Salesforce Analytics Architecture
Which Salesforce analytics tool answers which question, how data reaches each one, and how a semantic layer and governance keep the numbers consistent.
A Salesforce analytics architecture decides which tool answers which question, where the data for each tool comes from, and who owns each metric definition. Most organisations already have several tools: CRM reports and dashboards, perhaps CRM Analytics, Tableau, a marketing analytics product and a data warehouse. Without a plan, the same metric is calculated four ways. This page is a reference design; we validate the APIs, edition and permissions for each implementation.
Business context
The usual complaints are predictable: pipeline in the CRM dashboard does not match the finance report, marketing claims more conversions than sales recorded, and nobody knows which "active customer" count is right. The fix is less about choosing one tool and more about three decisions:
- Operational vs analytical reporting. Live CRM reports suit day-to-day work; cross-source analysis belongs on harmonised data.
- Where data is harmonised. In Data 360, in a warehouse, or both, with zero copy between them.
- Where definitions live. A semantic layer, so "active customer" is defined once.
GeekMidst can design and implement this architecture across Salesforce reporting, CRM Analytics and Tableau. See also Tableau and CRM Analytics consulting.
Salesforce analytics reference architecture
Reference architecture: a reusable pattern, not a live client integration
Select a component to see what it is responsible for and how data moves to and from it.
Components and data flows (text version)
| Component | Type | Responsibility |
|---|---|---|
| Sales and Service Cloudlive records | Salesforce product | Operational data: leads, opportunities, cases, activities. The source for pipeline and service metrics. |
| Marketing CloudEngagement or Next | Salesforce product | Sends, opens, clicks and journey data. In Engagement, system data views such as _Job, _Open and _Subscribers can be queried with SQL. |
| External platformsERP, ads, Braze, web analytics | External or custom system | Finance, media and engagement data. Braze can share data to Snowflake or stream engagement events through Currents. |
| CRM reports and dashboardsin the org | Salesforce product | Live operational reporting on CRM records, respecting each user's sharing and field access. |
| CRM Analyticsdeep CRM-embedded analytics | Data and analytics | Formerly Tableau CRM and Einstein Analytics. Still marketed; Revenue Intelligence is built on CRM Analytics Plus. |
| Marketing Cloud Intelligenceexisting customers, Intelligence+ | Data and analytics | Formerly Datorama. Offered to existing customers as Intelligence+ within Marketing Cloud Next. No end-of-sale notice found. |
| Data 360DMOs, calculated insights | Data and analytics | Harmonised, identity-resolved customer data and calculated insights such as lifetime value. |
| Data warehouseSnowflake, Databricks, BigQuery | Data and analytics | Enterprise history, finance and non-customer data. Works with Data 360 through zero copy. |
| Semantic layerTableau Semantics or warehouse | Data and analytics | Holds governed metric and dimension definitions once, so every dashboard uses the same calculation. |
| Tableau and Tableau NextCloud, Server, Next | Data and analytics | Cross-source analysis. Tableau Next is API-first agentic analytics on the Salesforce Platform and Data 360. |
| Agents and MCP clientsgoverned access | AI and agents | AI assistants that ask analytics questions through governed tools such as Tableau MCP, under each user's permissions. |
| Business userssales, service, marketing, finance | People | Read dashboards and ask questions. Certified sources are marked so users know which numbers to trust. |
Data flows
- Sales and Service Cloud → CRM reports and dashboards: live records. Reports run directly on CRM data with the running user's access.
- Sales and Service Cloud → CRM Analytics: CRM data sync. CRM Analytics loads CRM data on a schedule. Show the refresh time on dashboards and alert on failed syncs.
- Marketing Cloud → Marketing Cloud Intelligence: campaign data. Marketing Cloud Intelligence connectors bring campaign data in for existing customers.
- External platforms → Marketing Cloud Intelligence: media data. Ad platform and media data for cross-channel marketing analysis.
- Sales and Service Cloud → Data 360: data streams. Zero-cost connectors bring CRM objects into Data 360.
- Marketing Cloud → Data 360: engagement data. The Marketing Cloud Engagement connector brings engagement data into Data 360.
- External platforms → Data warehouse: ELT pipelines. External data loaded into the warehouse by your pipelines, with their own monitoring.
- Data warehouse → Data 360: zero copy. Query federation reads warehouse data live without storing it; data sharing can send insights back. File federation is labelled beta on Salesforce's page.
- Data 360 → Semantic layer: harmonised objects. Customer data with resolved identity feeds the semantic models.
- Data warehouse → Semantic layer: warehouse models. Finance and history tables feed the same semantic layer.
- Semantic layer → Tableau and Tableau Next: governed metrics. Dashboards use certified definitions instead of their own calculations.
- Tableau and Tableau Next → Agents and MCP clients: analytics tools. Tableau MCP and the Tableau Next hosted MCP server expose governed analytics to AI clients; check status and licences.
- Tableau and Tableau Next → Business users: dashboards. Published dashboards and Slack-integrated insights, filtered by user access.
- Agents and MCP clients → Business users: answers with sources. Answers cite the metric and source used, so users can check them.
Which tool for which question
| Tool | Best for | Data | Status and licence notes |
|---|---|---|---|
| CRM reports and dashboards | Operational lists and team dashboards | Live CRM records | Included; Winter '27 adds record previews and joined-report options in Beta |
| CRM Analytics | Deep analytics embedded in CRM | Synced CRM and external data | Separate licence; Revenue Intelligence uses CRM Analytics Plus |
| Tableau Cloud or Server | Cross-source visual analysis | Warehouse, Data 360, files | Tableau Desktop is now free, per tableau.com |
| Tableau Next | Agentic analytics on the Salesforce Platform and Data 360 | Data 360 | Sold standalone, in Tableau+, and included in certain editions (Core, Advanced, Max) |
| Marketing Cloud Intelligence | Cross-channel marketing performance | Marketing and media connectors | Existing customers, as Intelligence+ within Marketing Cloud Next |
| Data 360 calculated insights | Customer metrics such as lifetime value | Harmonised DMOs | Requires Data 360 |
Data flow, step by step
- Operational users work from live CRM reports, which always reflect current records and sharing.
- CRM Analytics syncs CRM data on a schedule for deeper analysis inside Salesforce.
- Data 360 ingests CRM and Marketing Cloud Engagement data, resolves identity and computes calculated insights.
- The warehouse holds finance and long-term history; Data 360 reads it through zero copy instead of duplicating it.
- The semantic layer defines metrics once over both, and Tableau dashboards and AI tools read those definitions.
- Users see a certified badge on governed content and a refresh time on every dashboard.
Trust boundaries and authentication
Each tool enforces access differently. CRM reports use org sharing and field-level security. CRM Analytics and Tableau need their own row-level security, which must be designed to mirror the org's rules rather than assumed. AI access should run per user: Salesforce Hosted MCP Servers, GA since 29 April 2026 for Enterprise Edition and above and intended for customers with Flex Credits, run each call as the authenticated user, and hosted Tableau MCP uses per-user OAuth 2.1. Tableau says hosted Tableau MCP is generally available, but another section of the same page says "coming soon", so treat it as likely GA, not confirmed. The analytics/tableau-next hosted server's status is not stated. Never give an AI tool direct warehouse credentials or unrestricted SQL; expose governed, read-only tools first.
Failure scenarios and how to handle them
| Scenario | Effect | Handling |
|---|---|---|
| Two dashboards disagree | Loss of trust | One semantic definition per metric, with a named owner |
| Dataset sync fails | Stale numbers shown as current | Show refresh time and alert on failed syncs |
| Row-level security gap | Users see data they should not | Mirror org rules and test with real user profiles |
| Marketing SQL query fails | Missing engagement data | SQL Query activities deactivate after 24 consecutive failures and time out after 30 minutes; monitor and optimise |
| AI answer without source | Unverifiable numbers | Require answers to cite the metric and source |
Operational monitoring
Track dataset and extract refresh success, data stream refresh in Data 360, dashboard usage (so unused content can be retired), query performance and row-level security test results after each change. Review certified content quarterly with metric owners.
Alternative approaches and trade-offs
- CRM reports only. Cheapest and live, but limited for cross-source analysis and history.
- Warehouse-first. Strong where a data team already runs a warehouse; Data 360 then focuses on activation and identity. See Data 360 vs a traditional data warehouse.
- Tableau Next on Data 360. Keeps analytics close to customer data and agents; check edition entitlements first.
Implementation checklist
- 01
List key questions
Write down the decisions each team makes and the metrics behind them.
- 02
Assign metric owners
One owner and one written definition per metric.
- 03
Choose tools per question
Operational, embedded, cross-source or marketing analytics.
- 04
Design the data path
Decide where harmonisation happens and use zero copy where it fits.
- 05
Mirror security
Design row-level security to match org access and test it.
- 06
Certify content
Mark governed data sources and dashboards.
- 07
Plan AI access
Start with read-only, per-user tools and log usage.
Related services and solutions
Tableau and CRM Analytics consulting
Dashboards, datasets and governance
→Marketing analytics and reporting
Campaign and engagement reporting
→Marketing Cloud Intelligence consulting
Support for existing customers
→Tableau MCP integration
Governed AI access to analytics
→Customer 360 architecture
Systems of record behind the metrics
→Data 360 architecture
Calculated insights and harmonised data
→Questions we hear
Is CRM Analytics being replaced by Tableau Next?
Is Marketing Cloud Intelligence still sold?
Do we need a warehouse if we have Data 360?
Can AI assistants query our dashboards safely?
Sources
- https://www.salesforce.com/analytics/tableau-next/
- https://www.salesforce.com/kr/products/crm-analytics/overview/
- https://salesforce.com/blog/marketing-cloud-next-details/
- https://www.salesforce.com/data/connectivity/zero-copy/
- https://tableau.github.io/tableau-mcp/docs/hosted-tableau-mcp
- https://developer.salesforce.com/blogs/2026/04/salesforce-hosted-mcp-servers-are-now-generally-available
Get one version of the numbers
We can map your reporting needs to the right tools, design the semantic layer and set up governance. Consultations are remote, from India.
