Keeping thousands of catalogue images consistent across dozens of suppliers

Write the house standard down first

Before triaging a single image you need a target to triage against. A house standard is short and specific: background, aspect ratio, padding around the product, minimum pixel dimensions, file format, colour profile, and rules about branding, watermarks and text in the image. Everything downstream, what you accept, what you fix, what you reject, follows from it.

Be realistic about the standard you can actually hold. A pure white background at a fixed square ratio with ten percent padding is achievable for most catalogue products and is what marketplaces and search results tend to expect, but it is unachievable for a five-tonne machine photographed in a factory. Define a primary standard for the majority of items and a documented exception for large equipment shot in situ, rather than pretending one rule covers a range that spans washers and bandsaws.

Include the things nobody writes down: whether the product is centred or offset, whether it faces left or right, whether cables and hoses are coiled or extended, whether an item is shown with or without its packaging. These conventions are what make a category page look deliberate, and they are invisible until one supplier's images break them.

• Background: colour or value, and whether shadows are allowed
• Aspect ratio and minimum resolution for the primary image
• Padding: how much clear space around the product
• Orientation convention: which way the product faces
• Branding: no supplier logos, watermarks, borders or burned-in text
• Colour profile and file format for handoff to the web platform
• Exception rules for large equipment and in-situ photography

Triage into three piles, not two

Every supplier image lands in one of three categories, and the trap is trying to make it two. The first pile is usable as-is: correct background, adequate resolution, clean of branding, close enough to the standard that nobody notices. The second is normalise: fundamentally sound photography that needs the background evened out, the exposure corrected, the crop and padding adjusted, or a watermark-free version requested from the supplier. The third is reshoot: too low resolution to rescue, wrong product variant, obviously wrong colour, or so far from the standard that fixing it costs more than photographing the item.

Set an explicit resolution floor and use it as the first gate, because no amount of editing adds detail that was never captured. An image below the floor goes straight to the reshoot or request-again pile regardless of how good it looks at thumbnail size, since it will fail the moment a customer zooms in on a product page.

Triage by category rather than by supplier. Customers compare products within a category, so the inconsistency they actually notice is between two bearings on the same page, not between a bearing and a hose reel three clicks away. Working category by category also lets you finish something visibly rather than spreading effort evenly across a catalogue that never looks finished.

The licensing question nobody asks until it is a problem

A manufacturer sending you a product image is not the same as a manufacturer granting you a licence to use it, modify it and publish it. Many distribution agreements include marketing asset provisions, many do not, and some manufacturers explicitly prohibit modifying their photography or removing their branding from it. The right to edit is a separate question from the right to publish, and both vary by contract and jurisdiction.

Ask for the terms in writing when you onboard a supplier, alongside the data feed and the price file, rather than discovering the position when someone objects. Specifically ask whether you may crop, recolour the background and remove watermarks, because those are exactly the operations normalisation requires. Where a supplier's image includes third-party content, another brand's component, a recognisable person, a photographer's credit, the position gets more complicated and is worth checking properly.

This is legal territory that differs significantly between markets, so treat the above as considerations rather than answers and take advice from someone qualified in your jurisdiction. What is universally true is that the conversation is cheaper at onboarding than it is after ten thousand product pages are live.

Normalising without altering the product

Normalisation is the largest pile and the one where discipline matters most. The permitted operations are the ones that change the photograph rather than the product: replacing or evening out a background, correcting exposure so a dark image matches a bright one, correcting a colour cast from the supplier's lighting, straightening, recropping to the house padding and ratio, and removing dust or bench clutter.

The forbidden operations are the ones that change what a buyer would specify. Do not adjust a product's colour beyond correcting a lighting cast, because colour frequently denotes a material, a grade, a pressure rating or a hazard classification. Do not remove a fitting, port, cable, bracket or fastener that ships with the item. Do not smooth a cast surface into a machined one, or make a painted finish look plated. Do not change proportions to fit a crop; a hose photographed slightly narrower to fit a square frame is now a hose that is not what the customer receives.

The hardest judgement is a supplier image where the product genuinely looks wrong, a grey casting rendered blue by fluorescent light, for instance. Correcting the lighting cast to neutral is legitimate; deciding what colour you think it should be and setting it there is not. If you are unsure, ask the supplier for the correct finish description and correct toward that, or move the item to the reshoot pile and photograph it yourself with a grey reference in the frame.

Doing the work at catalogue volume

Be honest with yourself about throughput before committing to a project plan. Editing tools that operate one image at a time are fine for a hundred hero products and are not a strategy for forty thousand SKUs. Foxi AI is in that category: it edits a photo you already have from a plain-language description, everyday looks cost about five credits and results typically return in under a minute, but there is no batch mode or API, so every image is an individual pass. That makes it well suited to high-value or high-traffic products, a category's top sellers, or the specific images that let you launch a new range, and unsuited to a full-catalogue sweep.

Plan accordingly: pick the images that carry the most weight rather than trying to normalise everything. The first hundred products in your best-selling categories, the items that appear in search results most often, and anything you are actively promoting will change how the catalogue reads far more than the long tail will. The tail can be improved as suppliers refresh their assets.

For the reshoot pile, build a small in-house setup and a fixed procedure rather than commissioning ad hoc photography. A bay with a consistent background, marked camera and product positions, fixed settings and a written frame list will produce catalogue images that match each other indefinitely, and it turns reshooting from a project into a routine task someone does for an hour a week.

Pushing the standard back up the supply chain

The cheapest image to normalise is the one that arrives correct. Send your house standard to every supplier as a one-page specification with example images, include it in onboarding and in annual reviews, and give suppliers a simple way to send updated assets. Manufacturers generally want their products presented well and often have better source files than the ones that reached you through three intermediaries.

Ask specifically for unbranded, unwatermarked, full-resolution originals rather than the versions on their public website, which are usually compressed and cropped for their own layout. This one request often solves more problems than any amount of editing, and it costs nothing but an email.

Do not promise anyone, internally or externally, that consistent imagery will lift sales or search performance. Nobody can demonstrate that reliably, and the honest case is simpler: a coherent catalogue lets a customer compare products on their merits instead of being distracted by mismatched pictures, and it stops your own listings from looking like they were assembled by accident.

Frequently asked