving exactly as designed.

And anyone doing competitive research can point it at a competitor. It only reads public data, the same data any shopper's browser downloads, so there is nothing private about it. Seeing how a large store in your category handles variant imagery is a decent benchmark for how much effort the category expects.

Related Tools

• Variant Image Checker, the single product version of this audit with a full per variant table.
• Variant Image Calculator, works out how many photos your option structure needs before you shoot.
• Product Image Audit, a ten point checklist covering alt text, filenames and image quality.
• Variant Combination Calculator, how many variants a given option structure produces and where it hits the ceiling.
• Collection Analyzer, product counts, variant counts and grouping potential for one collection.
• Store Analyzer, a broad overview of any Shopify store including products, theme and pricing.