Amazon advertising at portfolio scale: budgets, structure and bulk operations for hundreds of SKUs

A brand with 20 products can manage its advertising campaign by campaign. A brand with 500 cannot. Nobody can review 500 products’ search terms every week, so in a big catalogue the work moves up a level: deciding which products deserve which kind of advertising, how much each group may spend, and building a structure clean enough to change in bulk.

This guide covers that layer. For the ad types and what they cost, see our guide to Amazon advertising in the UK. For the campaign-level craft (match types, negatives, bidding strategies and placements), see our Amazon PPC strategy. For the operating model behind the whole account, see running a complex Amazon account.

Give every SKU an advertising role

In our operating model, every SKU has a role based on its contribution after fees and advertising: hero, core, tail or test. Advertising follows the role, not the other way round. These are our rules, not Amazon categories:

Two checks run underneath every role. No ads on products that cannot sell: out of stock, suppressed, or without the Featured Offer (our guide to the Amazon Buy Box explains why an offer loses it). And no ads on products that lose money on every sale once fees and advertising are counted, unless the loss is a deliberate launch investment. In large accounts we audit, the biggest single source of waste is rarely a bad bid. It is spend still flowing to products in one of those two states because a campaign built years ago was never switched off.

Allocate budget top-down by margin and role

Bottom-up budgeting, where each campaign gets whatever it asked for last month, rewards whatever spent most last month. Top-down budgeting starts from the total you are willing to spend, splits it by role, then by product group, and only then by campaign.

Each role also needs a target ACoS, worked out from margin rather than habit:

A hero range with a 32% margin that should return 6p in every £1 of ad sales has a target of 26%. Amazon Ads reports attributed sales without VAT⁶, so work margin out on the ex-VAT price; our break-even ACoS guide shows the method, and the FBA calculator for the UK gives each product’s margin after Amazon’s fees.

The plan is a starting point, not a forecast. Move budget between roles monthly, on profit, and only within a role weekly.

Portfolio budget caps: a ceiling, not a steering wheel

Portfolios group campaigns so you can see and cap them together. A cap can cover a date range or recur monthly. Amazon is explicit about how it behaves: the portfolio cap overrides campaign budgets, and once it is exhausted “all campaigns associated with that portfolio will be paused, even campaigns that still have budget remaining"³.

That makes a cap useful as a hard limit, for example on a launch group, and risky on a hero portfolio, where a cap reached on the 24th switches off your best campaigns for the rest of the month. We put caps on launches and tests, and manage hero and core budgets at campaign level.

Amazon is also testing a feature that lets Sponsored Products campaigns in a portfolio use budget that other campaigns in the same portfolio left unspent earlier in the month, so a campaign can spend up to 100% more than its average daily budget on a given day. It is in beta and not available to everyone⁴. If you have it, group only campaigns with similar targets in the same portfolio, so spare budget cannot flow from a hero to the tail.

A campaign structure you can manage in bulk

The structure we use has one rule: each campaign has one product group, one intent and one targeting type, so its results mean one thing. Intent matters most. Brand, category and competitor searches have very different economics, and blended together a cheap, efficient brand term hides an expensive competitor term.

Name every campaign so it can be filtered. At scale, campaign names are your database. Our convention, which is our practice and not an Amazon requirement:

So “SP_HERO_KNIVES-CHEF_CAT_EXACT_2610” tells anyone with the bulksheet open what the campaign is for and which rules apply to it.

Split variations that behave differently. When two colours or sizes share a campaign, budget tends to follow whichever has the longer sales history, not whichever converts better. Our PPC strategy guide describes an audit where exactly this happened. In a big catalogue, check it systematically: compare conversion by child ASIN in your Business Report⁹ with ad spend by child ASIN, and split any family where the two point in different directions.

Bulk operations: change hundreds of campaigns at once

Bulk operations let you download your campaigns as a spreadsheet, edit it, and upload it to create, update or archive campaigns, ad groups, keywords and other entities, for Sponsored Products, Sponsored Brands and Sponsored Display. Sellers reach it from the campaign manager, under All campaigns. A download can include up to 60 days of performance data, and a Sponsored Products sheet can include search term data with the related campaign IDs⁵.

The weekly bulk routine we run:

1. Download a bulksheet with the last 30 days of data, plus a Business Report and an inventory report for the same dates.
2. Join them on SKU or ASIN, so every ad row carries stock cover, Featured Offer percentage and margin.
3. Apply the rules by filter: pause ads where cover is below lead time or the product is suppressed; lower bids on targets above the role’s target ACoS with enough clicks to judge; raise budgets on campaigns at their target that ran out of budget early; add negatives for search terms that spent without converting.
4. Upload, and keep both files with the date. The pair is your change log.
5. Review exceptions by hand: anything the rules would change by a large amount, and every hero.

Automation helps, and Amazon is adding more. Amazon Ads Agent’s conversational features, including natural language analytics and one-click recommendations for Sponsored Products targeting, bids and budgets, are rolling out over the coming months¹. Treat any recommendation, from Amazon or a tool, as a proposal checked against your role targets, not as a decision.

Measure the portfolio, not the campaign

ACoS judges a campaign. At portfolio scale, three measures matter more:

• TACoS by role: ad spend ÷ total sales for each role. Heroes whose TACoS falls as sales rise are building organic rank; heroes whose TACoS rises with sales are renting growth. Our TACoS guide covers how to set a ceiling.
• Profit after advertising by role: total sales × (margin − TACoS).
• What the ads add. Sellers’ Sponsored Products sales are credited on a seven-day, last-click basis, and brand ads generally over 14 days⁶, so attributed sales include some sales that would have happened anyway.

The portfolio gives you a way to test that last point that a small account doesn’t have. Pick two similar product groups, pause or cut the ads on one for two to four weeks, and compare total sales for the two groups against their own history. If the paused group’s total sales barely move, its ads were mostly buying sales you would have made anyway.

For deeper analysis, Amazon Marketing Cloud is a clean room for analysing Amazon Ads signals alongside your own data. Amazon says it is available at no cost to eligible advertisers, including in the UK, that advertisers using sponsored ads can access it self-service, and that templates mean no SQL is needed for common analyses⁸.

When DSP and DVA+ fit

On 29 September 2026 Amazon renamed its unified ad platform Amazon Ads Agent, with three campaign types: sponsored ads; Display, Video and Audio; and Full-Funnel Campaigns¹. The changes that matter for a large catalogue:

• DVA+ combines what was run as Sponsored Display, Sponsored TV and programmatic campaigns into one buying experience across streaming TV, online video, display and audio, on Amazon and the open internet. You choose a goal, products, creative and budget, and Amazon’s AI places the ads; “advanced settings” keep manual control for experienced buyers. It begins rolling out from late October¹.
• Existing DSP campaigns continue without disruption, with settings, audiences, creative and history carried forward, and current pricing continues within DVA+².
• Full-Funnel Campaigns, which combine sponsored ads, display, video and streaming TV in one AI-run campaign, are available only in the United States so far¹.
• Access: Display, Video and Audio campaigns are available self-service with budgets you set, or through managed service, which typically requires a minimum spend that varies by country⁷.

Our test for whether a brand is ready, which is our practice and not Amazon’s:

1. Search advertising is profitable at role level, and TACoS is stable or falling on heroes.
2. Hero listings convert well, so paid traffic lands on pages that sell.
3. Stock can absorb more demand without stocking out.
4. You can measure what it adds: a holdout group of products or a pre-agreed before-and-after comparison, plus new-to-brand sales.
5. The budget is ring-fenced with an end date, so a test cannot quietly take money from proven search campaigns.

Start with your heroes and the products whose customers buy again, where reaching new shoppers has the longest payback. If after the test period total sales for the tested products have not moved beyond what the comparison group did, the spend was not adding enough, however good the platform’s own attribution looks.

A portfolio checklist

1. Every SKU has a role, and every role has a budget and a target ACoS based on margin.
2. No spend on out-of-stock, suppressed or loss-making SKUs except deliberate launches.
3. Every campaign name carries ad type, role, product group, intent, targeting and start month.
4. Portfolio caps sit on launches and tests, not on heroes.
5. A weekly bulksheet routine joins ad data to stock and margin, with files kept as a change log.
6. TACoS and profit after advertising are reported by role every month.
7. DVA+ or DSP is a ring-fenced test on heroes with a holdout, not a default line in the budget.