Data Modeling, Identity, and Stitching
What you must understand
CJA's power is cross-device and cross-channel analysis, which depends on identity. Know the concept of the person ID (the namespace used to identify a person) and how it is set at the connection level. Understand stitching: when a person is known on some events (authenticated) and anonymous on others, stitching uses a temporary ID to link anonymous events to a known identity within a defined replay window.
Be able to explain the difference between field-based stitching and graph-based stitching, and when each applies. Know that stitching requires a defined identity namespace and a lookback/replay window, and that stitched data is not retroactively rewritten-new data is stitched going forward.
Traps to watch
- Assuming stitching fixes historical data automatically; it does not rewrite past events.
- Confusing the person ID with the marketing channel or with a visitor ID.
- Forgetting that identity must be present in the schema and mapped in the connection.
- Overlooking that stitching has limits (window length, namespace availability).
Rule of thumb
If a question is about linking anonymous and authenticated events into one person, the answer is stitching with a person ID namespace and a replay window. If it is about which field identifies the person, it is the person ID in the connection.