Fixing a Marketplace Catalogue in the Right Order: Identifiers, Words, Content, Then Price

A catalogue that has grown for years without an owner usually has every problem at once: duplicate pages, missing attributes, titles written for nobody, images from three different shoots, and products nobody has checked the margin on in a year. The instinct is to start with what is most visible, the titles and images. The faster route is a fixed order, because each step depends on the one before it.

First make sure every product is identified correctly and selling on the right page. Then find the words buyers use for each kind of product. Then write the content around those words. Then extend the work to search outside the marketplace. And then, with real traffic and conversion to read, decide what each product should cost, how it should be fulfilled, and whether it should stay. Each step done out of order is usually done twice.

Content written for the wrong page is wasted. Before any copy is touched, check that each product has the right barcode, sits on one detail page rather than two, belongs to the right variation family, and is not suppressed. Duplicates split a product's reviews and sales history; a child listing in the wrong family confuses shoppers and search alike; a barcode registered to somebody else can remove a listing months later. The codes and their traps are in ASINs, barcodes and product identifiers.

This is also when suppressed and inactive listings get cleared, because a listing that is invisible cannot benefit from anything done after it. How to find the reason for each and clear it is in the search suppressed listing guide.

The output of this step is a register: one row per sellable product with its identifiers on every channel, its status, and its variation family. Everything after it works from that register, which is why it comes first even though nobody outside the business will ever see it.

The first step goes faster when the problems are pulled into one list rather than found one listing at a time. Seller Central already keeps most of that list; it is spread across a few places.

Suppressed listings. The inventory view can be filtered to listings Amazon has removed from search, and each one carries the reason and the field that needs fixing. The same list downloads as the Listings Quality and Suppressed Listing Report, with a column for the corrected value, so a catalogue with dozens of suppressions can be fixed in one file and uploaded back rather than edited page by page.

What the catalogue actually says. The Category Listings Report shows each product as Amazon's catalogue holds it now, including values another seller or Amazon itself contributed. It is not switched on for every account: Seller Support activates it on request, for categories where the account has active listings. It is the quickest way to see titles, bullets and attributes that no longer match what the brand uploaded.

Duplicates and broken families. Search the catalogue by barcode and by brand name: two pages for one product usually show up as two results with the same image. Then read each variation family on its own page, checking that every size and colour sits under the right parent and that no child is orphaned on a page of its own.

Put everything into one sheet, one row per product, with the problem and the fix. That sheet becomes the plan for the rest of the work and the record of what was done.

With the products settled, find the language. Research is done by kind of product rather than by listing: the phrases shoppers use, the questions they ask, the attributes they filter on, and the terms that already convert in the advertising. A catalogue of two hundred products often has only fifteen or twenty kinds, so the research is smaller than it looks. Where each kind of phrase belongs on a listing is in Amazon keyword research.

Doing this before writing matters because the words decide the structure: which attribute leads the title, which question earns a bullet, which use deserves an image. Copy written first and researched afterwards gets rewritten.

Now the visible work: titles, bullets, descriptions, the attribute fields, images and A+ content, each written from the research for its kind of product. Work through the catalogue by revenue, starting with the products that already sell, because an improvement on a product with traffic shows within weeks, while one on a product nobody sees shows nothing. What moves rank on a listing, and what only looks like it should, is in Amazon listing optimisation.

Use templates per kind of product rather than writing each listing from nothing. A template says what the title contains and in what order, what each bullet answers, and which image goes in which slot. It keeps the two hundredth listing as good as the first, and it lets the work be shared without the catalogue reading like five different authors.

Images follow the same order. Fix the main image first on the products that sell, then the slots that answer the objections the reviews raise; a full reshoot waits until the content work has said what the shots must show.

Not every listing needs the full treatment. Products that sell little and have no search demand behind them can keep tidy basics: correct identifiers, complete attributes, a clean main image and an accurate title. A writer's day spent on them is better spent on the products that carry the business. Revisit them once a year, or sooner if one starts to sell.

The steps need different people. Identifiers and listing health need somebody patient with Seller Central and flat files. Research needs somebody who reads search data and customer questions. Content needs a writer and a designer working from the same template. Store search needs somebody who understands crawling and structured data, and the economics need somebody comfortable with fee reports and margins. One person rarely does all five well, which is one reason catalogues stay neglected.

What matters most is that one person owns the order: decides when a product moves to the next step, keeps the register current, and stops the visible work starting before the foundations under it are done.

A brand with its own store applies the same research there, in a different shape: collection pages that answer a category, product pages written in the buyer's words rather than the supplier's, and structured data that states the facts plainly. The two disciplines share the research and little else, as eCommerce SEO and marketplace optimisation explains.

It comes after the marketplace work because the marketplace usually carries the sales, and because the research and the product register built there make the store work faster. A brand whose store carries most of its sales should reverse those two steps; the order follows the revenue.

The order holds on every marketplace, because every marketplace has to know what a product is before it can show it to the right buyer. What changes is where the problems are reported and which part of the content carries the most weight.

Walmart. Seller Center's Listing Quality dashboard scores each published item and breaks the score into parts: content and discoverability sit alongside price, shipping, and ratings and reviews. Missing attributes for the item's category pull the content score down, which makes the dashboard a ready made list for the first and third steps. A bulk attribute editor lets the same attribute be filled across many items at once, which matters on a catalogue of hundreds. The identifier step is the same as on Amazon: an item matched to the wrong listing, or two listings for one product, has to be settled before any copy is rewritten.

TikTok Shop. Seller Center rates each product's listing quality in tiers, from Poor to Excellent, with a diagnosis of the title, images, attributes and description and suggested fixes for each. The research step reads differently there, because most buyers arrive from a video rather than a search box: the words to use are the ones creators and buyers use in the videos and comments, and the content step includes the short videos and images that a product page leads with.

A brand selling on all three can run the steps once across the catalogue: identifiers and families settled everywhere, one piece of research per kind of product, then content written for each marketplace's shape from the same research.

With the catalogue identified, described and searchable, each product finally has honest numbers: real traffic, real conversion, real advertising cost. That is when to decide price, fulfilment method and whether a product stays, product by product, on the margin left after every fee, the advertising and the returns. The fees and where each appears are in Amazon seller fees explained.

There is one exception to doing this last. Before spending effort on a product's content, check quickly that it can make money at all at its current price. A product that loses money on every sale should not be polished first; it should be repriced, refulfilled or dropped, and the content time spent on one that can earn it back.

Most catalogue cleanups that fail do the right work in the wrong order, or undo it on the way. The same few mistakes come up again and again.

Rewriting a page that is about to be merged. When two pages for one product are merged, one of them goes. Copy written for that page is lost, and so is the time. Settle duplicates first, then write for the page that stays.

Uploading a full file over live listings. A full update in a listings file replaces the whole record, so a blank cell can clear a value that was right. A partial update changes only the fields sent. For a cleanup, send partial updates unless the aim really is to replace everything, and test the file on a handful of products before the rest.

Changing the product type or the barcode to fix a category. Changing what a page is, rather than what it says, can break it or split its reviews from the product. A product in the wrong category is usually a case for Seller Support, not an edit.

Fixing a page the brand does not control. Without Brand Registry, the brand is one contributor among several, and another seller's values can win on the detail page. Registering the brand first makes the rest of the work stick.

Judging it too early. A changed title can take days to show in search, and rank follows sales over weeks. Read the results against the same weeks before the work started, not against the day after.

The first step is the slowest relative to how little it shows, and the one most often skipped. The content step is the largest. The economics step is the quickest once the numbers are clean. A catalogue worked in this order shows its first results in the products that already sold, then in search coverage across the range, and last in margin.

Measure each step by what it was for: fewer suppressed and duplicate listings after the first, more search terms each product is found for after the second and third, sessions and conversion by product after the content, and contribution margin by product after the last. A single revenue number at the end hides which step did the work.

Each step is a service of its own in our commerce services: product listing optimisation, keyword research, eCommerce SEO and unit economics, worked in this order.