Same-session audience architecture and data handling

Learn how Hightouch evaluates same-session audiences, how realtime events work, what data is stored, and what kind of performance you can expect.

What you'll learn

After reading this article, you'll understand:

• How realtime events and your schema drive same-session evaluation
• How Hightouch combines live behavior with cached warehouse data
• What data is stored, why, and how long it's retained
• Latency characteristics, refresh behavior, and current limitations

Overview

This article covers how same-session audiences work under the hood—event processing, warehouse caching, data storage, and performance characteristics.

The diagram above shows how streaming events flow through Hightouch's evaluation engine, combine with cached warehouse attributes, and produce audience membership updates delivered via realtime syncs or the Membership API.

Key concept: event-driven evaluation

Same-session audiences are event-driven.

Rather than recomputing entire audiences, Hightouch:

• listens only to events defined in your schema
• evaluates only the user associated with each event
• uses a cached set of warehouse attributes
• updates audience membership incrementally

This approach avoids querying your warehouse for every event while still incorporating warehouse-backed logic.

How same-session audiences work

Same-session audiences rely on three components working together:

1. Realtime events (defined in your schema)
2. Audience logic (defined in Customer Studio)
3. Delivery (realtime syncs or Membership API)

Realtime events and your schema

Your schema defines the events that Hightouch evaluates in real time.

To track real-time behavior, you create a schema model that tells Hightouch:

• which event source to listen to
• which event type to capture (for example, track or page)
• which fields are available for filtering
• which timestamp represents when the event occurred
• how the event maps back to a user (for example, via user_id)

You can also apply filters so that only relevant events enter the evaluation pipeline.

Examples of realtime events:

• Page view
• Product searched
• Checkout started

Once defined, these events appear in Customer Studio and can be used directly in audience conditions.

In practice:

• The schema defines what events exist
• Audiences define how those events are evaluated

Audience conditions

Once a realtime event is available in the schema, marketers can use it in audience logic.

For example, an audience might include event conditions like:

• performed Page view at least 3 times in the last 10 minutes
• performed Checkout started in the last 5 minutes
• performed Search for running shoes during the current session

Same-session audiences combine live event conditions with warehouse-backed traits and attributes, such as:

• plan_type = free
• region = North America
• lifetime_value > 500

This allows evaluation of:

• current session behavior (events)
• historical context (warehouse data)

Membership availability

When a user enters or exits a same-session audience, Hightouch can make that membership available in two ways:

• Realtime syncs to supported destinations
• Membership API lookups from your application

Supported realtime destinations

For the full list of supported realtime destinations, see Create a realtime sync in the setup guide.

Evaluation flow

When a new event arrives, Hightouch evaluates only the affected user.

1. Event ingestion

Hightouch ingests streaming events from supported sources such as:

• Hightouch Events
• Segment
• RudderStack
• Snowplow

Only events that match a realtime event defined in your schema are processed for same-session evaluation.

If filters are configured on the event model, only matching events are evaluated.

2. User-level evaluation

For each event, Hightouch:

1. Identifies the associated user (for example, via user_id)
2. Evaluates whether the event contributes to any audience conditions
3. Retrieves relevant cached attributes from the warehouse
4. Recomputes audience membership for that user

If the user newly qualifies for an audience—or no longer qualifies—membership is updated immediately.

3. Membership availability

After evaluation, audience membership can be accessed in two ways:

Realtime syncs

Hightouch sends membership updates to supported realtime destinations.

When a user enters or exits an audience, Hightouch emits a membership event (e.g., User123 entered audience High-intent pricing visitors). These events can trigger downstream messaging, suppression, or activation workflows.

Supported destinations receive audience membership changes immediately after evaluation.

Depending on the destination configuration, this may include:

• Audience entry events when a user qualifies for an audience
• Audience exit events when a user no longer qualifies
• Mapped events or properties configured in the sync

Membership API

Applications can query current same-session audience membership via a low-latency API.

The API accepts a user_id and returns the list of qualifying audience IDs.

Common use cases:

• website personalization
• in-app messaging
• feature flagging
• session-based targeting

Same-session audiences and Personalization API

Hightouch supports two complementary systems for real-time use cases:

• Same-session audiences determine whether a user currently qualifies for an audience based on live session behavior.
• Personalization API retrieves warehouse-backed profile attributes used to personalize the experience.

These systems often work together in a single request flow.

Example flow:

1. Call the Membership API to check if a user qualifies for an audience
2. If they qualify, call the Personalization API to retrieve profile attributes
3. Use both to determine and personalize the experience

Data storage

To support low-latency evaluation, Hightouch caches some warehouse data into the Hightouch platform. Your warehouse remains the source of truth.

Event evaluation state

Raw event payloads are not retained after aggregation.

Warehouse attribute cache

Examples of cached attributes include:

• plan_type
• region
• loyalty_tier
• lifetime_value

This cache allows Hightouch to combine live session behavior with profile context already defined in your warehouse.

Audience membership store

Latencies

These are target performance characteristics, not guarantees.

Constraints and limitations

Same-session audiences are optimized for fast, session-based audience evaluation.

Current constraints:

• Realtime event conditions must include a lookback window
• Maximum lookback window is 24 hours
• Only event count conditions are supported for realtime evaluation today (no arbitrary aggregations like SUM yet)
• User property conditions (e.g., from identify calls) must go through the warehouse; they cannot be evaluated in real time via streaming

Example supported patterns:

• Viewed a page 3 times in the last 15 minutes
• Searched for a product during the current session
• Started checkout in the last 5 minutes

Key storage principles

Same-session architecture follows a warehouse-first model:

• Your warehouse remains the system of record
• Hightouch stores only the minimum state needed to evaluate same-session membership
• Full customer profiles are not copied into the same-session engine
• Event state expires automatically based on the configured lookback window
• Data is isolated by workspace within Hightouch-managed infrastructure
• Same-session processing infrastructure runs in US East and EU West regions
• Bring Your Own Bucket (BYOB) is not currently supported for same-session audiences

What's next?

After reviewing the architecture, the next step depends on your role.

If you're configuring schema and event models

• Define data schema →
  Set up parent, related, event, and realtime event models for Customer Studio.

If you're building audience logic

• Audiences →
  Create audiences using realtime events, traits, and warehouse-backed filters.

If you're activating same-session audiences

• Set up same-session audiences →
  Configure realtime events, delivery patterns, and downstream activation.

If you're personalizing experiences

• Personalization API →
  Retrieve warehouse-backed traits and profile data for live experiences.