Step 1: Export your catalog into an auditable sheet
You cannot audit what you cannot see side by side. Before looking at a single image, export a row per SKU with the columns you will score against. Every major platform supports this: Shopify via the products CSV export or Admin API, WooCommerce via a product CSV export plugin, BigCommerce via its export template, Amazon via a Category Listings Report.
The minimum useful export:
| Column | Why it matters | Source |
|---|---|---|
| SKU / product ID | Join key for everything else | Platform export |
| Trailing 90-day revenue | Drives fix priority | Analytics / platform reports |
| Sessions and conversion rate | Finds high-traffic, low-converting listings | GA4 / platform |
| Image count | Coverage gaps | Platform export |
| Primary image URL | Lets you pull dimensions and file size programmatically | Platform export |
| Alt text present | SEO and accessibility | Platform export |
| Return rate | Strongest signal of misleading imagery | Orders / returns report |
Resolution and file size are the two fields most exports omit, and they are the two most mechanically checkable. A short script that walks every image URL, reads the dimensions and the byte size, and writes them back into the sheet turns the slowest part of the audit into a background job. If you would rather not write code, most image CDNs will report both in their asset dashboards.
Add a "last updated" date column if your platform exposes it. Images older than about three years are disproportionately likely to fail current marketplace specs, because the specs tightened and the images did not.
Step 2: Score every SKU against six dimensions
Scoring needs to be blunt enough that two people reach the same answer. Six dimensions, pass or fail on each, gives you a score out of six per SKU and a defensible ranking.
1. Resolution. Pass if the longest side is at least 2000px. This is the practical floor for zoom on desktop and the explicit requirement on Amazon and several other marketplaces. Anything under 1000px fails hard and should be flagged separately — it is almost certainly a supplier image.
2. Coverage. Pass if the listing has at least the category baseline: typically a hero on white, two to three alternate angles, one detail or texture shot, and one scale or in-context shot. Apparel and furniture need more; simple accessories need fewer.
3. Consistency. Pass if the hero image matches its siblings in the same collection on background, crop margin, framing angle, and color temperature. This is the dimension that fails most often and is hardest to see one SKU at a time — you need a contact sheet of thumbnails per collection to catch it.
4. Accuracy. Pass if the image color matches the physical product and the image shows what actually ships. Cross-reference SKUs with return rates above your catalog average; misleading imagery shows up in returns before it shows up anywhere else.
5. Compliance. Pass if the image meets the specs of every channel it syndicates to — no watermarks, logos, borders, props in the main image, or text overlays where prohibited, and correct background and fill ratio.
6. Technical hygiene. Pass if the image is served in a modern format at a sane file size, has descriptive alt text, and has a filename that is not IMG_4471.jpg.
The shape of that distribution is the useful part. Consistency and technical hygiene fail constantly and are cheap to fix in bulk. Accuracy fails rarely but each failure is expensive, because it is generating returns.
Step 3: Rank the fix list by revenue, not by score
A SKU scoring 2/6 that sells four units a year is not the problem. A SKU scoring 4/6 that carries eight percent of revenue is. Multiply the failure count by trailing revenue and sort descending — that column is your work order.
Three cohorts usually emerge, and they want different treatment:
Fix first
- High traffic, low conversion, failing coverage — visitors are arriving and not being convinced
- Any SKU failing accuracy with an above-average return rate
- Anything failing compliance on a channel that can suppress the listing
- Top 20% of revenue with any hero image failure
Batch later
- Consistency failures across a whole collection — fix as one job, not SKU by SKU
- Missing alt text and bad filenames — scriptable across the entire catalog
- Format and compression issues — a CDN configuration change, not an image change
- Long-tail SKUs under a revenue floor you set deliberately
Set that revenue floor explicitly and write it down. Most catalogs have a tail of SKUs where the cost of reshooting exceeds the lifetime margin, and pretending otherwise is how audits stall at 60% complete.
High-traffic, low-conversion SKUs are the most commonly misdiagnosed cohort. Before blaming the images, confirm the product is in stock, priced competitively, and shipping in a reasonable window. Images are frequently the cause, but they are not always the cause.
Step 4: Close the gaps without booking a studio
Most audit findings do not require new photography. Sorting fixes by what they actually need keeps the remediation cost proportionate to the problem.
| Finding | What it actually needs | Effort |
|---|---|---|
| Inconsistent backgrounds across a collection | Background replacement on existing shots | Low |
| Wrong crop margin or aspect ratio | Programmatic recrop with padding rules | Low |
| Missing alt text / bad filenames | Bulk generation and rewrite | Low |
| Oversized files, legacy formats | CDN format negotiation and resize presets | Low |
| Under 2000px on the longest side | AI upscaling if the source is clean; reshoot if it is soft or compressed | Medium |
| Color drift between shoots | Color correction against a reference SKU | Medium |
| No lifestyle or in-context shot | Generated scene from the existing product cutout | Medium |
| Image does not match what ships | New photography of the correct unit | High |
| No detail or texture coverage at all | New photography | High |
The split matters because the top half of that table is where the volume is. In a first audit, roughly two-thirds of failures are consistency, cropping, format, and metadata issues — all fixable from images you already own. Only accuracy failures and genuinely absent coverage force a camera.
For the middle band — background swaps, recrops, scene generation, catalog-wide color matching — batch tooling is what makes the numbers work. Platforms like Retouchable process a full collection against one set of rules, which is the only sane way to fix a 64% consistency failure rate without spending a month in Photoshop.
Step 5: Write the standard so the catalog stops rotting
An audit that produces a fix list but not a standard buys you about eighteen months before the same failures reappear. The deliverable that prevents the next audit is a one-page image spec, enforced at the point images enter the catalog.
The spec should be short enough that a supplier or a freelancer reads all of it. Cover: minimum and maximum dimensions, aspect ratio, background specification, product fill ratio within the frame, required shot list by category, file format and naming convention, alt text pattern, and color reference. One page, no prose.
Then attach it to the process:
- Intake gate. No SKU goes live without passing the six-dimension check. If your platform supports it, make image count and resolution a required field.
- Supplier terms. If suppliers provide imagery, the spec goes in the purchase agreement. Supplier images are the single largest source of resolution and watermark failures.
- Quarterly spot check. Re-audit the top 50 SKUs by revenue every quarter rather than the whole catalog annually. It catches drift earlier for a fraction of the effort.
- Re-audit on trigger. Run a fresh pass after a theme migration, a marketplace spec change, or a channel launch — those three events invalidate assumptions wholesale.
Catalogs that enforce an image spec at intake report noticeably fewer listing suppressions and far less emergency remediation before peak season. The audit is the one-time cost; the spec is what stops you paying it again.