Amazon Ads Off-Amazon Audiences: When to Test

What actually changed this week

A purchase-data provider's off-Amazon shopping audiences are now available inside Amazon Ads DVA+. The provider announced it on November 4. You can now build campaigns using signals from purchases made outside Amazon.

• You can segment by individual retailer, combinations of retailers, purchase category, and online versus in-store shopping.
• The provider states its signals refresh daily and come from permissioned consumer spending data. Both claims are the provider's own, not independently verified.
• Per industry reporting, activation spans Amazon's ad network, including streaming video and third-party inventory.
• This applies to DVA+. It is not a new targeting toggle inside every Sponsored Products campaign.
• The integration expands a prior relationship that had covered offline sales-lift measurement.

That last point matters more than it looks. Measuring an outcome after the fact and choosing an audience before an ad serves are two different things. Before this, the relationship was about measurement. Now you can pick the audience up front and run the buy through Amazon's demand-side platform.

Who should and should not test off-Amazon audiences

Test this if you already run disciplined ads, you can afford a budget you might lose, and you can name a group you are clearly missing. Skip it if your current campaigns are leaking money or your listing does not convert the traffic you have now.

Good fit if:

• You have a category where rival shoppers clearly exist off Amazon.
• You can run a holdout or a credible comparison.
• Your margins give you room to pay for prospecting.
• Your listing already converts normal traffic at a healthy rate.

Bad fit if:

• Your Sponsored Products campaigns are not profitable yet.
• You would divert acquisition budget just because this sounds relevant.
• Your listing has weak images, thin copy, or missing keywords.
• You cannot measure incrementality and would read ROAS alone.

Do not move existing budget into this just because a new audience feels more relevant. Fund it with money you set aside for a test.

Write a sharp hypothesis before you spend a dollar

Do not test "reach competitor shoppers." That is not a hypothesis. Define one category and one testable idea you can prove or kill.

A sharper version looks like this:

• Category buyers: "Shoppers who buy in my category off Amazon but have never bought my brand will convert at a higher new-to-brand rate than my current prospecting."
• Multi-retailer shoppers: "Buyers who shop my product type across two or more retailers are high-value and worth a premium to reach."

These are suggestions, not confirmed capabilities at every level of detail. Before you build anything, confirm which audience definitions, lookback windows, and exclusions your account can actually use. The tool may not slice the way your hypothesis assumes.

How to isolate real incrementality

Use a holdout group or a credible comparison wherever the tools allow. That is the only honest way to separate sales your ad caused from sales that were coming anyway. The incrementality problem is biggest exactly here, because purchase-based targeting selects for shoppers who already have strong intent.

• A holdout keeps a slice of your target audience from seeing the ad. The gap in sales between the exposed group and the holdout is your real lift.
• Keep creative, offer, and measurement windows constant. If you run a discount at the same time, you cannot tell if better targeting or the discount drove the lift.
• Pick one variable to test. Change one thing, measure, then change the next.

No holdout, no honest read. If your partner cannot offer a holdout or a clean comparison, treat any "it worked" claim as unproven.

Why attributed revenue can lie to you

Attributed revenue can take credit for sales that would have happened anyway. A 5:1 reported return can still be a loss once you strip out the demand you already owned. Here is the illustrative example from the source. These numbers are an example, not reported real campaign results.

Read the last line again. The campaign looked like a 5:1 winner. It lost $2,000 before you even counted returns, data fees, and agency time. That is why attributed revenue alone is not a verdict. You need the incremental slice.

New-to-brand vs ACoS: read the right signal

For a prospecting audience aimed at rival-product buyers, new-to-brand tells you whether you are reaching genuinely new customers. ACoS alone can look healthy while you quietly pay to reclaim demand you already had.

• New-to-brand: if the goal is to win buyers you do not have, this is the number that proves it.
• ACoS: useful, but it does not tell you if the sale was incremental. A low ACoS on returning buyers is not expansion.

Decide before launch what you are measuring: Amazon contribution, total brand contribution, or another outcome. Then read the metric that matches that goal. Do not pick the goal after the data comes in.

Where the sale lands when you sell across channels

If you sell across multiple retailers, an ad can lift total brand demand while producing a different result for your Amazon account specifically. The sale may land somewhere other than Amazon.

• A buyer sees the ad, then buys your product at another retailer. Total brand demand went up. Your Amazon numbers did not.
• This is fine if your goal is total brand contribution. It is a problem if you funded the test to grow the Amazon account.

Pick the outcome you care about first. The same campaign can be a win for the brand and a loss for the Amazon P&L. Both can be true at once.

Guardrails: budget caps, stop conditions, documentation

Set your guardrails before launch, not after. The test should have a budget cap, a stop condition, and a written record so you can judge it honestly.

• Full-cost budget: include media, data, agency, creative, discounts, returns, and fulfillment. Not just the media spend.
• Stop condition: decide in advance what weak contribution or excessive frequency looks like, and pull the plug if you hit it.
• Pre-launch record: write down the audience definition, campaign setup, all cost components, attribution windows, and your success threshold before you start.

Documenting the success threshold before launch stops you from moving the goal posts when the data disappoints. That is the single easiest way to fool yourself.

Tighten keywords and listings so bought traffic converts

Buying a premium audience does not fix a weak listing. If your detail page does not convert, you are paying top dollar to send people to a page that loses them. Audit your keywords and listing content so purchased traffic has a reason to convert.

• Make sure your title and bullets name the problem the buyer is solving.
• Cover the real search terms buyers use, not just the words you like.
• Fill keyword gaps your rivals already own.
• Check that your images answer the objections a new-to-brand buyer has.

Do this before you fund the audience, not after. A converting listing lowers the cost of every click, paid or organic. For the mechanics, see why Amazon listings get traffic but fail to convert and how to lower TACoS by growing organic traffic.

Questions to force before signing any proposal

Before you commit real budget, get the hard numbers in writing. The announcement did not publish a seller-facing price list, a universal minimum spend, or an eligibility matrix. So you must confirm the terms yourself.

• What is the usable audience size, in writing?
• What is the total cost, including data and agency fees?
• What is the measurement design? Is there a holdout?
• Which audience definitions, lookback windows, and exclusions can my account use?
• What are the current DVA+ pricing, minimum spend, and eligibility rules?

Do not reuse older DSP budget or access figures. They may not match current DVA+ terms. Confirm pricing, minimum spend, and eligibility directly rather than working from old examples.

What this is not: access to customer data

Access to an audience is not access to identifiable customer records. You can target a group. You do not get their names, cards, or spending files.

• The provider's permission claim describes how it collects data, not any right for you to obtain or reuse consumer financial data.
• You are buying reach to a defined segment through Amazon's buying model, full stop.

Keep that straight when a proposal makes it sound like you are getting customer-level intel. You are not.

Conclusion

Off-Amazon purchase audiences are a real new prospecting lever. They are also a fast way to pay a premium for sales you already owned. The deciding factor is not the audience. It is your discipline. Small budget, a holdout, new-to-brand as the signal, and a success threshold you set before launch.

Fund it only after your foundation is solid. A converting listing and a clean keyword base make every buy cheaper. Start with your 2026 Amazon SEO and ads playbook and the honest math on sales strategy over ad spend. Then test the audience with money you can afford to lose.

References

1. Amazon Ads, online advertising for businesses of all sizes. sell.amazon.com
2. Amazon official announcements and small business updates. aboutamazon.com/news/small-business