Data 360 (the platform formerly called Data Cloud) is often sold as "a CDP" and then implemented like a data warehouse. It's neither, quite. It's a data platform built into Salesforce that ingests, harmonises and unifies data so that it can be acted on — by segments, flows, analytics and, increasingly, AI agents.

Here's how I explain the layers to architects new to it.

The five layers

flowchart LR
  subgraph CON[1 · Connect]
    DS[Data streams · connectors]
    ZC[Zero-copy federation]
  end
  subgraph HAR[2 · Harmonise]
    DLO[(Data Lake Objects)]
    DMO[(Data Model Objects)]
  end
  subgraph UNI[3 · Unify]
    IR[Identity resolution]
    UP[Unified profiles]
  end
  subgraph ANA[4 · Analyse]
    CI[Calculated insights]
    VS[Vector / search index]
  end
  subgraph ACT[5 · Act]
    SEG[Segments & activation]
    DA[Data actions · Flows]
    AG[Agent grounding]
  end
  DS --> DLO --> DMO --> IR --> UP
  ZC --> DMO
  UP --> CI & SEG
  DMO --> VS --> AG
  CI --> SEG & DA & AG
From raw data to action — the Data 360 layers

1 · Connect

Data streams bring data in from Salesforce orgs, cloud storage, web/mobile SDKs, MuleSoft and other connectors. Zero-copy federation lets you query data that stays in your lakehouse or warehouse (Snowflake, Databricks, BigQuery, Redshift and others) without moving it.

2 · Harmonise

Ingested data lands in Data Lake Objects (DLOs) in its source shape. You then map it to Data Model Objects (DMOs) — a canonical model (Individual, Contact Point Email, Sales Order and so on). This mapping is the most important design work you'll do.

3 · Unify

Identity resolution uses match rules (exact, fuzzy, normalised) and reconciliation rules (which source wins for each attribute) to produce unified profiles. Get this wrong and every downstream insight is wrong too.

4 · Analyse

Calculated insights compute metrics over unified data — lifetime value, engagement scores, propensity signals. Unstructured content (documents, transcripts, knowledge) can be chunked and indexed for vector and hybrid search, which is what makes retrieval-augmented AI possible.

5 · Act

Segments activate to marketing and advertising destinations; data actions trigger Flows or external events; and agents use the data for grounding. This is where the value shows up.

Where teams usually get the model wrong

Do
  • Start from the use case and work back to the data you need
  • Design DMO mapping with the business glossary in hand
  • Test identity resolution on real, messy data before go-live
  • Plan consumption and credit usage as an NFR
Don't
  • Ingest "everything" on day one and figure out use later
  • Create custom DMOs when a standard one fits
  • Use overly loose fuzzy matching — false merges are worse than duplicates
  • Copy data that zero-copy could have left in place

How it connects to Agentforce

An agent is only as good as what it knows. Data 360 gives agents a unified, governed view of the customer and searchable access to unstructured knowledge. In practice, the quality of your harmonisation and identity resolution shows up directly in how accurate and personal your agents feel.

Data 360 is not a place to store data. It's a place to make data useful — so design it around the decisions it needs to power.
Data 360Data ArchitectureZero Copy
Pranav Sanvatsarkar

Pranav Sanvatsarkar is a Solution Architect with 14+ years across telecom, healthcare, public sector and non-profit, focused on AI, agentic automation and Salesforce architecture. Say Hello!