Amazon's new campaign recommendation panel reaches every US advertiser this month. Read it, don't bulk-apply it.

What is actually arriving in October, and what is not

On 22 September 2026 Amazon Ads published a post about new sponsored ads tools for small businesses. One sentence in it sets the clock: "This intelligence dashboard will be available to all sponsored ads advertisers in the United States in October 2026." We are halfway through that month. If you sell in the US and run Sponsored Products, the panel is either on your campaign overview page already or it will be before November.

What it is, in Amazon's own words: "Sponsored ads advertisers now benefit from a new intelligence dashboard which surfaces insights and actions across targeting, bidding, and budget controls." And: "Available on the campaign overview page, it analyzes activity, explains how each campaign is performing, what is impacting results, and how to improve." The worked example Amazon gives is that "it might flag that a top-performing keyword is losing impression share and recommend a bid adjustment to recapture it or recommend adjusting targeting to help reach customers who had browsed but not purchased."

This is not Amazon's first recommendation engine, and it is worth being clear about that before anyone sells you a webinar on it. Amazon's own campaign-performance guide already tells advertisers "You'll find these campaign recommendations in your advertising account in the campaign manager" and that you "can edit them to suit your needs". What changes is placement and persuasion. The new panel sits on the page you open every morning, and each item comes with a projected outcome attached: "Each action is explained with visuals illustrating likely impact, and advertisers can apply, edit, or skip each action."

That difference matters more than it sounds. A recommendation buried in a sub-menu gets ignored. A recommendation on the overview page, with a chart showing what it will supposedly do to your sales and a one-click Apply next to it, gets taken — especially by someone checking their account at eleven at night in the middle of peak. The feature is not the algorithm. The feature is the button.

One of these recommends. The other one acts.

The same announcement describes a second, separate feature, and the two are already being muddled in commentary. Auto-optimization is the one that changes your campaigns itself: "Auto-optimization continuously finds and acts on opportunities to keep campaigns performing strongly as shopping habits shift." Amazon's full description of what it does is specific — it "expands product targeting to help advertisers engage with audiences who are considering similar or complementary products and categories, adds relevant keywords and pauses underperforming ones, and identifies seasonal opportunities—all within the advertiser's existing campaign settings."

And the date on it is not this month: "Auto-optimization will be available to advertisers in the United States in January 2027." So nothing arriving in your account in October adds a keyword or pauses one on its own. If you read somewhere this week that Amazon is now rewriting your keyword lists automatically, that is the January feature being described three months early. The thing you are getting now stops at a suggestion.

It is still worth knowing the shape of auto-optimization now, because the decision about it lands in a quieter month. Amazon says you "Enable auto-optimization with a single click in your campaign settings", that advertisers "can see how each action performs and edit or reverse actions at any time", and that a detail view "shows every keyword added, its spend, and conversions earned". That is an opt-in feature with a visible log and a reverse button — a reasonable design, and a genuinely testable one, which is exactly why it should be tested in January rather than accepted on faith.

The review-then-apply pattern runs across the rest of the platform too. Amazon's page for the conversational Ads Agent experience says advertisers "get to review a summary of proposed changes and approve them before they're applied, so they stay in control", and that the conversational experience itself is still expanding "to eligible advertisers through a rolling closed beta". Which is the real story of October: the chat assistant is still gated, but the recommendation panel is the piece that reaches everybody.

"Recapture lost impression share" means something different in October

Take Amazon's own example seriously, because it is the recommendation you are most likely to see: a top-performing keyword is losing impression share, so raise the bid to win it back. In February that is often a good trade. In the second half of October it usually describes a different situation entirely — you have not got worse, your competitors have got more aggressive, and the clearing price for that keyword has risen for everyone bidding on it.

The number that decides whether to accept the recommendation is your break-even ACOS, and the panel does not know it. Impression share is a diagnostic, not a target; nothing good happens because a percentage went up. A recommendation engine pointed at recaptured share will cheerfully buy that traffic back at a cost per click that puts the keyword underwater, and between 15 October 2026 and 14 January 2027 it sits further underwater still, because Amazon's holiday peak fulfilment fees take a bigger bite out of the margin on every FBA unit it sells.

So before you accept a bid increase, the thing to go and get is not more data from the ad console. It is your current contribution margin per unit, after this quarter's fulfilment costs, which tells you the maximum CPC that keyword can carry. Then the recommendation becomes a question you can answer in ten seconds rather than a chart you are inclined to trust.

The second half of Amazon's example — "adjusting targeting to help reach customers who had browsed but not purchased" — is a perfectly sensible idea with a timing problem. Broadening targeting on your highest-traffic ASINs in late October means the test learns on your most expensive clicks of the year and you read the result through the noisiest data you will see all year, with peak competition, peak CPCs and your own stock position all moving at once. The recommendation is not wrong. It is unreadable until January.

Amazon's beta numbers came from Amazon's beta

The announcement carries one performance claim: "In beta testing, campaigns that adopted keyword auto-optimization saw a first sale 40% faster and a 13% increase in sales." That is a real published claim and we are quoting it accurately. What is not published alongside it is a sample size, a category breakdown, a baseline definition or any methodology, and the advertisers who volunteer for an advertising beta are rarely a representative sample of advertisers.

Read the first figure carefully as well, because it is narrower than it sounds. "A first sale 40% faster" is a launch metric — it describes how quickly a new campaign or a new product gets its first conversion. If you launch products regularly that is a real benefit worth wanting. It says close to nothing about what the feature would do to a mature campaign that has been running on the same twenty converting keywords for a year.

The same page also claims that "advertisers using first-party data with Amazon Ads see a 30% increase in ROAS compared to those who don't". That one is correlational by construction. Advertisers who have website activity and email lists to upload are, almost by definition, advertisers with an established business off Amazon, better brand recognition and more repeat buyers. Comparing them with advertisers who have none of that measures the businesses, not the feature.

None of this makes the tools bad. It makes vendor percentages useless as a forecast. If you decide to work through a batch of recommendations, write down first what you expect to happen to your own numbers — blended TACOS, units, contribution margin — and check that against reality in three weeks. The honest version of a benchmark is your own account last month.

Ads data manager is the one that is already live, everywhere

One item in the same announcement needs no waiting and got the least attention: "Ads data manager expansion to sponsored ads is available now to all advertisers, globally." Amazon describes it as a way to combine your own data — "like website activity and emails" — with Amazon shopping, browsing and streaming signals, and says "No additional budget or data expertise is needed."

Whether that is useful to you depends entirely on one question: do you actually hold any first-party data? If every sale you make happens on Amazon, you have no website traffic worth uploading and no mailing list, then there is nothing to connect and "available now to all advertisers" is not an instruction aimed at you. That describes a large share of the sellers who read a line like that and feel behind. If you do run a DTC site or a real email list, this is the one item from the announcement worth an hour of your time this quarter.

One caveat before you upload anything. Amazon's assurance is that "All data usage is privacy-safe and exclusively for each advertiser's own campaigns", and there is no reason to doubt the technical description. But sending your customer records to an advertising platform is a legal decision as well as a marketing one, and the answer depends on where your customers live and what your own privacy policy told them you would do with their details. That is a check to run before the upload, not afterwards.

What to do this week

First, go and look. Open the campaign overview page and find out whether the panel has reached your account yet, because a rollout stated as a month is a gradual one. Then read every recommendation it offers and apply none of them. What you are learning on this pass is what kind of recommendations it makes about your account — whether it has spotted something you genuinely missed, or whether it is mostly offering to spend more money faster.

Second, decide your triage rule before the button tempts you, because deciding in the moment is how peak budgets disappear. A rule that works: accept budget increases on campaigns already converting above break-even, since they are reversible and change nothing structural. Edit, never apply as offered, any bid change — check the proposed cost per click against your break-even ACOS with this quarter's fulfilment costs in it. Skip every targeting expansion and new match type until January. One-click Apply is designed to remove friction; your rule is the friction, and it has to exist in writing before you are tired.

Third, keep a record of what you accepted and when. In six weeks you will be trying to explain a change in performance, and "Amazon recommended it" is not a record. If the panel does not give you a change log you can export, a dated line in a spreadsheet for every recommendation you apply costs you fifteen seconds and saves you a week of guessing in December.

Fourth, put the real decision in your calendar for January. That is when auto-optimization is due in the US, when it will be a single click in campaign settings, and when your data stops being distorted by peak. Note now what you would measure: the keywords it adds, what they spent and what they converted, all of which Amazon says the detail view will show. Judging those against the keywords you would have added yourself is a fair test of the feature. Judging it against 13% is not a test of anything.