Profile fragments and the union schema, in plain English

Why this matters

Understanding profile fragments and the union schema matters because it helps analysts trace where data comes from. Without this clarity, it's hard to know if a value in a customer profile originated from a specific source—like a website visit or a mobile app. This ambiguity can lead to misinterpretations of customer behavior and flawed targeting decisions.

If the union schema isn't properly configured, fields from different datasets may not combine correctly, causing missing or duplicated data. This can result in incomplete or inaccurate profiles, which in turn affects campaign performance and personalization accuracy. Knowing how fragments are merged and which schema fields are included ensures data integrity and trust in the insights derived from the Real-Time Customer Profile.

The key ideas

A Real-Time Customer Profile builds one unified view of a person by gathering data from many sources. Each source contributes a piece of data called a fragment — the part of a person’s information stored in that specific dataset. These fragments may contain different details, like a phone number from a marketing campaign or a location from a mobile app.

The union schema combines all the fields from different datasets that belong to the same type or class. For example, if multiple datasets track email addresses, their fields are grouped together in the union schema so that all relevant data types are available in the profile. When a new dataset is added to Profile, its fields are automatically included in this union schema.

To build a complete profile, the system uses the identity graph to find all matching fragments across sources. Then, a merge policy decides how to resolve conflicts — for instance, which email address to keep if two sources provide different ones. Looking at the fragments of a single profile shows exactly where each value came from, making it easy to trace data origins and understand how the profile was built.

How to apply it

To apply this in practice, start by identifying which datasets are Profile-enabled. Each of these contributes a fragment—specifically, a piece of data about a person stored in that source. When multiple datasets hold data about the same person, those fragments are collected and combined. The union schema plays a key role here: it combines the fields from all Profile-enabled schemas that belong to the same class, creating a unified structure that defines how data is grouped and matched.

When a new schema is enabled for Profile, its fields are automatically added to the union schema, expanding the scope of data available in the merged profile. This means the profile structure grows to include more data points, but only if the schema shares the same class.

To see where a value originated, examine the fragments of a specific profile. This gives you direct visibility into which source contributed each piece of data. For example, if a customer’s email appears in a marketing dataset, you can trace that fragment back to its origin. This transparency helps in auditing data sources and ensuring accuracy when building customer journey insights.

Mistakes to avoid

• Avoid assuming that all data in a profile comes from a single source. Each fragment represents data from a specific dataset, and relying on one source can lead to incomplete or inaccurate views. Always check the fragments to confirm where a value originated.

• Do not treat the union schema as a static structure. When a new schema is enabled, its fields are added to the union schema, altering the available data fields. Without updating your understanding of the schema, you may miss new data points or misinterpret existing ones. Regularly review the union schema and associated fragments to stay aligned with current data sources.

Quick checklist

• Verify that the dataset you're using is Profile-enabled, as only such datasets contribute fragments.
• Confirm that the new schema you're adding has been properly enabled and will contribute its fields to the union schema.
• Check that the identity graph is active and configured to link profiles across datasets.
• Ensure your merge policy is set to handle conflicts appropriately, especially for sensitive or critical fields.
• Review the profile's fragments to confirm which data sources are contributing values to the merged profile.
• Validate that the union schema includes all required fields from all relevant datasets sharing the same class.