Real-Time Segmentation in Data 360

"Real-time segmentation" is easy to ask for and easy to misunderstand. Some Data 360 paths are fast enough to react within a session. Others run every 15 minutes, every hour or on a publish schedule. Knowing which is which decides whether a use case works.

This page explains streaming and batch segments in Data 360, what Salesforce documents as near real time, the use cases that justify it, the cost considerations, and how we test it. For building segments in general, see our segmentation and calculated insights service.

The business problem

A customer abandons a basket, buys the product you are about to promote, or raises a complaint, and your marketing still treats them as they were yesterday. Batch segments are fine for a weekly newsletter. They are not fine for in-session offers, post-purchase suppression or service-aware messaging.

What is near real time, per Salesforce

The real-time layer accepts the Web SDK, Mobile SDK and server-to-server ingestion. Real-time data graphs "are required to use other real-time capabilities", which include real-time identity resolution, calculated insights, segments and data actions. Salesforce's real-time ingestion page gives no beta or GA label, so we confirm the current state for your org before design.

Proposed architecture

There is no single diagram for this page; real-time segmentation sits inside the Data 360 reference architecture and the real-time personalisation architecture. This is a reference design, validated per implementation.

Use cases that justify real time

• In-session personalisation: change content or offers while the visitor is still on the site. See Salesforce real-time personalisation.
• Post-purchase suppression: stop promoting a product the customer has just bought.
• Abandonment triggers: react to a basket or form left incomplete.
• Service-aware messaging: hold promotional messages when a case is open.

If a use case only needs same-day accuracy, a batch segment on the standard path is simpler, cheaper and easier to debug.

Data flows and integration patterns

Events flow from the SDK into the real-time layer, are matched to a profile through the real-time data graph, and update segment membership or trigger a data action. Batch data still matters: real-time segments can combine a live event with attributes that arrive on the standard path. Know which attributes are fresh and which are hours old. For transforms, see batch vs streaming transforms; for identity, see real-time identity resolution and data graphs.

Security boundaries

The Interactions SDK sends events only if a customer has consented to tracking and stops when consent is revoked. Keep that behaviour intact in your tag management, and keep real-time consumers inside the same permission model as batch ones.

Failure handling

Real-time paths fail differently from batch ones. Watch for SDK tags removed by a site release, consent banners that block events unexpectedly, data graphs missing an attribute the segment depends on, and consumers that cannot keep up. Every real-time use case needs a defined fallback, such as default content or a batch segment.

Cost and capacity considerations

Real-time features process data continuously, so their consumption can differ from, and may exceed, a nightly batch. Salesforce does not publish these costs on the pages we checked, and we do not quote prices. Practical guidance: model consumption with your Salesforce account team, limit real-time data graphs to the attributes a decision needs, and keep identity resolution scheduling deliberate. Salesforce notes that turning off "Run jobs automatically" can reduce identity resolution cost.

Deployment model

We configure in a sandbox or test Data 360 environment, working remotely from India. Site tags go through your web release process, not ours.

Testing strategy

Measure latency from the browser event to the consumer's action, not just inside Data 360. Test with consent granted, refused and revoked. Run a load test at expected peak, and test the fallback by disabling the real-time path.

Operational monitoring

Track event volumes by source, the share of events matched to profiles, segment membership changes and consumer errors. A sudden drop in events usually means a site change, not a change in customer behaviour.

Prerequisites and limitations

Data 360 must be provisioned, with real-time data graphs for real-time features. Winter '27 adds Agentic Segmentation and Activation, a conversational way to build audiences; Salesforce has not stated its status, so we do not rely on it for production design yet.

Sources

• https://developer.salesforce.com/docs/data/data-cloud-ref/guide/c360a-real-time-ingestion-api.htm
• https://trailhead.salesforce.com/es/content/learn/modules/real-time-use-cases-in-data-cloud/explore-real-time-features-in-data-cloud
• https://developer.salesforce.com/docs/data/data-cloud-int/guide/c360-a-mobile-web-datastream.html
• https://developer.salesforce.com/docs/data/salesforce-interactions-sdk/guide/c360a-api-salesforce-interactions-web-sdk.html
• https://trailhead.salesforce.com/content/learn/projects/explore-data-cloud-core-functionality/unify-your-data