Our client, a large national CEE retailer with over 26,000 stores, collected detailed purchase-level data through its loyalty program, with each customer tied to a stable identifier, typically a phone number. Transaction history, including product mix, frequency, basket composition, price sensitivity, store location, and seasonality, was processed using internal data science models to derive attributes describing customer behavior and context.
These attributes represented inferred characteristics such as age group, purchasing power, household indicators, category affinities, visit patterns, and lifestyle signals. Some were straightforward, for example consistent purchases of baby products or pet food, or a sustained preference for premium brands. Others were more subtle, such as baskets that repeatedly combined distinctly “male” and distinctly “female” product categories, which often pointed to shared households rather than individual shoppers. All attributes were probabilistic, strengthening as patterns repeated over time.
The retailer wanted to improve its own marketing by enriching retail-derived attributes with signals from other enterprises and from different types of enterprises. Retail data alone provided strong insight into purchases and price sensitivity, but additional context such as mobility, media exposure, financial behavior, or brand interaction would allow for more precise targeting and personalization. At the same time, retailer wanted to offer its own attributes for external targeting and collaboration, allowing partners to reach relevant retail audiences.
However, this created a trust problem. Customer data was sensitive, regulated, and commercially valuable. Neither the retailer nor its partners were willing to share raw data, identifiers, or customer profiles, and no party was comfortable relying on a central intermediary. Any solution had to enable collaboration while keeping data local, controlled, and safe by design.
Our client used the CDP as a controlled layer for collaboration and monetization of retail data. Instead of sharing datasets, the retailer exposed internally derived attributes through the platform and allowed both internal teams and external partners to work with them when defining targeting logic. Each enterprise decided which attributes could be used and under what conditions, while all raw data and models remained inside its own infrastructure.
Advertising agencies and partners interacted with retailer’s data by defining segments as combinations of attributes, for example customers with high purchasing power who frequently buy specific product categories. When a segment was executed, the computation ran locally at the retailer and at other participating enterprises. The only output of this process was a set of phone numbers corresponding to customers matching the criteria, which could then be activated through marketing platforms such as Google or other channels.
This approach allowed the retailer to both enrich its own marketing with external signals and offer offline retail audiences for targeting, without exposing customer data or relying on a trusted intermediary. Collaboration happened at the level of logic and results, not data exchange, making it safe, compliant, and commercially viable.
