Salesforce Data 360: From Data Streams to Actionable Insights

Organizations today collect customer data across a multitude of systems. CRM platforms, marketing tools, advertising platforms, data warehouses, and external applications can all contain valuable data about the same customer. But when these systems operate independently, organization don't have a unified view of their customers or a complete understanding of the people they serve.

Salesforce Data 360 helps organizations connect, harmonize, unify, analyze, and activate that data. Data 360 creates a unified layer across the existing systems, instead of replacing them, which allows businesses to turn isolated data fragments into one meaningful customer profile that can help them create targeted audiences and build analytics and AI-powered experiences.

Data 360 is a customer data platform designed to connect data from multiple systems and organize it into a unified and consistent customer view. Data 360 acts as a source of reference rather than a source of record. The original information can remain within its original system, while Data 360 brings that data together for analysis and purpose.

Connecting Data Across the Organization

The Data 360 process starts with importing data streams. A data stream defines how data enters the platform from its original source. These sources could be Salesforce CRM records, marketing applications, advertising platforms, commerce systems, APIs, uploaded files, or cloud data platforms.

Data streams may operate by a scheduled batch ingestion or near-real-time streaming. Zero-copy connections are also supported by certain cloud data platforms, which allows Data 360 to query data without requiring the organization to duplicate and move the underlying data.

Once a stream is configured, the incoming data ends in a Data Lake Object, also known as a DLO. The DLO represents the raw version of the source data and within a DLO, fields generally retain the names, formats, and structures used by the original system.

Harmonizing Data with a Shared Model

Different systems often describe the same data in different ways. One system may use "first name" while another may use "First Name". Even though these two fields represent the same data, they cannot be used consistently until they are mapped into a shared model.

Data Model Objects, also known as DMOs, build that shared vocabulary. Mapping DLO fields to DMOs transforms source-specific data into a standardized structure that can be used throughout Data 360. This means customer records from several systems may all be mapped to the Individual DMO, while email addresses may be mapped to the Contact Point Email DMO.

The difference between DLO and DMO is important. The DLO is the raw landing layer, and the DMO is the harmonized layer that is used downstream in Data 360. Downstream capabilities such as identity resolution, segments, and calculated insights generally rely on the mapped DMOs.

Some data may also require some reshaping before it can be mapped effectively, which can be accomplished through transforms. A streaming transform can handle simpler, near-real-time operations, while a batch transform supports more complex logic such as joins and aggregations. The right transform approach depends on the structure of the source data and how quickly the result must become available.

Creating Unified Customer Profiles

After the data has been harmonized and mapped to a DMO, identity resolution determines which records represent the same customer or person. A customer may be a lead in one system, a contact in another, and a website visitor in a third. Data 360 uses matching rules to link those records to a Unified Individual. These rules may consider email addresses, names, phone numbers, physical addresses, dates of birth, or industry-specific identifiers in this process. Organizations can also use fuzzy matching to account for variations such as "Matt" and "Matthew"

Reconciliation rules determine which value should represent a single-valued attribute, once records are matched, to select a source of truth. An organization may prioritize its CRM, the most recently updated value, or the value that appears most frequently across systems.

Identity resolution requires ongoing oversight. If a rule is too strict, it may fail to connect records belonging to the same individual, and if a rule is too broad, it may incorrectly merge different individuals. Organizations should therefore treat identity resolution as a continuing process of monitoring, testing, and refinement.

Turning Unified Data into Audiences and Insights

Once unified individuals are created, organizations can create segments based on customer attributes and behaviors. A segment might identify customers who abandoned a shopping cart during the last 2 days, exclude anyone who did complete a purchase, and send the resulting audience to a marketing journey. To support operational use cases, segments can also identify customers who have opted out of communications and send that information back to the original CRM or other source platform.

Organizations can also perform analytics in Data 360 through calculated insights, streaming insights, Salesforce reports, Tableau, and Tableau Next. Through these capabilities, organizations can analyze unified information that provides a more complete view of customer activity and business performance than what is possible within isolated systems.

Grounding Agentforce in Trusted Data

Through AI Models, Data 360 provides the ability to ground Agentforce agents in trusted data with retrieval-augmented-generation, also known as RAG, to find relevant content and provide it to the large language model when a user asks a question. Knowledge is divided into smaller passages known as chunks which are embedded as vectors and stored in a search index. When a user asks a question, a retriever searches the index and returns the most relevant content. A prompt template then gives those results to the large language model, so it is able to generate an answer that is based on trusted data.

Chunk design has a significant impact on retrieval quality. Each chunk should be focused enough to answer a specific type of question, while still containing enough context to stand on its own. Including identifying information, such as a program, product, customer, or policy name, can help prevent the agent from retrieving the wrong passage for the use case.

Key Takeaways

Salesforce Data 360 provides the foundation for turning disconnected data into coordinated business action. By connecting different data sources, harmonizing records into a shared model, building unified individuals, creating segments, generating insights, and grounding Agentforce agents, organizations can build a more consistent and useful understanding of their customers.

For successful implementation, organizations should begin with a clearly defined outcome rather than an attempt to connect to every available data source at once. With the right architecture and governance, Data 360 can help transform existing technology investments into more personalized experiences in a more efficient process that provides measurable business value.