Data Cloud July 2025 Release Highlights

Things are heating up this July as Data Cloud announced the pending release of 19 new features and enhancements. Some of the most exciting updates are the ability to declare a Secondary Index within a DLO, improving query performance and reducing costs for high-volume use cases, Data Cloud to Data Cloud sharing via Zero Copy allowing for more connectivity and new use cases across orgs, the introduction of image processing for unstructured databases as well as increased file support and new and improved chunking methods for text, lifted restrictions for activation paths in activation, and the exciting introduction of household support for identity resolution. Read the summary below for a concise run-down of each feature, or watch the full 4 hour video on the Data Cloud Video Hub.

Granular Consumption Data in Digital Wallet: Phased Release

Starting this release, the Digital Wallet will begin to phase out its aggregated consumption metrics in favor of more granular, resource-based metrics. With this new level of detail, admins will be able to understand with greater precision the processes which are consuming data cloud services credits. Admins and analysts will be able to make use of new fields which track the name and ID of each resource consuming credits. Since this data is available in a DLO it can be used for custom reporting or even calculated insights to track, detect, and act upon anomalous behaviors.

View Semantic Layers in Companion Orgs

Data Cloud One enables multiple Salesforce Orgs with the ability to view, edit, and collaborate on the data managed in the Data Cloud. With this newest update, the Semantic Layer is now viewable from companion orgs, enabling Tableau Next and Tableau Cloud reporting. The ability to edit and create models will be available in an upcoming release.

Security Improvements: Data Resilience Managed Capability (DRMC)

In order to maintain the highest standards regarding data availability, integrity, and recoverability for our most regulated customers, Data Cloud will be able to utilize DRMC (Data Resiliency and Metadata Catalog) as a backup and recovery solution. Designed to protect mission-critical customer data against threats like cyberattacks and hardware failures, DRMC uses AWS-based S3 replication and immutable backups in secure “Bunker” accounts. Data is continuously replicated and locked to prevent tampering, with support for automated, near real-time recovery.

Share Data Cloud Data with other Data Cloud Orgs via Zero Copy

A new beta Zero Copy sharing capability enables secure, near real-time data sharing between different Data Cloud tenants or organizations without needing to replicate or re-ingest the data. The zero-copy architecture uses query federation to allow target orgs to access shared objects virtually from the source org, reducing storage costs, ELT duplication, and latency. This feature enhances operational efficiency, supports privacy-preserving collaboration in clean rooms, and unlocks advanced cross-org analytics and activation scenarios.

New Marketing Intelligence Connectors

Three new beta connectors will be available for Marketing Intelligence: Instagram, YouTube Insights, and Google Ads Manager. These connectors enable users to seamlessly ingest, transform, and harmonize key marketing performance data from popular platforms.

Image Processing: Unstructured Data

Image Processing will be available as an open beta in the feature manager, expanding unstructured data capabilities to include image understanding alongside text, audio, and video within the current unstructured pipeline. The new capability supports OCR (image-to-text conversion), image captioning using GPT4o to describe image content, and image similarity, which enables searching for visually similar items from a catalog. It can process images from sources like CRM or cloud storage, parse and index them, and make the extracted insights searchable for agents and applications. Supported formats include JPEG, PNG, JPG, and PDFs with embedded images. When the image processing option is checked when building your search index, the processed image results appear as new chunks within your chunk DMO.

Federated Authentication for Amazon Kinesis Connector

AWS credentials for Kinesis connections no longer need to be stored in Data Cloud. Instead with this beta feature, Salesforce can be registered as an identity provider in AWS IAM, allowing admins to create a scoped Identity & Access Management (IAM) role with only the necessary permissions for Data Cloud to access Kinesis. After the role is created, Data Cloud only needs the role’s ARN (Amazon Resource Name) to establish the connection. This helps centralize credential management within AWS, supports least-privilege security practices, and enables role edits directly in AWS without requiring updates in Data Cloud.

Connector Framework for Partners

Salesforce partners can white-label existing Data Cloud connectors and distribute them via AppExchange to increase the visibility and branding of their products. Partners can package a branded version of a generic connector (e.g., the Apache Iceberg connector) without building a new one from scratch. These connectors inherit updates to the base connector automatically, and can be provided at an equal or lower release status of the underlying connector (GA, Beta, etc). Future versions of the framework will expand to allow partners to build entirely new connectors.

Data Graph - Tagging Root Keys in Participating DMOs

Data Graphs with deep nested data can now be refreshed faster by allowing customers to directly tag a root key (like an Individual Id) within lower-level DMOs. When a DMO changes, the system must traverse multiple foreign key relationships to find which Data Graph record to update which increases time and cost for large, multi-level graphs. By adding the primary key of the root DMO as a field in the participating DMO and tagging it as the root key, the system can immediately identify the correct Data Graph record to update, eliminating the need for lengthy traversal. This optimization is especially valuable for deep data graphs with frequently changing DMOs, reducing refresh latency and costs.

Unstructured Data: New File Types and Section Aware Chunking

Data Cloud's vector database now supports Microsoft Office Files (docx, pptx, xlsx), emails (eml, msg with attachments), and XML files. The update also debuts Section-Aware Chunking, which packs related paragraphs or sections together and allows for overlapping text for better context. This addresses limitations of the passage extraction method that can produce smaller chunks. With this new chunking capability, Data Cloud enables more accurate vector embeddings, improving retrieval quality and the completeness of Agentforce-generated answers.

Unstructured Data: Process History for Search

Search indexes now feature a Process History tab which provide granular job metrics including timestamps, number of processed and skipped records, total data size, number of chunks created or removed, and job status. Metrics are available for all jobs completed since June 2025, with the most recent 1,000 runs retained. This feature gives customers transparency into the indexing process, making it easier to validate that new unstructured data has been successfully processed into the vector database while getting vital details that impact overall consumption.

Unified Household

Individuals can now be grouped through Identity Resolution into unified households using either party identification or contact point address details. This enables use cases where the point of contact or analysis originates from the household rather than the individual or an account. Like Unified Individuals or Accounts, Unified Households are connected through a bridge record, in this case the Household Member Link, which links the Individual to the Household. Individuals can belong to multiple households if they possess multiple addresses. For this initial release, only one match rule can be configured per rule set to avoid overly complex groupings. The ability to segments and activations on households and unified households is planned for the next monthly release.

Secondary Indexing for DLOs

This feature greatly improves query performance and reduces credit consumption for large Data Lake Objects (DLOs) when filtering on non-primary key fields (e.g., global account ID). It works by using Batch Data Transforms to create a duplicate indexed DLO partitioned on the chosen field, allowing the query engine to avoid full table scans and directly retrieve matching rows. The index is rebuilt daily and is ideal for highly selective queries (returning a small percentage of rows). Due to the increased processing costs, ROI is typically achieved in scenarios with over 10,000 queries per day per DLO. While only one index per DLO and single-field indexing are supported in this release, the feature fully respects governance and data space permissions, giving admins a scalable way to speed up and lower the cost of large-table queries.

Additional Segment Scheduling

Customers can gain more control with their segments by scheduling refreshes at a wider variety of supported frequencies without having to refresh segments manually or via flow. Segments can now be scheduled in daily, weekly, or monthly intervals, with precise controls that dictate the day of week, time of day, and range of the automated schedule.

Segmentation Data Preview

With the release of this feature, segment users can preview sample results of a segment before publishing it, helping avoid unnecessary processing costs and delays. Using a new component called Data Lens, the feature provides structured data profiling and visualization similar to Tableau Prep. Users can see a table of up to 1,000 sampled rows, while the profile pane above summarizes attribute distributions using up to 1 million rows. This enables users to validate filters, confirm data quality, and explore attributes interactively before finalizing their segments. The feature is enabled by default, requires no additional license, and is billed at 2 credits per 1 million records processed, making it easier to QA segments efficiently and confidently.

Related Attribute Multi-Path Support

Single-path limitations during attribute activation are a problem of the past, allowing marketers to pull attributes from multiple Data Model Object (DMO) paths in a single activation. This unlocks greater flexibility for building personalized campaigns that combine data from different areas of the model, such as purchase history and engagement data, without requiring data flattening or post-processing. The key enhancements include:

• Multipath Support: Marketers can now select attributes across multiple DMO paths, enabling more comprehensive and personalized activations.

• Decoupled Filtering: Filters can now be applied without including the filtered attributes in the activation payload, reducing unnecessary data transfers.

• Improved Grouping UI: The new grouping interface better reflects the final JSON structure of the activation payload, making it easier to visualize the output.

• DMO Counter: A counter shows how many DMOs are involved in the activation, including intermediate joins, with a guardrail limit of 8 to help users manage complexity.

Activation Triggered Flows

This pilot feature allows users to activate segments to external systems without relying on manual exports or complex middleware. It enables marketers to define an activation in Data Cloud and automatically send the resulting audience payload to any API-based endpoint, orchestrated through Salesforce Flow.

The feature supports two options for connecting to external systems:

1. External Services + Named Credentials – Users configure authentication (API key, OAuth 2.0, or basic auth) via named credentials, create an external service, and add an HTTP callout action in the new Activation Trigger Flow.

2. Mulesoft Connectors in Flow – Users leverage out-of-the-box Mulesoft connectors (e.g., HubSpot, Twilio, Zendesk) to map Data Cloud attributes directly to the target system’s schema.

Activated segments should use Data Cloud as the activation target. Once the activation runs and updates the Audience DMO, a platform event triggers the flow, which can read the payload and pass it to the endpoint via action. The pilot currently supports batch activations only, limited to Individual and Unified Individual DMOs, with a rate limit of 100k/hour and a maximum of two related attribute hops, with GA rates in the millions.

GetSegment Connect API

Segment IDs can now be used to query segment details in the Connect API. Previously, developers had to first look up the API name and then call the API again to retrieve the segment details. With this enhancement, a single API call using the segment ID returns the desired segment details, simplifying programmatic access and reducing steps for automation and integration use cases.

GetAll Segment Filter Enhancements

The Get All Segment Connect API now includes more advanced filtering options. Users can filter on three new fields: SegmentStatus, LastPublishedEndDateTime, and SegmentOn. Two new operators, not equal to (!=) and in, have been added, and the total number of filters allowed per request has increased to 10. LastPublishedEndDateTime field currently supports only the not equal to operator. Additionally, multiple filter conditions must be combined using AND in all caps. These updates allow for more precise, efficient segment retrieval without changes to the API endpoint structure.