Running a complex Amazon account: an operating model for large catalogues

A complex Amazon account is rarely a hard one. It is usually an unmanaged one. Once a catalogue passes a few hundred SKUs, or adds a second marketplace, a Vendor Central account or a TikTok Shop, nobody can keep it in their head. Without a system, the account sets the agenda: mornings go on suppressed listings and stockouts, and the work that grows the business waits.

What replaces memory is an operating model: a role for every product, an owner for every kind of change, a short list of numbers with a definition of “normal” for each, and a fixed rhythm for looking at them. This is the model we run on large accounts, section by section.

Run the account by exception

You cannot look at every SKU every week, and you don’t need to. The job of the operating model is to show you the handful of products that are broken or drifting and hide the hundreds that are fine.

So the model has one design rule: for every metric, decide what normal looks like, then only look at the products outside it. A product holding the Featured Offer, converting at its usual rate, with stock cover inside its band and a positive contribution after advertising, needs nothing from you this week. Everything below is about deciding what “normal” means and making the exceptions visible.

Complexity also multiplies rather than adds. On an account that also runs Vendor Central or TikTok Shop, one stockout can mean a missed purchase order, a channel taking orders it can’t fill and ad spend still sending shoppers to a page that can’t sell. So the rhythm later in this guide checks the connections between areas, not just each area.

Segment every SKU by profit and role

The first mistake in a big catalogue is treating every product the same: the same listing effort, the same advertising targets, the same stock policy. The second is ranking products by revenue. Revenue hides Amazon’s fee stack, so the products that sell most are not always the products that earn most.

So rank the whole catalogue by contribution per SKU over the last 12 months, then give each product one of four roles. The roles are our practice, not an Amazon category.

Two rules make the roles work. First, a product’s role decides its advertising target: a hero can run up to its break-even ACoS, a core product should stay well inside it, and a tail product should rarely pay for clicks at all. Our guide to break-even ACoS shows how to work out the number for each product, and our Amazon PPC strategy shows how to split budget by margin. Second, tests graduate or stop. A test that has not earned a core or hero role after its agreed period moves to the tail or is discontinued.

To see how concentrated your sales are before you start, download the Detail Page Sales and Traffic by Child Item Business Report and run it through the Deep X-Ray in our free Profit X-Ray. It shows how many listings make 80% of your sales and what share your top five carry, and flags listings with traffic but no sales and listings losing the Featured Offer. The file is read in your browser and never uploaded.

Catalogue governance: one source of truth, owned fields, controlled change

Big catalogues rot quietly from the data up. Attributes go stale on re-uploads, titles get overwritten by a feed, a compliance field goes blank and a listing drops out of search. None of it shows in the revenue line until it has been costing money for months.

The first reason is structural. On a shared product page, Amazon builds the listing from several contributors. It “automatically combines the best product data from various seller submissions”, weighing sales volume, refund rates, customer feedback and A-to-z claims⁹. For brand-registered products, contributions from sellers with a brand representative role are the ones displayed⁹, which is one of the practical reasons to enrol in Brand Registry¹¹.

That leads to three rules we apply on every large account.

Keep the master copy outside Seller Central. Seller Central is where you publish product data, not where you store it. Keep every SKU’s title, bullet points, item highlights, attributes, images, compliance documents and pack data in a product information system or a disciplined master spreadsheet. When a live listing looks wrong, compare it with the master and you know exactly what changed.

Edit in bulk, with a log. Amazon itself recommends the product spreadsheet rather than the single-listing form for variation families with more than 250 child ASINs⁶, and the same logic applies to any large edit. Each uploaded file, kept with its date, becomes your change history. Edits made one listing at a time in the browser leave no trail.

Give every field an owner. Not every person or system should be able to change everything.

On a small team one person may hold several of these seats, and with an agency some sit with us. What matters is that each seat exists, so that when something breaks nobody spends a week asking whose job it was.

Attribute completeness belongs in the same discipline: every blank field is a question the page can’t answer and a filter the product can’t match. Our listing optimisation guide covers the content rules, including the 75-character title limit and the new item highlights field from 27 July 2026, and our Rufus guide covers how Amazon’s shopping assistant reads the page.

Variation strategy at scale

A variation family groups sizes, colours or other options under one product page. Done well, it concentrates traffic onto one strong page. Done badly, it buries the best seller among dead options or breaks Amazon’s rules.

Amazon UK’s rules are short. Every product in a family must be the same brand and product type; Amazon says variations “must be functionally identical products"⁶. You must use an approved variation theme for the product type and must not put more than one characteristic into one variation attribute⁶. The parent is a non-buyable container with no price or stock, and each child needs its own product identifier⁶. Amazon reserves the right to remove families or children that don’t comply, and families with more than 4,000 child ASINs are not displayed on the detail page⁶.

The bigger change in 2026 concerns reviews. Amazon announced that from 12 February 2026 reviews would only be shared between variations “with minor differences that don’t affect functionality"⁸. Amazon’s UK help now states the same rule and says you will receive an email 30 days before any change affects your products⁷. In practice, families split by colour or size usually keep shared reviews, while families split by flavour, formulation or specification may now show separate review counts on each child.

For a large catalogue that means three jobs:

1. Audit families that group genuinely different products. If a newer variant relied on a sibling’s reviews, plan for it to stand on its own: its own launch budget, its own review requests, its own conversion work.
2. Remove or fix dead children. A suppressed or permanently out-of-stock child clutters the shopper’s choice and the stock plan. Move it out or fix it.
3. Choose the hero child deliberately. Set the main image and the order of options so the shopper lands on your best-converting variant, not on whichever child was created first.

Featured Offer control

The Featured Offer, Amazon’s current name for the Buy Box, is the offer shown with the Add to Basket and Buy Now buttons². On a large account it is won or lost per child ASIN, so a family can look healthy at parent level while one child has lost it. This section covers what a catalogue owner needs; our full guide to the Amazon Buy Box explains the selection factors, the July 2026 change and how to win it back in detail.

Amazon changed the rules this year. On 20 July 2026 it removed the seller eligibility requirements for the Featured Offer for sellers in the UK and EU⁴, and its UK help now says offers from all sellers are eligible to be evaluated³. Selection did not change⁴. Amazon ranks offers on:

• Price: the total of item price and delivery charge².
• Delivery promise: faster offers, and offers with a specific date rather than a range, are more likely to be featured².
• Free delivery: offers with free delivery are more likely to be featured².
• Seller performance: including order volume and chargeback rate².

Amazon says it does not consider customer reviews, Prime eligibility or fulfilment method when other factors are equal². An offer that is out of stock can’t be featured, and Amazon may feature no offer at all if none meets its requirements, even when there is only one offer on the page².

The number to watch is Featured Offer percentage in the Business Report: the share of page views where your offer held the Add to Basket button. It falls when you are out of stock, when another seller wins, or when your offer isn’t available through the Featured Offer, and it is reported by SKU and ASIN¹.

Now that every seller’s offer is evaluated, resellers who used to fail the eligibility step can compete on price and delivery for your listings. That makes price control more important, and it is a distribution job, not an Amazon one. Amazon states that breaches of exclusive or selective distribution agreements are not intellectual property infringement and that it would not be appropriate for it to help enforce them¹⁰. Control who can buy your products wholesale, write your pricing terms into reseller agreements, and keep Amazon’s brand protection tools, such as Transparency and Project Zero¹¹, for counterfeits. Our guide to Amazon Transparency and Project Zero explains when each one pays.

Account health

Account health sits above everything else in the model, because a deactivated account earns nothing.

Amazon’s Account Health Rating runs from 0 to 1,000 and reflects your policy record over the last 180 days. New sellers start at 200. A score of 200 to 1,000 is “Healthy”, 100 to 199 is “At Risk” of deactivation, and 99 or lower means the account is eligible for deactivation or already deactivated⁵. Violations carry points by severity (critical, high, medium or low), repeat violations of the same policy count for more each time, and for higher-severity policies as few as two repeats in 180 days can put the account at immediate risk⁵. Amazon may also deactivate immediately, whatever the score, if it suspects fraudulent, deceptive, illegal or harmful activity⁵.

Large catalogues are more exposed simply because there are more listings to break a rule: a missing safety document, a restricted claim in a bullet point, an image that fails a check. So the model treats account health as a daily 10-minute check by one named owner, and treats any new violation as that day’s first job, not the weekly meeting’s. If an account is deactivated anyway, our guide to appealing an Amazon account suspension covers the plan of action.

The rhythm: weekly, monthly, quarterly

Metrics without a rhythm become a dashboard nobody opens. This is the cadence we run; adjust the times to the size of the account.

The monthly review is where Brand Analytics earns its place. The Search Query Performance dashboard, available to brand representatives in Brand Registry, shows impressions, clicks, basket adds and purchases by search query¹², which tells you whether a hero’s slowdown is lost visibility or lost conversion.

The quarterly review is where the tail gets cut. Discontinue deliberately: sell stock through rather than paying for it to sit in storage, point real demand at a sibling you are keeping, and keep any product with a genuine strategic reason, such as a line a key trade customer expects. Our inventory management playbook covers cover bands and storage costs in detail.

The reporting that matters: profit per SKU

If you change one thing in your reporting this year, report contribution per SKU instead of sales per SKU.

A useful monthly report fits on one page:

• Contribution by role, with the share of the account’s total each role earns.
• The top 20 SKUs by contribution, next to their rank by sales. The gaps between the two lists are where the decisions are.
• SKUs with negative contribution after advertising, with the reason: price, fees, returns or ad spend.
• Revenue concentration: how many SKUs make 80% of sales, and how that is moving. Buyers ask the same question when they value the business: our Amazon business valuation calculator and our guide to selling an Amazon business show why.
• Featured Offer percentage and conversion (Unit Session Percentage) for heroes, both from the Business Report¹.

Sales still matter, because rank follows them. But a record sales month in which a third of the catalogue loses money is not a good month.