Retail media has multiplied the number of platforms a brand or agency needs to report on. Amazon, Walmart Connect, Instacart Ads, and Criteo each have their own console, their own API, and often their own definition of basic metrics like "spend" or "conversion." Teams end up building the same report five times a week, once per platform, then manually stitching totals together in a spreadsheet.
The core problem: every platform models data differently
The technical challenge isn't just pulling data from five APIs, it's reconciling them into one consistent model. Attribution windows differ. Some platforms report same-day conversions, others report 7-day or 14-day windows. Campaign structures don't map one-to-one. Without a normalization layer, a "unified" dashboard is just five dashboards pasted next to each other, giving a false sense of comparability.
What a real unified reporting layer requires
- A canonical data model campaign, spend, impressions, and conversions defined once, with each platform's data mapped into it
- Explicit attribution-window handling so cross-platform ROAS comparisons are actually apples-to-apples
- Scheduled multi-source ETL pulling from each platform's API on its own cadence and rate limits
- A single BI layer dashboards built once against the unified model, not once per platform
- Version-controlled metric definitions so "ROAS" means the same thing across every report a client sees
This is the architecture we built for BTR Media, a retail media platform helping brands manage campaigns across Amazon, Walmart, and Instacart from one system. The unified data model became the foundation for both the reporting layer and, later, cross-platform automation rules.
“By reducing manual effort and improving visibility, teams can scale campaigns efficiently while maximizing return on ad spend across every retail media platform they run.”Techesthete, on the BTR Media platform
Start with the metric definitions, not the dashboard
It's tempting to start a unification project by designing the dashboard first. The harder, more important work is agreeing on metric definitions before any data is pulled: what counts as a conversion, what attribution window is standard, how blended ROAS is calculated across platforms with different fee structures. Get that wrong and every dashboard built on top of it inherits the same inconsistency.
