Experience Platform · Profile and data lake
Profile or data lake?
Laura’s shop stores browsing, baskets and purchases. Not all of that data needs to occupy Profile. Choose what stays there and compare volume with the uses you retain. In this example, CJA still reads every event from the data lake.
Example assumptions
Example shop, volumes, event sizes and licence. Each event type goes to its own dataset, all with the same expiry, and 90 days are counted as 3 months.
Licence · Total Data Volume
What counts towards the licence
Only what sits in the Profile store counts. The data lake does not. Example licence: 900 GB.
Use cases · requirements
What still works
Each use case reads from one place. If the data is not there, or its window does not fit the expiry, it fails.
Architecture
Where it fits in the platform
Map of Adobe Experience Platform with Journey Optimizer and Customer Journey Analytics, top to bottom in five stages. Capture: the Web and Mobile SDK, on the client or the server, sends XDM events, and systems such as the CRM or the warehouse deliver files. Ingestion: the Edge Network receives the XDM event in its datastream and streams it to Profile, and also to Data Lake datasets. Sources receives files and API calls through its connectors and Data Prep, and in these examples loads them into the Data Lake in batch, although it also supports streaming connectors. Profile and Data Lake: Identity Service links identifiers from several namespaces in a graph and passes it to Profile. Profile joins attributes and events in real time, and is fed by the enabled datasets of the Data Lake, which keeps the history. Audiences and governance: Segmentation receives the profiles and evaluates audiences with three methods. Governance applies labels, policies and consent. Applications: audiences go out to Destinations, in Real-Time CDP, where Governance checks its policies at activation. Journey Optimizer receives profiles and events to act in its journeys in real time. Customer Journey Analytics ingests the Experience Platform datasets selected in a connection and exposes them for analysis through its data views.
The recommendation
Only what activates or personalises goes into Profile. The rest stays in the data lake
Total Data Volume measures what you keep in Profile, not what you collect. An event that no audience or journey uses takes up licence and adds nothing.
- 01
Profile holds what activates
Product views and purchases, for audiences and to know whether someone already bought. The cart event triggers the journey even if it is not in Profile. What feeds no audience and no condition does not need Profile.
- 02
The rest goes to a data lake only dataset
A datastream configuration override sends those Web SDK events to another event dataset that is not enabled for Profile. That dataset must be configured in the datastream first.
- 03
Expiry follows the longest window
Repurchase after 12 months needs at least 13 months. Review the audiences before shortening it: deletion from Profile is permanent, and a window that does not fit gives wrong results.
- 04
CJA reads everything from the data lake
The CJA connection reads Experience Platform datasets whether or not they are enabled for Profile. An audience that does not need real time can come from Data Distiller.
Explore this decision
- What needs to leave the warehouse?
Ingestion, Federated Audience Composition and Data Mirror solve different problems. See what moves and what to check before choosing.