Salesforce
Data 360 (Data Cloud) Implementation

Data 360, formerly Data Cloud, brings customer data from Salesforce clouds, websites, warehouses and other systems into one place, unifies it into profiles, and lets teams and AI agents act on it. Salesforce calls it "the real-time data engine that powers the entire Salesforce platform" and "the essential data foundation for Agentforce".

On its Data 360 page Salesforce groups the product in three areas. The table uses Salesforce's own wording.

Salesforce also says Data 360's primary focus is activating data, which is why an implementation is judged by what the profiles drive, not by how much data is loaded.

A Data 360 implementation runs in six build steps after scoping: connect sources through data streams, map them to the data model, set identity resolution rules, define calculated insights, build segments, and activate them to target systems. Each step produces something you can check before the next one starts, which keeps errors from spreading downstream.

Scoping comes first: the use cases, the source list, the consent rules and an estimate of consumption. When sources sit in systems without a ready connector, our MuleSoft team and Salesforce integration services build the feeds.

Identity resolution uses a ruleset with two kinds of rule. Match rules decide which source records belong to the same person; reconciliation rules decide which value wins when those records disagree. The result is a unified profile that updates as source data changes, while Data 360 still stores every unique source value.

Trailhead lists three match methods. Exact needs the data to match precisely. Fuzzy allows typos and slightly different spellings. Normalized matches the same information regardless of formatting and is available for email, phone and address. You can have several match rules, each another chance for records to match, and a rule is met only when all its criteria are.

Reconciliation rules can pick the last updated value, the most frequent value across linked records, or a value by source sequence, so a trusted system outranks a less validated one. They apply at object and field level. Trailhead notes that a unified profile is made of the unified individual and the unified link individual objects, which join it back to the source records.

For a full walk-through with worked examples, read Data 360 identity resolution explained.

Zero copy lets Data 360 query data where it already lives instead of copying it in. Salesforce lists out-of-the-box support for Snowflake, Databricks, Redshift and BigQuery, says zero copy "removes the need for traditional ETL data replication", and calls it bi-directional, so Data 360 insights can also be read by platforms such as Snowflake.

Salesforce describes two federation types. Query federation sends queries to the external source over JDBC, and no data is stored in Data 360. File federation, which the zero-copy page marks as beta, reads the external storage directly without using the warehouse's compute. Salesforce's architecture guide compares the options:

In practice the choice is made per data set during design, because it sets freshness, cost and where governance is applied.

Data 360 gives Agentforce agents two kinds of context: structured data from unified profiles and data model objects, and unstructured content such as cases, knowledge, files and web pages through retrieval-augmented generation (RAG). A search index stores chunked, vectorized content; a retriever links it to the prompt template an agent action uses.

• Search index. Data 360 chunks the content and creates vector embeddings. Trailhead describes vector search for general long-form questions and hybrid search, which adds keyword matching, for the most accurate results.
• Retriever. Data 360 creates a retriever for each index as the bridge to prompt templates; the retriever fills the prompt with the most relevant information.
• Agent action. Trailhead's example builds an action called Answer Questions with Case that uses the prompt template.
• Unified profile. Trailhead notes RAG built on Data 360 has context from unified profiles, so answers can reflect who the customer is.
• Agentforce Data Library. For unstructured sources, a Data Library automates the data stream, mapping, search index and retriever.

Agentforce calls also pass through the Einstein Trust Layer. See Agentforce implementation and our guide to Data 360 for Agentforce grounding.

Data 360 governs access with data spaces, permission sets and policies, and records consent as data model objects tied to each contact point. Data spaces are logical partitions: users see only the spaces they can access. Policies combine tags, classifications and user attributes at field, object or record level, with attribute-based access control.

Trailhead is explicit that data spaces control what a user can access, not where data is located, so they are not a data residency tool. Data can be tagged by hand or with AI, and Platform Encryption for Data 360 supports customer- or externally managed keys.

For consent, Trailhead's Marketing Cloud Next module says consent is tied to the contact point, such as an email address, not just the person. The Communication Subscription Consent object records the status, date and source of consent, consent can be bulk imported from CSV or created by flows, and sends check consent at send time. A new email address does not inherit old consent, which is a design point for identity resolution.

Effort depends less on data volume than on the number and condition of sources, how much custom mapping they need, how hard identities are to match, how many insights, segments and activations the first release includes, and whether consent and governance rules already exist. A scoping call turns these factors into a plan and a usage estimate.

Usage is a separate cost line. Salesforce's pricing page lists Flex Credits at $500 per 100k and profile-based options at $240 and $420 per 1k profiles a year, says data ingestion is free, and notes the page is for information only and subject to change. Design choices such as refresh frequency affect consumption, so we estimate it during scoping. To start, book a discovery call.

Choose a Data 360 partner on evidence: certified Data 360 consultants you can verify on Trailhead, a plan that starts from your use cases, clear answers on identity rules, zero copy and consent, a usage estimate before the build, and integration skills for the sources that lack connectors. Ask for each of these before you sign.

1. Credentials you can check. Salesforce offers a Certified Data 360 Consultant credential; ask for names and verify them in Trailhead's credential lookup.
2. Use cases first. The plan should name the segments, service views or agent context the data must serve.
3. Identity and consent design. Ask how match and reconciliation rules will be tested against your real records, and how consent per contact point is handled.
4. Ingest or federate. A good partner explains, per data set, why it should be ingested or queried through zero copy.
5. Consumption estimate. Ask for an estimate of Data 360 usage tied to the design, not only a services quote.
6. Integration depth. Check that the team can build MuleSoft or API feeds where no connector exists.
7. Partner standing. Techila's AppExchange listing shows the Summit tier and a 4.82 rating from 19 reviews.

If you want a fixed first step rather than a full program, the Data 360 Quickstart covers one use case end to end.