On this page (16)
Salesforce Data 360 (formerly Data Cloud) and Snowflake can share data in both directions without copying it. Zero-copy data federation lets Data 360 use tables that stay in Snowflake, so warehouse data such as orders or product usage can enrich customer profiles, segments and AI agents. Zero-copy data sharing goes the other way, exposing Data 360 data to Snowflake for analytics and data science. Salesforce and Snowflake made the bidirectional integration generally available, and it gives teams a way to connect CRM and warehouse data without building and maintaining pipelines.
Which direction do you need?
| Direction | Called | Typical use |
|---|---|---|
| Snowflake to Salesforce | Zero-copy data federation | Bring order history, product usage, billing or loyalty data into Data 360 for unified profiles, segments, calculated insights and agents |
| Salesforce to Snowflake | Zero-copy data sharing | Analyze CRM, engagement and unified profile data alongside warehouse data in Snowflake, for BI and data science |
Salesforce's zero-copy federation also works with other platforms, including Google BigQuery, Databricks and Amazon Redshift, so the same pattern applies if your warehouse is not Snowflake.
What are good first use cases?
- Service agents and AI agents seeing recent orders, shipments or usage from the warehouse on the customer record, without nightly copies.
- Marketing segments built on purchase or product-usage data that only exists in Snowflake.
- Account health or churn signals calculated from usage data and surfaced to customer success in Salesforce.
- Revenue and pipeline analysis in Snowflake that joins CRM opportunities with finance data.
What should you plan for?
- Identity: decide how warehouse records match Salesforce accounts and contacts, such as by customer ID or email, before building segments.
- Data model: map Snowflake tables to Data 360 data model objects deliberately, starting with the few that drive your use case.
- Performance: federated queries run against Snowflake; for frequently used data, Salesforce offers acceleration, which caches a copy in Data 360, so test response times for agent and console use.
- Cost: both platforms meter usage. Data 360 consumes credits and Snowflake bills compute for queries, so estimate both for your query patterns.
- Governance: apply access controls in both systems and agree which team owns each shared dataset.
How do you start?
- Pick one use case with a clear owner and metric, such as reducing service handle time with order data.
- List the Snowflake tables it needs and the Salesforce objects they relate to.
- Set up the connection, federate only those tables and build identity rules.
- Test with real users and real volumes, then expand table by table.
What decisions come before the first project?
- Which system is the source of truth for each data set you plan to share?
- Who owns access approvals in each platform, and how will they stay in step?
- Which business question will the first use case answer, and how will you know it worked?
- How fresh does the data need to be: real time, hourly or daily?
- Who watches consumption and cost on each side once the data is live?
Who does what on the project?
| Role | Responsibilities |
|---|---|
| Business owner | Defines the use case, the success measure and who uses the result |
| Data or analytics engineer | Prepares and documents the Snowflake tables, manages warehouse access and cost |
| Salesforce architect | Maps data to Data 360 objects, designs identity resolution and how results reach users |
| Security and governance | Approves access in both systems and reviews what each team can see |
| Salesforce admin | Builds the screens, segments or agent actions that use the shared data |
Signs zero-copy is not the right first step
If CRM data itself is unreliable, with duplicates, missing owners and inconsistent fields, sharing it with a warehouse spreads the problem further. Fix the core records first. Likewise, if no one can name a decision that would change with the shared data, start with that conversation rather than the integration.
How do you keep cost and performance under control?
- Federate only the tables a use case needs, and review usage before adding more.
- Test response times where people will feel them, such as service consoles and agent answers.
- Use acceleration for data that is read constantly, and leave rarely used data federated.
- Set budgets and alerts for consumption in both platforms, with a named owner for each.
- Review query patterns monthly and retire shared data nobody uses.
How do you measure the first use case?
Agree the measure before building. For a service use case, measure handle time or repeat contacts before and after agents see warehouse data. For marketing, measure segment performance against a version built without the shared data. For customer success, measure whether health signals predicted renewals better than before.
Frequently asked questions
Do we need Data 360 to use Snowflake data in Salesforce?
Not always. Smaller needs can be met with integrations that bring summarized data into Salesforce fields. Zero-copy through Data 360 suits teams that need richer profiles, segments or AI agents grounded in large warehouse data sets.
Does zero-copy remove the need for data engineering?
No. It removes much of the pipeline work, but tables still need to be modeled, documented, secured and kept accurate. Plan for data engineering time on the warehouse side.
Can the same approach work with other warehouses?
Yes. The same pattern applies to other supported platforms, so the planning steps here carry over if your warehouse is not Snowflake.
How long does a first use case take?
A narrow use case with clean, well-understood tables can show results in weeks. Identity resolution across messy sources is usually what extends the timeline.
A first-project timeline
| Period | Focus | Output |
|---|---|---|
| Weeks 1–2 | Agree the use case, measure, tables and owners | A one-page brief everyone has signed off |
| Weeks 3–5 | Connect, federate the chosen tables, map to Data 360 objects, build identity rules | Unified records for a test group of customers |
| Weeks 6–8 | Surface the data where users work and test with real users and volumes | Response times and accuracy confirmed in the console or agent |
| Weeks 9–12 | Measure against the baseline and decide what to add next | A go or no-go on expanding, with cost data from both platforms |
Red flags
- A plan to federate every warehouse table before any use case is defined.
- Nobody on the warehouse side assigned to the project.
- Identity rules agreed without testing on real, messy records.
- No budget owner for consumption in either platform.
Zero-copy removes a lot of pipeline work, but not the need for clear ownership, careful modeling and a measurable first goal. Projects that keep those three in view tend to expand smoothly.

