Most mid-size companies I talk to already have Data 360 provisioned. It came with a renewal or an Agentforce deal, someone clicked through setup, and now there's an empty Data Cloud app and a vague goal of "one view of the customer." The question is where to start without burning months and consumption on things nobody uses.
Short answer: pick one decision the profile has to support, connect only the two or three sources that decision needs, map them cleanly to the standard data model, get identity resolution right on real records, and ship one activation back into Sales or Service Cloud. Segments, extra sources and agent grounding come after that first activation is in use.
Start With One Decision, Not the Diagram
Every Data 360 kickoff has a slide with ten sources feeding one profile. That slide is how projects stall. Before connecting anything, write down the one thing someone should be able to do once the profile exists:
- A service agent sees every Contact, Lead and order history for the caller on one page
- A rep sees product usage and open support issues on the Account before a renewal call
- Marketing builds one audience from CRM and web data instead of rebuilding it in spreadsheets
That decision tells you which sources matter, which match rules you need, and what the activation looks like. Anything that doesn't serve it waits for phase two.
1. Data Streams: Two or Three Sources, Not Ten
A data stream is a source connected to Data 360 and landing in a data lake object. For most mid-size orgs the first set is short:
- Salesforce CRM through the built-in connector. Salesforce ships starter data bundles for Sales and Service Cloud objects, which save a lot of manual mapping
- One system outside Salesforce that holds what CRM doesn't: an ERP, a product database, a warehouse table or an ecommerce platform
- Maybe a third if the first decision really needs it, like web engagement or a marketing tool
Two habits keep this phase cheap. First, only ingest the CRM objects identity resolution needs. On the Contact 360 project, Data 360 ingested Account, Contact, Lead and the Contact Point objects, and the page read opportunities and activities live from Salesforce. Second, if the outside data already lives in Snowflake, Databricks, BigQuery or Redshift, check whether zero copy can read it in place before you copy it. I cover that tradeoff in Data 360 vs a data warehouse sync.
Set refresh schedules to what the decision needs. A renewal view rarely needs anything faster than daily.
2. Data Model Mapping: Stay Standard
Data lake objects get mapped to data model objects: Individual, Contact Point Email, Contact Point Phone, Party Identification, Account and the rest. This is where most of the long-term cost hides, because every shortcut here shows up later in identity resolution and in every query.
What I check before calling mapping done:
- Standard objects first. Use Individual and the Contact Point objects as designed. Custom data model objects are fine for data that has no standard home, not for avoiding the learning curve
- Your master ID goes into Party Identification. If an ERP or MDM system owns customer IDs, they need to land as Party Identification rows, not as a custom field nobody can match on
- Normalize before you match. Emails lowercased and trimmed, phones in one format, placeholder values filtered out
- Know what's blank. Count how many records have an email, a phone and a master ID before the first identity run, so you have a baseline
3. Identity Resolution: Check It Against Real Records
Identity resolution is where the profile becomes useful or becomes a liability. A ruleset with match rules decides which records are the same person, and reconciliation rules decide which values win.
For a first ruleset I keep it small:
- Exact match on the strongest ID you have, usually a master or ERP customer ID, with match on blank turned off
- Normalized email, tightened with last name, because shared mailboxes merge people on email alone
- Fuzzy name matching only after the exact rules are proven on real data
Then pull a sample of unified profiles and look at them with someone who knows the customers. Merges that look fine on a dashboard often fall apart when a sales rep reads the names. If Data 360 is already shared with marketing, remember that changing a shared ruleset changes their profiles too, and get their sign-off first. The full story of how that played out is in my identity resolution with MDM and Databricks post.
Ask whether anything will read the reconciled values at all. If the page only needs to know which records belong together and can show live Salesforce values, you can skip most reconciliation rules.
4. One Activation People Use
The first activation is what turns Data 360 from a platform into something people notice. For a mid-size org that's usually inside Salesforce, not an ad platform:
- Related lists and copied fields on the Contact or Account page, through Data 360 enrichments
- A Lightning page component that shows every linked record for the person
- One calculated insight, like lifetime value or open issues across systems, next to the record
Ship one, train the people who use it, and watch whether they do. If reps ignore it, you've learned something cheap. If they ask for more, phase two has a clear priority list.
What Can Wait
- Segments and marketing activation, unless marketing is the first decision
- Agentforce grounding. The data model you build now is the one agents will read later, so getting it right is the prep work
- More sources. Add them one at a time, each tied to a question someone is asking
- Unstructured content like Knowledge or Confluence pages, which I wrote about in Confluence to Salesforce Knowledge through Data 360
Keep Consumption Predictable From Day One
Data 360 bills on consumption, and early habits set the curve. What drives usage most in a first build is how much you ingest, how often streams refresh, how often identity resolution and insights run, and how many queries hit the data. Fewer sources, daily refreshes where daily is enough, and reading CRM data live instead of copying it back keep the first months predictable. I walk clients through those drivers before go-live, not after the first bill.
How Long the First Phase Takes
The scope above maps to my Data 360 Launch package: up to three sources ingested and mapped, identity resolution for one unified profile, up to five calculated insights or segments, and one activation back into Sales or Service Cloud, with two weeks of hypercare. It typically takes 5 to 7 weeks. What moves the timeline is the number of sources outside Salesforce, how clean the IDs are, and whether a shared ruleset needs another team's sign-off.
If you've got Data 360 turned on and aren't sure what to build first, book a call and I'll tell you where I'd start. More on how I run these projects on the Data 360 consulting page.
