The customer data platform conversation in 2026 is no longer “do we need one.”
It is “do we already have one and not know it.”
Salesforce Data Cloud (now Data 360) and Microsoft Fabric with Customer Insights are the two heavyweight unification platforms enterprises are forced to compare, often because they already pay for both and need to pick a primary.
These are not the same product even when the marketing slides rhyme.
Who this is for
Enterprises with customer data spread across CRM, web, product, support, marketing, and warehouse — and a real activation goal (personalization, agent grounding, audience export) rather than a vanity unification project.
TL;DR pick
- Salesforce is your activation surface (Sales, Service, Marketing, Agentforce) — Data 360.
- Your analytics stack is already on Microsoft Fabric / OneLake — Customer Insights on Fabric.
- You need both worlds and refuse to pick — pick anyway. Run one as primary, the other as a federated source. Dual-primary CDPs do not work.
Comparison at a glance
| Dimension | Salesforce Data 360 | Fabric Customer Insights |
|---|---|---|
| Storage model | Zero-copy via Bring Your Own Lake (BYOL), native object store | Fabric OneLake (delta/parquet), single tenant lake |
| Identity resolution | Rulesets + ML matching, deterministic + probabilistic | Match rules + Synapse ML, configurable graphs |
| Activation | Native to Salesforce clouds, plus segments to Meta/Google/etc. | Native to Dynamics 365, Power Platform, plus exports |
| Compute model | Calculated insights, streaming + batch, CDP-native query | Spark/SQL on Fabric, Dataflows, KQL for telemetry |
| AI grounding | First-class grounding for Agentforce / Atlas | First-class grounding for Copilot Studio |
| Streaming | Native streaming ingestion + change data capture | Real-time intelligence (KQL), Eventstream |
| Governance | Field-level security inherited from Salesforce model | Purview lineage, sensitivity labels, Entra ID |
| Cost model | Credit-based (CDP credits), storage + compute + activation | Fabric capacity units (F-SKUs), Customer Insights add-on |
| Org-fit | Marketing- and service-led customer activation | Data-engineering-led, lakehouse-anchored |
Storage model: zero-copy vs lakehouse-native
Data 360 leans into zero-copy.
Bring Your Own Lake support means Snowflake, BigQuery, Databricks, and Iceberg-backed lakes can register tables as data lake objects without physical movement.
Data 360 then runs queries against those tables in place.
The pitch is “no second source of truth”; the catch is that performance and cost depend on the upstream lake’s behavior.
Fabric Customer Insights is the opposite end.
OneLake is the source of truth, in delta format, governed by Microsoft’s stack.
If your data is already arriving in Fabric via Mirroring or Shortcuts, Customer Insights sits naturally on top.
If your data lives in a non-Microsoft warehouse, you are mirroring or shortcutting it in, and the unification story works but the lake is Microsoft’s.
Neither model is wrong. They suit different data org maturities. Lakehouse-first enterprises pick Fabric; activation-first enterprises pick Data 360.
Identity resolution: where the marketers care
Data 360’s identity resolution is rule-based with optional ML augmentation.
You configure rulesets (deterministic matches first, then fuzzy), promote candidate matches, and review with a clear UI.
The model is “explicit and auditable” — a marketer can defend why two records merged.
Customer Insights ships configurable match rules and supports Synapse ML for probabilistic matching when the data has the volume to justify it.
The defaults work for simple cases; deeper matching configurations require a data engineer.
For most regulated industries (financial services, healthcare, public sector) the auditability of Data 360’s ruleset model is a procurement advantage.
For data-engineering-led teams who already have ML matching pipelines, Customer Insights’ openness to custom logic is the advantage.
Activation: where the value actually lands
A CDP that cannot activate is a warehouse with a marketing problem. Both platforms ship activation; their surfaces differ.
Data 360 activates into Marketing Cloud, Service Cloud, Sales Cloud, and Agentforce natively.
External activations target Meta, Google Ads, LinkedIn, TikTok, and others via packaged connectors. Calculated insights flow directly to agents for grounding.
Customer Insights activates into Dynamics 365 Customer Insights — Journeys, Sales, Service, and Power Platform apps.
External activations cover the major ad platforms and Microsoft Advertising. Copilot Studio uses Customer Insights as a grounding source.
Activation latency on both platforms is now near-real-time for streaming use cases (sub-minute) and minutes-to-hours for batch segments.
Where they diverge is the friction to reach a third destination — both rely on packaged connectors plus partner ETL for the long tail.
A small config lens
Conceptually similar shapes for ingesting a transactional event:
-- Data 360 — streaming insert via Data Stream
INSERT INTO order_events__dlm (
order_id, customer_email, amount, created_at
) VALUES (
'O-1023', '[email protected]', 129.00, CURRENT_TIMESTAMP
);
-- Fabric Customer Insights — write to OneLake via T-SQL or Spark
INSERT INTO order_events
SELECT 'O-1023', '[email protected]', 129.00, GETUTCDATE();
-- Then mapped into a Customer Insights data source.The SQL is familiar on both sides.
The platform-shaped work — DLM modeling on Data 360, semantic modeling on Customer Insights — is where the platforms feel different.
AI grounding: why this matters in 2026
This is the new procurement question.
Buyers are not asking “can it unify data”; they are asking “can my agent ground on it without leaking.”
Data 360 is purpose-built to ground Agentforce.
Calculated insights, semantic objects, and per-record access scopes feed directly into the Atlas reasoning engine. The grounding latency is small enough for in-conversation use.
Customer Insights grounds Copilot Studio and Copilot for Dynamics 365 the same way.
The data is in OneLake, Copilot retrieves through the Fabric data agent, and policy is enforced by Purview labels and Entra ID.
If your agent strategy is Agentforce, Data 360 is the obvious answer. If it is Copilot Studio, Customer Insights on Fabric is. Mixing across vendors works but you pay an integration tax.
Governance and privacy
Data 360 inherits Salesforce’s sharing model — field-level security, sharing rules, and record visibility carry across to the unified profile.
Strong if your data already lives in Salesforce; more work to mirror if it does not.
Customer Insights inherits Microsoft’s stack — Purview for lineage and DLP, Entra ID for identity, sensitivity labels for data classification.
The governance surface is broader because Fabric covers more than just customer data.
Both support consent management, opt-out propagation, and right-to-be-forgotten workflows.
Neither is a substitute for a real consent platform if you operate in heavily regulated geographies.
Streaming and event handling
Data 360’s streaming ingestion supports near-real-time data streams with change data capture from Salesforce sources and external connectors. Most marketers will find this fast enough; product engineers will sometimes want lower-latency primitives.
Fabric’s Real-time Intelligence (KQL on Eventstream) is the more mature event surface. If your use cases include product telemetry, fraud signals, or IoT data merging into the customer profile, Fabric’s event stack is the cleaner story.
For pure marketing personalization, both are sufficient. For event-driven product-led growth use cases, Fabric edges ahead.
Cost model: credits vs capacity units
Data 360 prices on CDP credits — a basket measure across storage, compute, profile counts, and activation volume.
Pricing is workload-shaped and rewards efficient modeling. Cost models predict well once you have 90 days of production data.
Fabric Customer Insights prices on Fabric capacity units (F-SKUs) plus Customer Insights licensing.
Capacity is shared with the rest of Fabric — a heavy analytics workload can crowd out CDP processing or vice versa unless you isolate capacities.
Neither is “cheaper.” Both reward thoughtful modeling and punish kitchen-sink ingestion.
Admin UX
Data 360 has converged on a workspace UX that looks and feels like Salesforce Setup with a CDP brain.
Pleasant for Salesforce admins, more foreign for data engineers.
Customer Insights inside Fabric is more notebook-driven and data-engineer-shaped.
If your team already operates in Power BI, Synapse, and Fabric, the experience is continuous.
Who should pick which
- Heavily Salesforce-anchored, marketing-led activation goals — Data 360.
- Microsoft 365 / Fabric anchored, data-engineering-led — Customer Insights on Fabric.
- Multi-cloud, no dominant CRM — pick by activation surface. If you cannot name where the segment will be used in the next 90 days, you are not ready to buy a CDP.
- You already pay for both — pick a primary based on which activation surface drives revenue, federate the other.
- Regulated industries — both work, but Data 360’s record-level visibility model is the more familiar story for financial services and healthcare CISOs.
- Composable / reverse-ETL believers — neither replaces a Hightouch / Census conversation if you want to keep your warehouse as the source of truth without a packaged CDP brain.
- Real-time product personalization — Fabric’s event stack is slightly ahead; Data 360 closes most gaps for typical marketing use cases.
The pick
Default to the platform that owns your activation.
The unification work is real but it is one-time; activation is daily.
For Salesforce shops the answer is Data 360. For Microsoft shops it is Customer Insights on Fabric. For everyone else, run a 90-day pilot with both — but with a defined activation outcome, not a generic “single customer view” goal.
For a related lens on data platforms, see our Salesforce Data Cloud vs HubSpot and Dynamics 365 vs Salesforce breakdowns.
