How to Target Marketplace Audiences in Open Web Performance Campaigns

On-site ad inventory on Amazon, Walmart, and similar platforms has become crowded and expensive. Every brand selling on these marketplaces is bidding for the same spots, pushing CPCs up and capping your budget’s reach.

Outbidding everyone for the same impressions is a poor fix. A better solution is to find and reach those same marketplace shoppers elsewhere: reading the news, watching videos, or scrolling a publisher’s website.

This approach works because the audience signal originates from verified purchase behavior; it’s not a guess. A shopper who abandons a cart — or searches the same product category three (or four) times — doesn’t lose that intent when they close the tab.

Off-site retail media extensions and first-party data integrations let you:

• Follow that intent across the open web.
• Attribute resulting sales back to specific SKUs.
• Adjust bids in real time using AI-powered programmatic advertising.

Third-party cookie deprecation has eliminated the probabilistic targeting infrastructure that open web campaigns once depended on; a first-party-anchored approach tied to dynamic retargeting is now driving return on ad spend (ROAS) gains for performance marketing teams.

Activating Retailer First-Party Data for Off-Site Reach

Retail media networks have transactional truth, which most demand-side platforms (DSPs) lack. Their first-party data isn’t modeled or inferred, it reflects actual purchase history, product search behavior, and loyalty program activity from shoppers who’ve already shown what they want to buy.

When you partner with a retail media network (RMN) directly — or access their audience segments through a DSP with a certified data connection — you can serve ads to shoppers across open-web inventory (e.g., news sites, content platforms, or streaming services) outside the marketplace itself. This access distinguishes off-site retail media from standard programmatic display. The targeting foundation is built on verified accounts and not from guesses cobbled together from browser signals.

The mechanics vary by network:

• Amazon DSP exports first-party audience segments into the open web supply.
• Walmart Connect uses The Trade Desk’s DSP infrastructure to push Walmart shopper segments into programmatic environments.
• Target Roundel has partnerships with Google and The Trade Desk that enable similar reach extension.

All three share a basic pipeline: Verified purchase intent flows from the marketplace into open web placements, bypassing the third-party cookie ecosystem that browsers have been dismantling since Apple’s 2017 ITP changes and Google Chrome’s 2024 depreciation rollout.

In practice, you can target a shopper who viewed a product on a marketplace three times this week on Amazon and reach them while they’re reading an article on a publisher site. (And you don’t need a cookie to make the connection.) Audience match rates tend to run higher under this model than under cookie-based targeting, as the identity graph traces back to an authenticated marketplace account, and not a browser fingerprint that decays across devices.

Start by looking at where your sales happen. If 80% of your transactions run through one marketplace, activate that network’s off-site extensions first; then, build a multi-RMN data pipeline.

Beyond retail media networks, performance advertising platforms like Realize offer their own first-party data marketplace as an alternative or complementary targeting layer. Realize’s Data Marketplace includes Taboola first-party audiences built from partnerships with over 11,000 premium publisher sites and Connexity eCommerce signals, alongside third-party segments from providers like Bombora and Eyeota. These segments cover 200+ categories spanning demographics, interests, and intent signals, giving advertisers access to verified audience data outside the walled gardens of individual marketplaces.

Syncing Your First-Party Integrations for Dynamic Retargeting

RMN audience data covers shoppers who’ve interacted with your products on the marketplace. Your own first-party data (e.g., CRM lists, offline conversions, and server-side pixel data) covers a different but equally valuable group: customers you already know from your own channels. Combining both gives you a sharper retargeting layer.

The technical foundation for this integration has three components: server-side pixel tracking, CRM list matching, and data clean rooms.

Server-side pixel tracking

Server-side pixel tracking routes conversion data through your server infrastructure, rather than relying on browser-based tags. Browser privacy restrictions and ad blockers don’t impact it. It captures conversion events that client-side tracking misses and generates a cleaner signal for downstream audience building. As browsers limit the lifespan and scope of first-party cookies, server-side tracking is becoming the standard for getting a trustworthy conversion signal.

CRM list matching

CRM list matching lets you upload customer records (e.g., email addresses, phone numbers, and hashed identifiers) to DSPs and programmatic platforms. The data is then matched against authenticated user profiles within their supply networks. Since these audiences come from purchase relationships, they tend to beat probabilistic third-party segments on match and downstream conversion rates. Results still depend on list size and recency, so a stale CRM file won’t perform as well as one that’s regularly refreshed.

Platforms like Realize enable advertisers to upload CRM lists directly for matching against authenticated user profiles across the open web, and to build lookalike audiences from those matched segments. This extends the value of first-party data beyond retargeting known customers to finding new prospects who share similar characteristics, a useful complement to marketplace audience campaigns focused on shoppers who’ve already interacted with your products.

Data clean rooms

Data clean rooms provide a privacy-compliant environment for combining your data and a retailer’s first-party data without sharing raw records from either party. Amazon Marketing Cloud, Walmart Luminate, and clean room infrastructure providers (e.g., Snowflake, LiveRamp, and Habu) support this process.

The output is an audience that reflects the intersection of the retailer’s purchase signals and your CRM history: high-intent shoppers you’ve already reached through open web inventory and creative tailored to their position in the purchase cycle.

The product catalog feed makes dynamic retargeting work. Connect your catalog to the DSP or creative platform in real time, and ads populate with the specific SKUs each user looked at (instead of a generic brand message). A shopper who views a specific product sees that product, its current price, and available inventory, pulled live from the catalog at render time. Dynamic retargeting (rather than regular display retargeting) drives higher click-through and conversion rates for open web campaigns built around marketplace audiences.

Four Strategies for Retargeting Marketplace Audiences Effectively

Reaching the right audience just opens the door. Bidding strategy, creative relevance, and exposure management dictate the campaign performance.

1. Segment audiences by granular purchase intent

Not all marketplace audience segments have equal conversion potential, and bidding like they do wastes budget. The most important segmentation variable is proximity to purchase:

• High-intent segments that justify aggressive bids and conversion-focused creative include cart abandoners, repeat product viewers, and recent category searchers.
• Mid-intent segments, including shoppers who visited a product page without adding to cart — or browsed a category without clicking a specific product — are further out. Soften the bid and lean into consideration-stage creative (e.g., comparisons, reviews, or use cases).
• Passive browsers are shoppers who landed on a category page once or showed a single interaction signal weeks ago. Passive browsers convert at a low enough rate that you can exclude them from your retargeting window. Keeping them inflates impressions without moving conversions, which distorts your ROAS math.

Try this bid structure:

• Cart abandoners at 2-3x your base bid.
• Repeat viewers at 1.5x.
• Category browsers at base or slightly below.

This tiering reflects actual purchase probability, rather than giving every retargeting impression equal weight.

For advertisers using marketplace audience segments on the open web, Realize supports AND logic when layering multiple audience criteria, e.g., combining a demographic segment like “Age 25-34” with an interest signal like “Auto” to target only users who match both conditions. This allows for more precise segmentation than broad OR-based targeting, though advertisers should monitor reach estimates to avoid over-narrowing their audience.

2. Leverage AI-powered bidding for real-time optimization

No human can manually track device, time of day, content environment, recency of marketplace activity, and current behavioral signals fast enough to price impressions correctly at scale. AI-powered bidding systems can. They read those signals and adjust bids accordingly, bidding aggressively on a cart abandoner who’s returned to search the category and read related content, or pulling back on lower-probability inventory.

Actual results support this claim:

• Minor Hotels used AI-driven automated bidding through the Realize performance advertising platform and achieved 5x ROAS on a campaign targeting high-intent travel audiences.
• Also using Realize, PortAventura World paired precision retargeting with automated bidding and beat its other performance channels by 44% on ROAS, with a 47% lower cost per acquisition in 2024.

For marketplace audience campaigns specifically, the signal worth feeding into your bidding system first is recency: how long ago the marketplace interaction happened and how close the user is to converting. A cart abandoned two hours ago presents a different targeting opportunity than one abandoned a week ago. A bid system that prices the distinction correctly, in real time, will outperform one that doesn’t.

3. Use Dynamic Creative Optimization (DCO)

Generic retargeting creative ignores the one thing you know about the user: what they looked at. Dynamic creative optimization (DCO) solves this issue by connecting your live product catalog to the ad template, so each impression shows the specific SKUs someone viewed, related products, current pricing, and inventory status.

You’ll need a structured product feed, creative templates with dynamic slots for product image, name, price, and CTA, and a way to map browsing history to SKU identifiers. Once in place, someone who saw a product sees the same product in an ad with a price drop flagged (if there’s been one) or a low-stock warning (if that’s the case).

For brands with larger catalogs, DCO also enables related product recommendations. You can show complementary items alongside the item a shopper already viewed, which can lift average order value beyond the original product.

4. Cap frequency to prevent ad fatigue

A user who sees the same ad 20 times over three days isn’t any more likely to convert on the twentieth impression. Beyond a certain point, you’re paying to annoy people.

For high-intent segments, including cart abandoners, three to five impressions per day, across the first 72 hours, is a reasonable starting cap. Drop that frequency to one to two impressions per day over the following week if there’s still no conversion. Mid-intent segments should start lower.

If your campaign runs across multiple DSPs, frequency management should account for total exposure across all of them. Otherwise, a user can hit their frequency cap on one platform and keep seeing your ads, uncapped, on another. Pair frequency caps with a recency window (14 to 30 days from the last marketplace interaction) so you’re not chasing intent that’s already gone cold.

Closing the Loop: Measuring Open Web Campaign Performance

The hardest part of off-site retail media is proving it worked. A shopper sees your ad on a publisher site, then converts back to the marketplace. Without a measurement system connecting those two events, you can’t tell whether your open web campaign drove the sale, or whether it would’ve happened without your campaign.

Closed-loop attribution addresses this conundrum by matching ad exposure data against purchase event data inside a controlled environment. The retailer’s transaction records serve as the conversion source; the DSP (or RMN’s ad-serving logs) are the exposure record. When both share a common user identifier (typically via the retailer’s authenticated user ID), you can trace a sale back to the ad exposures preceding it.

This attribution runs through tools (including Amazon Attribution, Walmart’s closed-loop reporting, and Target Roundel’s performance insights) or clean data rooms where retailer and advertiser data join without either side exposing raw records. The output yields SKU-level attribution: which products converted as a result of open web ad exposure, the frequency and recency of exposure, and the ROAS each campaign element contributed.

This level of granularity enables decisions that aggregate reporting can’t support: which audience segments convert better with open web exposure, which creative formats shorten the path to purchase, and which SKUs benefit the most from being shown off-site.

Track these metrics as your baseline:

1. Matched conversion rate (the percentage of users exposed to open web ads who converted on the marketplace within the conversion window).
2. Incremental sales lift (conversion rate among exposed users versus a control group that wasn’t served ads).
3. ROAS at the campaign and SKU level.
4. Cost per conversion compared to on-site marketplace placements.

Together, they tell you whether your open web spend is adding legitimate new sales, or claiming credit for conversions that would have happened on the marketplace regardless.

Key Takeaways

Marketplace inventory has a ceiling. More brands competing for the same placement on platforms like Amazon lead to rising CPCs and a shrinking incremental reach. Off-site retail marketing breaks that ceiling by taking marketplace audience data into open web environments where there’s less competition for the same shopper.

The strongest approach combines two data sources: the retailer’s first-party purchase data, with your own CRM and conversion signals. Alone, neither of these data sources gets you as far. Together, through data clean rooms or API-level integrations, they produce high-intent audience segments built on verified behavior and existing commercial relationships, all without any dependence on third-party cookies.

From there, execution decides whether reach converts to revenue. AI-powered bidding catches high-value impressions that a manual process would miss. Dynamic retargeting turns a generic ad into a specific product response. Closed-loop attribution through RMN measurement tools ties it back to SKU-level sales. Now you have the evidence (not just your gut) to scale off-site spend with confidence.

Frequently Asked Questions (FAQs)

What is off-site retail media targeting?

Off-site retail media targeting is the use of a marketplace’s first-party audience data (e.g., purchase history, product search behavior, and category browsing) to serve targeted display or video ads to those shoppers as they browse the open web. Advertisers reach them in content environments outside the platform, rather than waiting for them to return to the marketplace. This strategy keeps the advertiser’s brands in front of high-intent shoppers throughout consideration.

How do first-party integrations improve open web retargeting?

Server-to-server tracking, CRM list uploads, and data clean room connections let advertisers match known customers with open web ad inventory, all without third-party cookies. Because identity signals come from authenticated user accounts or CRM records, rather than browser-level tracking, audience match rates are higher, and the resulting segments are more accurate. Dynamic creative can reflect the specific products and browsing history of each matched user, which tends to outperform generic, non-personalized retargeting on click-through and conversion.

Can I track marketplace sales from open web ad impressions?

Yes, you can track marketplace sales from open web ad impressions. Closed-loop attribution (through RMN measurement tools) connects off-site ad exposure to conversions back on the marketplace and provides tracking data down to the SKU, time between exposure and purchase, and which campaign elements drove it. This strategy makes ROAS reporting possible at the product level, not just the aggregate campaign level.

Why expand performance campaigns beyond the marketplace?

On-site marketplace ads reach shoppers who are already on the platform and close to buying. Open web campaigns reach those same shoppers earlier in the session (during the discovery and consideration phases), when fewer advertisers are competing (and CPMs are lower). That incremental reach from open web placements reduces your dependence on saturated marketplace inventory; closed-loop attribution provides the evidence that off-site spend is adding to on-site performance, rather than just overlapping with it.