ARCHITECTURE
Most companies run more architecture than they need. Some run less.
Every customer data architecture we’ve seen fits one of three patterns: the platform does everything, the platform pairs with a data warehouse, or the platform activates decisions made elsewhere. Each level is right for someone. The expensive mistake is building level three when you’re a level one company, or forcing level one to do a level three job.
THE THREE LEVELS
Three patterns cover almost every stack.
Level 1. One platform does it all.
Tracking, customer profiles, analytics, and campaign execution inside one application, typically Bloomreach Engagement. Right for most mid-size brands, lean teams, anyone whose reporting needs fit inside the platform.
Level 2. Platform plus data warehouse.
The triangle: your site and apps feed tracking into the platform, raw data flows onward into a warehouse such as BigQuery, and computed results flow back for activation. Right for brands with analysts, multi-source reporting needs, growing event volumes, or data the CDP should not hold.
Level 3. Enterprise decisioning stack.
A central data platform owns the data, a decision engine computes next best offer and next best action, and Bloomreach Engagement is the activation layer that delivers those decisions across channels in real time. Right for enterprises with data teams, regulated industries, and organizations where marketing is one consumer of a shared customer data asset.
The AI layer.
AI isn’t a fourth level. It attaches to whichever level you run: inside the platform, on the warehouse, or as the decision engine itself.
A note on names: a CDP is the customer data foundation. A platform that adds automation, personalization, and journeys on top, like Bloomreach Engagement, is a customer data and experience platform (CDXP). The three levels apply to both.
HOW TO CHOOSE
Pick the smallest level that answers the questions you have today.
The right level is the one that answers the questions you’re actually asking now, not the ones you might ask in three years. Build ahead of your questions and you pay for infrastructure nobody uses. Here’s the honest test.
If this is you: your marketing team gets what it needs from the platform’s built-in reports, and nobody is writing SQL against your data — you’re probably Level 1.
If this is you: a data or BI analyst keeps hitting the limits of the platform’s reporting and needs to join data from other systems — you’re probably Level 2.
If this is you: you have a data team, and departments beyond marketing (service, risk, finance) all consume the same customer data — you’re probably Level 3.
The two ways to get it wrong. A warehouse or a decision engine nobody queries is spend with no consumer. Forcing one platform to do the job of three is the opposite mistake, and you feel it as manual work and numbers that don’t reconcile. In practice, most teams we meet are already at the right level, or a step above what they need.
Moving up is a project, not a rebuild, when the level below was built cleanly. That’s the real reason to get the data model right at level one: it keeps the next level a step, not a restart.
WHAT WE DO
We design, build, and run all three levels.
We’ve implemented level one across 40+ accounts, connected platform-to-warehouse pipelines for brands whose reporting outgrew the platform UI, and delivered enterprise CDP work for banking and telecom clients. We’ll also tell you, in the first call, which level you actually need. Often the honest answer is a smaller one than you expected.
Not sure which level you are?
Book a discovery call. Describe your stack and your questions. We’ll place you on the map and tell you what we would build, or what we would remove.