Composable customer data platform
Petabyte-scale, first-party tracking events, cross-device and cross-brand identity resolution. Multi-brand activation.
- Petabytescale data platform
- Multi-brandidentity resolution
- First-partytracking events
The Group operates a multi-brand portfolio. Customer data was siloed per brand, often dependent on Google Analytics or Firebase, and resolved poorly across devices. The data didn't support the use cases the business already had — let alone the ones we wanted (cross-brand audience activation, multi-channel attribution, fraud signals across the ecosystem).
The decision: own the foundation. Build a composable Customer Data Platform (CDP) from first-party events, with identity resolution as a core service rather than an afterthought.
Three composable tiers. Each layer is replaceable; the contract between layers is the dataset.
- Foundation: First-party User Tracking Events (instrumented directly, not borrowed from Google Analytics or Firebase). Cross-device identity stitching plus cross-brand resolution across the multi-brand portfolio.
- Applications: Attribution modelling. Recommendations features (the customer-affinity vectors that fed Tensorflow Recommenders). Audience targeting for owned and paid channels.
- Activation: Braze for email and lifecycle. Social retargeting integrations (Meta, Google). Third-party data partner integrations for activation outside owned surfaces.
- Data governance framework — quality service-level agreements, schema review, lifecycle ownership — implemented in parallel so the platform stayed trustworthy as more applications consumed it.
Three-tier composable CDP: foundation (events + identity) → applications → activation.
Platform reached petabyte scale with improved data availability, governance, and reporting reliability across business units. Cross-brand customer journeys became visible for the first time — previously invisible interactions across the portfolio could now be reasoned about.
Scalable analytics architecture supported flexible reporting across the Group: same foundation, different application configurations. Merchandising, marketing, and fraud each consumed the platform without colliding.
The framing as 'composable' rather than 'best-in-class single CDP' was deliberate. Best-in-class single vendors couldn't handle the multi-brand identity case without compromising on at least one stack we already had. Build the foundation; let applications evolve.