Profiles Sync Tables and Materialized Views

Through Profiles Sync, Segment provides data sets and models that can help you enrich customer profiles using any warehouse data available to you.

Using a practical example of how Segment connects and then merges anonymous profiles, this page explains the tables that Segment lands, as well as the tables you materialize as part of Profiles Sync.

Case study: anonymous site visits lead to profile merge

To help illustrate the possible entries and values populated into Profiles Sync tables, view the event tabs below and consider the following scenario.

Suppose the following four events lead to the creation of two separate profiles:

// An anonymous visit to twilio.com triggers a Page call:

anonymous_id: 5285bc35-05ef-4d21
context.url: twilio.com
timestamp: May 2, 14:01:00

// Segment generates Profile 1, with a single known ID: 5285bc35-05ef-4d21

// Moments later, the same user signs up to Twilio with their email address.
// This triggers an Identify call:

anonymous_id: 5285bc35-05ef-4d21
context.url: twilio.com/try-twilio
timestamp: May 2, 14:01:47
email: jane.kim@segment.com

// Segment modifies Profile 1, adding an email address to the anonymous ID generated in Event 1.

// Weeks later, an anonymous visit to twilio.com triggers a Page call:

anonymous_id: b50e18a5-1b8d-451c
context.url: twilio.com/education
timestamp: June 22, 10:47:15

// Segment generates Profile 2, with a single known ID: b50e18a5-1b8d-451c.

// Moments later, the same user signs up for a Twilio webinar.
// This triggers an Identify call:

anonymous_id: b50e18a5-1b8d-451c
context.url: twilio.com/events/webinars
timestamp: June 22, 10:48:00
email: jane.kim@segment.com

// Segment understands that Profile 2 and Profile 1 are the same user.
// Segment merges Profile 2 into Profile 1.
// Profile 1 now has two values for anonymous_id: 5285bc35-05ef-4d21 and b50e18a5-1b8d-451c.

Initially, Segment generates two profiles for the first three calls. In the final event, though, Segment understands that Profile 2 should be merged into Profile 1. Segment then merges Profile 2 into Profile 1, merging away Profile 2 in the process.

Profiles Sync tracks and provides information about these events through a set of tables, which you’ll learn about in the next section.

Profile raw tables

Profile raw tables contain records of changes to your Segment profiles and Identity Graph over time.

With raw tables, you have full control over the materialization of Profiles in your warehouse, as well as increased observibility.

Raw tables contain complete historical data when using historical backfill.

The id_graph_updates table

The id_graph_updates table maps between the following:

• segment_id: the profile ID that Segment appends to an event or an identifier at the time it was first observed
• canonical_segment_id: the fully-merged segment ID (that is, the profile Segment now understands any events or identifiers to map to)

As a result, this table contains information about the creation and merging of profiles, as well as the specific events that triggered those changes.

Using the events from the profile merge case study, Segment would generate three new entries to this table:

In this example, the table shows profile_2 mapping to two places: first to itself, then, later, to profile_1 after the merge occurs.

Recursive entries

Segment shows the complete history of every profile. If, later, profile_1 merges into a different profile_0, Segment adds recursive entries to show that profile_1 and profile_2 both map to profile_0. These entries give you a comprehensive history of all profiles that ever existed.

If you’ll use Profiles Sync to build models, refer to the id_graph model, which can help you put together a complete view of a customer.

The external_id_mapping_updates table

This table maps Segment-generated identifiers, like segment_id, to external identifiers that your users provide. It has the following columns:

The anonymous site visits sample used earlier would generate the following events:

In this table, Segment shows three observed identifiers. For each of the three identifiers, Segment outputs the Segment ID initially associated with the identifier.

The profile_traits_updates table

The profile_traits_updates table maps each segment_id with all associated profile traits.

Segment updates this table:

• for each identify call that updates one or more traits for a segment_id.
• for any merge where traits from two previously separated profiles are now combined.

In the event that two profiles merge, Segment only updates the profile_traits_updates table for the canonical_segment_id, or the fully merged id.

From the profile_traits_updates table, use Segment’s open-source dbt models, or your own tools to materialize the profile_traits table with all profiles and associated profile traits in your data warehouse.

Event type tables

Event type tables provide a complete history for each type of event. Segment syncs events based on the event sources you’ve connected to Unify.

Identity Resolution processes these events, and includes a segment_id, enabling the data to be joined into a single Profile record.

Event type tables includes the following tables:

• Identify
• Page
• Group
• Screen
• Alias
• Track

These event tables are similar to the tables landed by Segment warehouse integrations, with the following exceptions:

• Events are combined in a single schema. For example, if you have three sources going into a single space, Segment produces one schema, not three.
• These tables have two extra columns:
  • segment_id: the profile ID at the time the event came through. That profile may have since merged.
  • event_source_id: the specific source ID of the incoming event

The previous result would generate two entries in the pages table:

And two entries in the identifies table:

All these events were performed by the same person. If you use these tables to assemble your data models, though, always join them against id_graph to resolve each event’s canonical_segment_id.

Profiles Sync schema

Profiles Sync uses the following schema: <profiles_space_name>.<tableName>.

If your space has the same name as a source connected to your Segment Warehouse destination, Segment overwrites data to the Event tables.

Track event tables

Track event tables provide a complete event history, with one table for each unique named Track event. Segment syncs events based on the event sources you’ve connected to Unify.

These tables include a full set of Track event properties, with one column for each property.

Segment’s Identity Resolution has processed these events, which contain a segment_id, enabling the data to be joined into a single profile record.

Tables Segment materializes

With Profiles Sync, you can access the following three tables that Segment materializes for a more complete view of your profile:

• user_traits
• user_identifiers
• profile_merges

These materialized tables provide a snapshot of your Segment profiles, batch updated according to your sync schedule.

Switching to materialized Profile Sync

If you’re not using materialized views for Profile Sync and would like to switch, follow these steps:

1. Enable Materialized Views through Selective Sync:
   • Navigate to Unify on the sidebar and select Profiles Sync.
   • Ensure you are viewing the Engage space you would like to enable materialized views for.
   • Go to Settings → Selective Sync and enable the following tables:
     • user_traits
     • user_identifiers
     • profile_merges
2. Request a Full Profiles and Events Backfill
   • After enabling the materialized views, you’ll need to ensure historical data is populated in the materialized tables.
   • Write to friends@segment.com and request:
     • A full Profiles Backfill to populate historical profiles data.
     • An Events Backfill to include any relevant historical events, including a date range for Segment to pull data in for the events backfill.
3. Verify Your Data
   • Once the backfill is complete, review the data in your warehouse to confirm all necessary historical information has been included.

Why materialized views?

Materialized views offer several advantages:

• Faster queries: Pre-aggregated data reduces query complexity.
• Improved performance: Access enriched profiles and historical events directly without manual joins.
• Data consistency: Automatically updated views ensure your data stays in sync with real-time changes.

The user_traits table

With the user_traits table, you’ll see all traits that belong to a profile, represented by the canonical_segment_id. Use this table for a complete picture of your Profiles Sync data with external data sources such as customer purchase history, product usage, and more.

• This view is a fixed schema, and contains a row for each trait associated with the profile.
• As new traits are added to the profile, new rows are added to the table.

When a merge occurs, two things happen:

1. Segment deletes the merge from profile in the table, along with with all the traits that belong to it.
2. Segment updates the merge to profile with the traits from the profile deleted in step 1.
   • For any conflicting traits, Segment appends the most recent trait to the profile.

This table has the following columns:

The user_identifiers table

The user_identifiers table contains all external ID values that map to a profile, which is represented by the canonical_segment_id.

With the user_identifiers table:

• There’s one row per identifier associated with the profile. This view has a fixed schema.
• As new identifiers are added to a profile, new rows are added to the table.

When a merge occurs:

1. Segment deletes the merge from profile in the view, along with all associated identifiers.
2. Segment updates the merge to profile with the identifiers that belonged to the profile deleted in step 1.

This table has the following columns:

The profile_merges table

The profile_merges table contains all mappings from a segment_id to a profile, represented by the canonical_segment_id. This mapping indicates that a profile has been created within Segment.

With the profile_merges table:

• There’s one row per profile associated with the canonical_segment_id that represents the profile. This view is a fixed schema.
• When a profile is created, a new row is created with the segment_id and canonical_segment_id having the same value.

When a merge occurs:

1. Segment deletes the merge from profile, along with all Segment IDs that belong to it.
2. Segment updates the merge to profile with Segment IDs that previously belonged to the profile deleted in step 1.

This table has the following columns:

Tables you materialize

You can materialize the following tables with your own tools, or using Segment’s open-source dbt models:

• id_graph
• external_id_mapping
• profile_traits

You might want to materialize your own tables if, for example, you want to transform additional data or join Segment profile data with external data before materialization.

Every customer profile (or canonical_segment_id) will be represented in each of the following tables.

The id_graph table

This table represents the current state of your identity graph, showing only where a segment_id is now understood to point.

The most recent entry for each segment_id from id_graph_updates reflects this. After the four example events, id_graph would show the following:

Segment drops most diagnostic information from this table, since it’s designed for reference use. In this case, you’d learn that any data references to profile_2 or profile_1 now map to the same customer, profile_1.

The external_id_mapping table

Use this table to view the full, current-state mapping between each external identifier you’ve observed and its corresponding, fully-merged canonical_segment_id.

In the case study example, you’d see the following:

The profile_traits table

Use the profile_traits table for a singular view of your customer. With this table, you can view all custom traits, computed traits, SQL traits, audiences, and journeys associated with a profile in a single row.

The profile_traits table contains the last seen value for any of your customer profile traits that Segment processes as an Identify call.

If Segment later merges away a profile, it populates the segment_id it merged in the merged_to column.

In the case study example, Segment only collected email. As a result, Segment would generate the following profile_traits table: