Retail Media’s Audience Fragmentation Between Online and In-Store – and How AI Can Solve It

Published, March 18, 2025

In 2026, we will look back and say: Retail Media is stuck. Except we ought to face and solve the next big challenge right now: unifying addressable audience segments across networks as well as across online and in-store channels. Despite the rapid growth of Retail Media, most brands are forced to operate in silos, with in-store and digital campaigns managed separately, limiting the effectiveness of Retail Media investments.

I recently read an article by Greg Deacon on InternetRetailing about the organizational hurdles of breaking down silos in Retail Media. He makes a strong case for how internal brand structures are often the biggest blockers to true cross-channel integration. That’s the organizational side of the problem. Here, I’m focusing on the tech side of the problem – how unifying audience segments across online and in-store is the only way to truly break these silos and create a seamless omnichannel media buying experience.

Even when we look at In-store Retail Media alone, the industry is still deploying legacy DOOH technology without questioning whether it can actually deliver the value, precision, and scalability that made online Retail Media so successful. For a deeper dive into why in-store Retail Media cannot simply replicate DOOH, check out my blog post from 2023.

Since then, a lot has happened: the CAPEX for sensor-based hardware kits has significantly decreased (by more than 70%), and major leading Retail Media organizations have started deploying technologies enabling real-time audience segmentation, activation, and measurement in-store. The most recent example is MAF’s Precision Media in their Carrefour UAE stores. 2024 was a year of progress because first-movers initiated this wave of innovation. And as we know, every innovation needs bold trailblazers, and the early majority will follow the first-movers soon after.

Now, building on the progress made on real-time in-store audience segmentation and activation, it’s time to look ahead and focus on the next big challenge: unifying audience segments across online and in-store to create a truly seamless omnichannel media buying experience.

In-store Attribution Is Needed But Not Enough

Currently, the industry’s biggest attempt to bridge the gap between in-store and online shopping is linking loyalty card data to purchases at physical checkout and linking back to online shopping history. That’s a good step and, fortunately, from a tech perspective, not a big challenge. However, it misses a crucial component of the bigger picture: real-time addressability and activation of the in-store shopper during the shopping session based on unified cross-channel audience segments.

To really bring digital advertising’s power into the physical store, we need more than just checkout-based attribution. Checkout data only captures completed purchases, failing to account for in-store browsing behavior, intent signals, and real-time engagement – elements that have been critical to Online Retail Media’s success.

DOOH Technology Is Needed But Not Enough

Right now, In-store Retail Media is defined by DOOH technology, not by what actually works for omnichannel advertising. The industry assumes that because digital screens are needed for In-store Retail Media, the related DOOH technologies are the only option to invest in without any changes or add-ons. That’s like saying banner ads targeted based on zip codes are the future of ecommerce advertising. It ignores the core mechanisms that made Online Retail Media dominant: real-time, retail/shopper-data-driven, audience-based activation.

The problem? DOOH technology was built for a different purpose. It wasn’t designed for real-time audience activation. It wasn’t designed to unify audience segments across online and in-store environments, and it certainly wasn’t designed to integrate with the sophisticated data-driven buying experiences that brands and agencies now expect. The result is that in-store advertising remains disconnected from the precision, measurement, and flexibility that define modern Retail Media.

The Current Market Thinking: A Fragmented Mess & How To Clean It Up

Let’s examine how the industry is approaching audience segmentation in Retail Media Networks (RMNs). Right now, the primary focus is on isolated first-party data within each retailer’s walled garden. Every RMN operates its own unique audience segmentation, bidding system, and measurement framework. As a result, media buyers face a fragmented landscape where they must manually reconcile audience definitions across hundreds of different RMNs, most built on different tech stacks.

The ad tech industry has recognized this problem and is already working on mapping audience segments across these walled gardens. The challenge for demand aggregators is that isolating and matching user IDs doesn’t work across Retail Media walled gardens due to privacy restrictions, retailers’ lack of willingness to open up, and fragmentation between networks. Isolating and matching IDs might still be viable for offsite activation but not for unifying audiences across Retail Media Networks. Even for offsite activation, the majority of targeting still relies on lookalike modeling rather than one-to-one ID matching.

The only remaining path forward for cross-RMN unification is using AI to match audience segments across networks. AI-driven generative modeling enables intelligent segment predictions, while AI-driven probabilistic modeling facilitates audience matching without requiring deterministic identifiers. Online innovators in our space will use AI-driven lookalike modeling to map audience segments across different Retail Media Networks, addressing the fragmentation issue. These efforts mark an important step toward solving the online part of the problem.

A similar challenge applies to in-store: there is no real-time access to a shopper’s profile during the in-store journey. So, if we are mapping audiences across walled gardens using AI, why aren’t we extending this logic to the physical world, where ID matching is neither feasible nor scalable?

The answer is simple: it’s possible. Using AI to predict in-store shopper segments in real-time and match them with already offered and monetized online segments is the only path to truly unified omnichannel Retail Media.

Unifying Cross-Channel Audiences Across Online & In-store

The good news is that, considering in-store as an isolated channel, the technological solution is already developed, deployed in major global store networks, and applied to hundreds of campaigns daily. This is the tech I explained in my introduction. The same approach can be applied to predicting to which online audience segments individual in-store shoppers belong and making them addressable based on those.

And this is how the four-step process works:

• Capture: The foundation of everything is capturing the right data in a privacy-compliant way and transforming the physical shopper into a digital profile on the local Edge-AI computer in real-time. Only computer vision can deliver the quantity and quality of data signals to understand the full spectrum of shoppers and their behavior in-store. No other technology can provide real-time insights into shopper movement, engagement, and other relevant insights.

• Segment: Using visual attributes such as age, gender, and group composition (e.g., couple, solo shopper, family with children), the AI estimates shopper profile attributes in real-time. Remember: in-store, there is no access to the shopper profile in real-time, so an AI has to predict the attributes. Combined with location-based contextual attributes, the AI – trained on historical retailer data – predicts shopper segments and aligns them with online audience segments.
  This is the key to unifying audience segments across in-store, online, and different Retail Media Networks. It enables a seamless, standardized, and unified audience structure that allows media buyers to execute true omnichannel campaigns without fragmentation.

• Activate: This is one of the most complex challenges to solve. The core question the AI must answer is: How many impressions will the next ad spot on a given digital signage screen generate, and which audience segments will it reach? This is a challenge with multi-dimensional complexity:

1. 1. Timing & Ad Activation: Unlike online advertising, where an ad is triggered when a user loads a page, in-store digital signage operates continuously. The AI cannot passively wait for a trigger – it must be active at all times, analyzing the environment well before the targeted ad spot begins. The next spot is initiated not by direct action but by the conclusion of the previous one.
   2. Dynamic Shopper Movement: Shoppers constantly move about rather than remain stationary in front of the screen. If an ad lasts 10 seconds, the AI must predict who will be within the screen’s viewability field for the duration of that spot, factoring in movement patterns and dwell times.
   3. 1-to-Some Addressability: Unlike online advertising, where each ad delivery corresponds to exactly one impression (hence the term “ad impression”), in-store advertising operates on a 1-to-some model. A single ad play can and should reach multiple shoppers simultaneously and generate multiple impressions.

To solve this challenge, computer vision leverages 3D space perception, walking path predictions, and Predictive AI to forecast audience composition for the next spot. Then, approximately half to one second before the ad plays, the AI informs the digital signage CMS, ad server, and SSPs about the audience composition for the next spot and recommends what available segment to target, ensuring maximum relevant impressions.

• Measure: By aggregating such in-store audience data with retailer’s transaction and shopper data, advanced audience and performance analytics are generated, enabling true omnichannel media performance insights. Key upper-funnel media metrics include the number of impressions, impressions within the screen’s view field, dwell times, and view times.
  This ensures comparability between online and in-store media measurement and delivers deeper audience insights than online. Additionally, because audience segmentation is unified across online and in-store, consistency is maintained even at the segmentation level. You can read about how Advertima solves the measurement gap in our In-store Retail Media Standards post.

The Future of Omnichannel Retail Media Is About Unifying Audiences Across Channels and Networks

Retail Media won’t reach its full potential until it embraces audience unification across online and in-store. Capturing, segmenting, activating, and measuring audiences seamlessly based on unified segments across both environments will create a genuine omnichannel media buying experience and unlock new revenue streams. This shift will ultimately define the next phase of Retail Media’s evolution. 

The technology is here. The buying experience is evolving. But adoption is not a given. It requires innovators and first-mover Retail Media executives to break down existing silos and embrace AI-driven audience matching across online, in-store, and RMNs. 

The key challenge now is whether retailers and tech providers will take the necessary steps. Media buyers are certainly ready to benefit from such advancements.