Amazon DSP Audiences and Targeting: A Practical Map
In-market, lifestyle, remarketing, lookalikes, AMC — the taxonomy is flat but performance is not. A ranked view of which DSP audiences deserve your budget first.
Amazon DSP audiences fall into two families: segments Amazon builds for you from its own shopping and streaming signals (in-market, lifestyle, interests, life events), and segments you build from your own signals (remarketing, advertiser uploads, lookalikes, AMC-derived audiences). The first family is cheap to switch on and easy to overspend against; the second is where most of the return sits. This article maps the whole taxonomy, then takes a position on which audiences deserve the impression at which stage of the funnel.
If you are still deciding whether the channel belongs in your plan at all, start with Amazon DSP explained and the budget-allocation view in DSP vs Sponsored Ads. What follows assumes you already have a seat.
The audience taxonomy, in one table
Amazon Ads groups its pre-built segments into four documented categories — in-market, lifestyle, interests and life events — all assembled from first-party signals across the Amazon store, Prime Video, Twitch and IMDb. On top of those sit the audiences you create. Naming shifts between the DSP console, Sponsored Display and the Ads API (the API exposes them as AUDIENCE_SAME_AS_IN_MARKET, _LIFESTYLE, _INTEREST, _LIFE_EVENT), so treat the labels below as the concept rather than a literal screen string.
| Audience | Built from | Typical size | Funnel stage | Our default view |
|---|---|---|---|---|
| In-market | Recent category shopping and browsing | Medium | Consideration | The most reliable prospecting audience on the platform |
| Lifestyle | Aggregated shopping and streaming behavior ("foodies", "sports enthusiasts") | Very large | Awareness | Use deliberately, cap it, expect weak short-term ROAS |
| Interests | Frequently browsed and bought topics | Large | Awareness | Similar to lifestyle; rarely worth a separate line item |
| Life events | Milestone signals (new parents, moving, traveling soon) | Small to medium | Awareness / consideration | Excellent when your product maps to the moment, noise otherwise |
| Views remarketing | Shoppers who viewed your detail pages | Small | Conversion | The workhorse. Fund it first |
| Similar-product remarketing | Shoppers who viewed comparable ASINs | Medium | Consideration | Strong conquesting play when your PDP is genuinely better |
| Purchase remarketing | Past buyers of your ASINs | Small | Retention | Underused; pays for itself on consumables |
| Brand halo | Shoppers who engaged with any ASIN in your brand | Small to medium | Conversion | Widens views remarketing without leaving your brand |
| Pixel / advertiser audiences | Your site tag, CRM lists, DTC data | Varies | Conversion / retention | Only worth the setup if you have real off-Amazon traffic |
| Lookalikes | Modeled from one of your own seeds | Large | Prospecting | Only as good as the seed |
| AMC custom audiences | SQL over your own event-level data | Varies | Any | The most precise option, and the highest effort |
| Third-party data providers | External providers layered into the DSP | Varies | Awareness | Adds data cost; justify it before you buy it |
Contextual targeting sits outside this table on purpose. Amazon documents it as a separate targeting mode — you are buying a page or a category context, not a shopper segment — and it combines with audiences rather than competing with them.
Which audiences are worth the impression
The taxonomy is flat. Performance is not. Ranked by how often each one earns its place in a plan we would defend to a CFO:
Fund first: views remarketing, brand halo, purchase remarketing. These are people who already met your product. Costs per impression are similar to everything else on the platform, but conversion rates are not close. If your DSP budget is under roughly $15k a month, there is a strong argument for spending all of it here and nowhere else, then widening once the retargeting pool is saturated — you can see saturation in the reporting when reach flattens while frequency climbs.
Fund second: similar-product remarketing and in-market. These are the two prospecting audiences with an actual intent signal underneath them. In-market is the one to scale; similar-product is the one to watch, because conquesting only works when the shopper who lands on your detail page finds a better offer than the one they just left. If your reviews, price or images lose that comparison, conquesting is a subsidy to your competitor. Audit the PDP before you fund the audience — our Amazon listing SEO work usually runs ahead of a conquesting push for exactly this reason.
Fund last, and only with a cap: lifestyle, interests, broad lookalikes. These are where DSP budgets quietly disappear. They are enormous, they deliver impressions without complaint, and they will return a flattering CTR alongside a return on ad spend that never justifies the line. They have a legitimate role — launches, category entry, staying visible off Amazon — but that role is a capped percentage of the budget, agreed in advance and measured on new-to-brand outcomes rather than blended ROAS.
The quiet budget burners are less obvious than "broad audiences": overlapping line items bidding against each other for the same shopper, remarketing windows so long the audience is mostly people who already bought, third-party data layered on top of Amazon segments that were already sufficient, and lookalikes seeded from a list too small to model anything useful. Each looks like activity in the console and like nothing in the P&L.
Amazon DSP placements and supply
Targeting decides who; placements decide where, and the two get conflated constantly. By default a DSP line item is eligible across Amazon owned-and-operated properties, Amazon Publisher Direct inventory and open exchanges — which means an untouched supply setting will send your budget wherever it clears cheapest. That is not automatically wrong, but it should be a decision.
Practical rules we apply to Amazon DSP placements:
- Separate Amazon owned-and-operated from open-web supply into different line items. They perform differently enough that a blended number tells you nothing, and you cannot rebalance what you cannot see.
- Treat streaming and audio as their own plan. Prime Video, Fire TV and Twitch inventory buy attention, not clicks; judging them on click-through is a category error.
- Cap frequency per audience, not per campaign. A retargeting pool of 40,000 shoppers absorbs a very different daily frequency than a lifestyle segment of several million.
- Leave pre-bid brand safety and invalid-traffic filters on. The saved CPM is never worth the alternative.
Reading the reporting without fooling yourself
DSP reporting gives you detail page view rate, add-to-cart, new-to-brand purchases, purchase rate, reach and frequency alongside ROAS. Two habits keep the numbers honest.
First, judge each audience on the metric matching its job. An awareness line item measured on same-week ROAS will always lose; measure it on detail page view rate and new-to-brand share instead. A retargeting line item measured on reach will always look small; measure it on purchase rate.
Second, accept that native DSP reporting cannot tell you how audiences interact — whether the shopper who converted after a DSP impression had already clicked a Sponsored Products ad three times that week. That question needs event-level data, which is exactly the gap Amazon Marketing Cloud fills, and why overlap and path-to-conversion analysis is the first thing we run when a DSP program has been live long enough to have history.
FAQ
What are the audience types in Amazon DSP?
Amazon Ads documents four pre-built categories — in-market, lifestyle, interests and life events — built from first-party shopping and streaming signals. Alongside them you can use remarketing audiences (views, purchases, brand halo, similar product), pixel and CRM-based advertiser audiences, lookalikes, AMC custom audiences and third-party data segments.
Is retargeting better than in-market targeting on Amazon DSP?
Retargeting almost always converts better, because those shoppers already viewed your product. In-market is the better scaling audience once the retargeting pool saturates — visible when reach flattens while frequency rises. Most healthy programs run both: retargeting funded first for efficiency, in-market layered on for incremental reach.
Do I need a minimum budget for Amazon DSP?
Amazon's managed-service option has historically carried a minimum around $50,000, varying by country, while the self-service minimum was removed in late 2025 and access now runs through the Ads console or an agency seat. Practically, the model needs enough conversion signal to optimize, so plan a sustained monthly budget rather than a test.
Where do Amazon DSP ads appear?
Across Amazon owned-and-operated properties (the Amazon store, IMDb, Twitch, Fire TV, Prime Video), Amazon Publisher Direct inventory, and third-party sites and apps via open exchanges. By default a line item is eligible for all of them, so restrict supply deliberately and split owned-and-operated from open web to keep reporting readable.
Where to start this week
Three things to do before you touch a bid. List every active line item with its audience and supply setting, and flag any two that could serve the same shopper. Check the lookback window on each remarketing audience against your actual purchase cycle. Then split one blended line item into Amazon owned-and-operated versus open-web supply and give it two weeks. If the account is inherited and you cannot reconstruct why the audiences were built this way, a structured account audit beats reverse-engineering it line by line.
