Most sellers treat product images as a launch task. You shoot the product, resize it, upload it, and move on. But a catalogue is a living thing: suppliers send replacement photos, staff upload phone shots in a hurry, marketplaces quietly tighten their specs, and a store migration silently breaks a few hundred image URLs. Two years later, nobody actually knows what is in the catalogue.

That used to be survivable. In 2026 it is not. Amazon runs periodic compliance sweeps that suppress listings which have been live and selling for months, Google Merchant Center has been flagging undersized images since April 2026 ahead of a hard 500 x 500 minimum on 31 January 2027, and marketplace moderation is increasingly automated rather than complaint-driven. The images that pass today are not automatically the images that pass next quarter.

This is a practical, repeatable audit you can run over 50 SKUs or 5,000 without opening every file by hand.

What actually goes wrong in a catalogue over time

Before you audit, it helps to know what you are looking for. In real seller catalogues, the recurring problems are remarkably consistent:

None of these announce themselves. They surface as a suppressed listing, a disapproved feed item, or a slow bleed in conversion that never gets attributed to images.

Step 1: Export an inventory of every image you have

You cannot audit what you cannot list. The goal of this step is a single spreadsheet with one row per image: SKU, listing title, image position, and image URL or file path.

On Shopify, a standard product CSV export from Products, then Export, includes the image source URL and image position columns alongside the handle and title. That export is not an image report, but it is the raw material for one. On marketplaces, use the seller centre bulk product export, which typically includes image URL columns for the main image and each gallery slot. If you keep masters locally, run a directory listing instead and treat the folder as the source of truth.

Two things worth capturing while you are here: how many images each listing has, and which listings have exactly one. Single-image listings are almost always underperforming, and they are the cheapest fix in the whole audit.

Step 2: Run the technical checks

Five machine-checkable faults account for most rejections. Work through them in order, because each one is cheaper to fix than the next.

Broken and corrupt files

Fetch every image URL and record the HTTP status. Anything returning 404, 403, or a redirect to a login page is invisible to the marketplace crawler, which is exactly how feed items get disapproved for an unreachable image. Also check that the file is complete: ImageMagick's identify command reports when an image is truncated or corrupt, which is a common outcome of an interrupted bulk upload.

Pixel dimensions

This is the single highest-yield check. Running identify -format "%w x %h %f\n" across a folder gives you width, height and filename for every file in one pass, which you can sort to find everything below your threshold in seconds. Filter for anything under 1,000 px on the longest side first, then anything under 1,600 px.

Aspect ratio

Flag anything that is not the ratio your primary channel expects. Most e-commerce marketplaces want 1:1, and a 4:3 image pushed into a square slot gets letterboxed or centre-cropped, which is how products end up with their tops sliced off in search results.

File size and format

Look for two opposite failures: files so heavily compressed that they show visible artefacts at zoom, and 12 MB uncompressed exports that slow your product pages down. Also check for formats a channel will not accept, and for HEIC files that slipped in from an iPhone.

Colour space

Any master saved in Adobe RGB or CMYK will render dull or shifted in a browser. Identify reports the colour space, so this is a bulk check rather than an eyeball one. Everything destined for the web should be sRGB.

Current minimums versus safe master sizes

Audit against the safe master column, not the minimum. Meeting the bare minimum means re-auditing every time a platform moves the goalposts.

ChannelStated minimumSafe masterNotes
Amazon1,000 px longest side (enables zoom)2,000 x 2,000Pure white main image (RGB 255,255,255), product fills 85% of frame, max 10,000 px
ShopifyNo hard minimum2,048 x 2,048Square masters keep theme grids consistent
Shopee500 x 500, 1:11,024 x 1,024 or higherNo watermarks, borders or added text; white strongly preferred
Lazada500 x 500, 1:11,000 x 1,000 or higherWhite background for the main image
Google Merchant Center100 x 100 (250 x 250 apparel)1,000 x 1,000Rises to 500 x 500 for all products on 31 January 2027; no promotional overlays

The Merchant Center change is the one to plan for now. Warnings labelled "image too small for upcoming enforcement" have been appearing in accounts since April 2026, and a catalogue full of 300 px thumbnails will not be fixable in a weekend.

Step 3: The checks a script cannot do

Once the technical faults are cleared, sample the catalogue visually. You do not need to review everything: take the top 50 SKUs by revenue, plus a random 5% of the rest, and check for background consistency, frame fill, text or badges burnt into main images, variant images matching the variant they are attached to, and packaging that still matches what you actually ship. Products shot against a different backdrop or at a different distance from the rest of the range make a catalogue look assembled rather than curated.

Step 4: Triage before you fix

Sort every fault into one of three buckets. Reprocess covers anything where the pixels are good enough and only the container is wrong: resizing, padding to square, converting format, converting colour space, stripping metadata. Reshoot covers genuinely low-resolution originals, wrong products, and images where the background cannot be salvaged. Accept covers faults on discontinued or near-dead SKUs, which are not worth the labour.

Be honest about the reprocess bucket. Upscaling a 400 px supplier JPEG to 1,600 px satisfies the dimension check while producing a soft, mushy image that fails the quality check instead. If the detail is not in the original, it cannot be added.

Step 5: Re-export everything at spec, in one pass

The mistake at this stage is fixing images one at a time in whatever editor is open. Set your target specs first, then process the whole reprocess bucket in a single batch so every output shares the same dimensions, padding, background and compression settings. Batch tools built for this, such as PixelPrep, will take a folder of mixed masters and produce marketplace-ready sets per channel, which is what keeps a 2,000-image fix from becoming a fortnight of manual work.

Keep the masters. Every future spec change is a re-export from masters if you have them, and a reshoot if you do not.

Step 6: Make it repeatable

Run the technical checks quarterly, and always after three specific events: a platform migration, a new sales channel going live, and any bulk import from a supplier feed. Save the export and check commands as a script so the next audit is an afternoon rather than a project.

Your catalogue audit checklist

  1. Export one row per image: SKU, title, position, URL or path.
  2. Flag listings with fewer than four images.
  3. Check every URL returns 200 and every file opens without corruption.
  4. List all images under 1,000 px, then under 1,600 px, on the longest side.
  5. Flag anything not in your primary aspect ratio.
  6. Flag files over 5 MB and files with visible compression artefacts.
  7. Confirm every image is sRGB and in an accepted format.
  8. Visually review top 50 SKUs plus a 5% sample for background, frame fill, overlays and variant accuracy.
  9. Triage into reprocess, reshoot, accept.
  10. Batch re-export the reprocess bucket to per-channel specs from masters.
  11. Re-upload, then confirm the listing quality or feed diagnostics dashboard clears within 48 hours.
  12. Diarise the next audit and save the scripts.

An image audit is unglamorous work that nobody asks for. It is also the only reliable way to find the listings that are quietly suppressed, the feed items that stopped serving, and the 300 px thumbnail that has been costing you conversions since 2024.