By Julian Weidner
In a world where online marketing channels thrive on granular tracking and data-driven attribution, in-store retail media has historically lagged behind. No more. Today, advanced technologies—from computer vision with on-device processing to edge computing and heatmaps—are helping retailers and brands connect the dots between an in-store impression and a final purchase.
The Evolving Landscape of In-Store Data Analytics
Traditionally, measuring in-store effectiveness meant approximate sales lifts or simple test-versus-control studies. With the rise of AI-driven displays and on-premise analytics, attribution in physical spaces has become far more precise. Retailers can track dwell times, foot traffic, and shopper interactions, often storing data locally to comply with privacy regulations and reduce network latency.
Laying the Foundation: Attribution & Incrementality:
Before diving into advanced methods of in-store data collection, let’s clarify two essential concepts:
Attribution: Assigning credit for a sale or conversion to a specific marketing touchpoint—in this case, a digital display or interactive endcap.
Incrementality: Measuring the additional sales or conversions directly attributable to the campaign—above and beyond what would have occurred otherwise.
In an in-store environment, this involves pairing exposure data (did the shopper see the ad?) with purchase data (did they buy the featured product?). Techniques like test/control store splits and time-based A/B testing help prove incremental lift.
Modern Technologies To Track Shopper Impact
Thanks to computer vision, machine learning, and edge computing, marketers now have a robust set of options to link ad exposures with real-world outcomes. Devices such as Azure Kinect DK, Nvidia Jetson Nano, and Google Coral Dev Board enable advanced data gathering methods with relative ease.
Computer Vision
On-Device Video Analysis: In-store cameras can perform anonymized customer tracking dwell-time, gaze, and sentiment analysis. By tracking each shopper during their visit to the store, retailers can connect in-store ad exposure to purchase behavior in a privacy-friendly manner.
Heatmaps & Traffic Flow Analysis
Overhead Cameras: Placed strategically around the store, these capture real-time footage to help visualize foot traffic and shopper movement patterns. By identifying hotspots (areas of high foot traffic or dwell time), retailers can reposition signage for maximum exposure or adjust content timing based on observed patterns.
Linking Shopper Engagement to Attribution and Incremental Sales
Lets run through a hypothetical scenario: A supermarket deploys an in-store retail media solution that is tied to computer vision based cameras. An in-store advertisement display is booked for a campaign. The data gathering process could look something like:
In-store Session Tracking Label
The system assigns a temporary label, customer_1, for their shopping session. This label is created for each customer entering the store. It is then deleted upon exiting the store. Preventing any long-term link to a particular individual. No personal data, such as biometric identifiers, are captured.
Engagement Detection
A camera at an endcap observes how shoppers interact with an advertisement. If a shopper is determined to 'gaze' at the ad for ~three seconds, they are then classified as “engaged.” The raw video is processed locally in real time, and no footage is stored.
POS Match
A second camera at the checkout uses the same short-lived label to record if customer_1 ended up purchasing the promoted product. Once the transaction is complete, the label is disconnected from the receipt data, allowing only aggregated insights (e.g., “One shopper who engaged with Display A bought Product X”).
Attribution Analysis
The system establishes a direct link between shopper engagement and purchase behavior by tracking whether customer_1, who interacted with the advertisement, ultimately buys the promoted product. This approach enables advertisers to measure how ad engagement correlates with actual sales, providing a transparent attribution model for in-store retail media. Incrementality Analysis
To measure whether the advertisement caused additional sales beyond organic purchasing behavior, an incrementality test is required. This involves a control group. There are a few ways this could be done.
Geo-based holdouts, where some stores run the display and others do not.
Time-based holdouts, where the display is active during certain periods and inactive during others.
Randomized suppression, where a subset of shoppers is intentionally prevented from seeing the ad.
By comparing purchase rates between exposed and control groups, advertisers can determine the true incremental lift generated by the advertisement. This ensures that reported sales impact is causally linked to ad exposure, rather than being a result of natural shopper behavior.
Balancing Data Accuracy With GDPR
While advanced tracking makes for powerful analytics, it also raises legitimate concerns around privacy and data protection. Noncompliance with regulations like GDPR can lead to direct legal consequences, including fines of up to 4% of a company’s worldwide annual turnover. Additionally, the reputational fallout from perceived privacy violations can be just as damaging—customers may lose trust, voice their dissatisfaction publicly, and ultimately shift their loyalty elsewhere. Thus, beyond legal obligations, transparent and ethical handling of data is key to maintaining both compliance and consumer confidence. The example above was written with GDPR in mind.
No Personal Data: Because the camera generates only a short-lived label with no persistent ID or biometric info, there’s no way to identify an individual.
Local Processing: All detection and labeling occurs on-premise, mitigating privacy risks and lowering reliance on external servers.
Session Delete: By deleting session at the end of the shopping session, the system avoids building ongoing profiles of repeat visitors, remaining well outside the scope of personal data under GDPR.
Aggregated Insights: Receipt data is used solely to calculate incremental sales; once the purchase is confirmed, the shopper label is discarded. Only aggregated, anonymous stats remain, such as “X shoppers engaged and Y of them purchased."
Data accuracy and GDPR compliance don’t have to be at odds; with the right technical approach and clear communication, retailers can deliver meaningful metrics that power better shopper experiences—without compromising consumer trust.
In-store retail media is no longer at a disadvantage compared to online marketing channels. By leveraging on-premise computing, daily-reset session labels, and local, ephemeral tracking (e.g., a short-lived “customer_1” label), retailers can measure incrementality and attribution with remarkable precision—yet still minimize privacy risks. Cameras process video in real time, discarding raw footage and replacing it with anonymous, short-lived identifiers that never link to personal data or biometric details. Once a shopper’s purchase is confirmed, the temporary label is stripped away, leaving only aggregated statistics to guide Return on Ad Spend (ROAS) analysis. This approach provides actionable insights—similar in depth to online analytics—while honoring GDPR’s principles of data minimization, transparency, and privacy by design.
Commentaires