Individual and Contact Point Email belong to the Profile category because they describe a person and how to reach them. Website browse clicks and order events belong to the Engagement category because they record actions at a specific timestamp. A product catalog or store location directory belongs to Other. If you mistakenly assign clickstream events to the Profile category, Identity Resolution will attempt to unify individual clicks as if they were separate human beings.
The Ask: A marketing team asks for a straightforward audience segment: "Customers who browsed products on our website in the last seven days but did not complete a purchase."
The Symptom: When the implementation team opens the Segmentation canvas, the website browse object does not support relative time-window filters like "in the last 7 days." They attempt brittle workarounds, including custom Calculated Insights and nightly CSV exports.
The Root Cause: During initial ingestion, the website browse Data Stream was created under the Other category instead of Engagement. Only Engagement objects enforce an Event Time field and expose native time-series filtering in Segmentation. Because a Data Stream's category cannot simply be edited after creation, every workaround was trying to compensate for a single dropdown choice made months earlier.
The Fix: The team creates a new website browse Data Stream configured with the Engagement category, designates the click timestamp as the
Event Time field, maps the Data Lake Object to an Engagement DMO, and rebuilds the segment. With the proper category in place, building the seven-day behavioral filter takes only minutes.
🔑 Key Points
- The P-E-O Classification: Every Data Stream in Salesforce Data Cloud must be classified as Profile (who), Engagement (what and when), or Other (reference data) at ingestion time.
- Locked Once Created: A Data Stream and its underlying Data Lake Object (DLO) lock in their category upon creation. Changing a category later requires creating a new Data Stream and rebuilding downstream mappings, Calculated Insights, and Segments.
- Identity Resolution Requires Profile: Identity Resolution match and reconciliation rules operate exclusively on Profile DMOs (such as
Individual,Contact Point Email, andParty Identification). - Behavioral Segmentation Requires Engagement: Only Engagement objects require an
Event Timefield, which unlocks recency and frequency filters (such as "occurred within the last 30 days") in the Segmentation builder.
🧩 The Core Concept: Understanding the P-E-O Model
When you ingest raw data into Salesforce Data Cloud—whether from Sales Cloud, Service Cloud, Marketing Cloud Engagement, Amazon S3, or the Ingestion API—you first land that data in a Data Lake Object (DLO) and then map its fields to a canonical Data Model Object (DMO) in the Customer 360 Data Model. Choosing the Data Category is one of the very first configuration steps when defining that stream, and it acts as an architectural contract for the rest of the platform.
Each of the three categories enforces specific schema expectations and unlocks different downstream capabilities:
- Profile (The "Who"): Represents a person, a business entity, or the identifiers and contact points attached to them. Common standard DMOs in this category include
Individual,Account,Contact Point Email,Contact Point Phone,Contact Point Address, andParty Identification. You must assign the Profile category to any customer or prospect record you plan to unify through Identity Resolution or select as the primary target population ("Segment On") of a segment. - Engagement (The "What & When"): Represents time-series interactions, behaviors, and historical events—such as website page views, mobile app taps, email opens, support cases, and purchase orders. Because events only make sense on a timeline, Data Cloud requires you to designate a valid date-time field as the Event Time when setting up an Engagement stream. This timestamp drives behavioral segmentation, engagement scoring, and chronological activity tracking.
- Other (The "Reference"): Represents supporting lookup and reference datasets that are neither people/entities nor time-series events. Classic examples include product catalogs (
Goods Product), store branch directories, warranty tier definitions, or currency lookup tables. Objects in the Other category enrich Profile and Engagement records through data model relationships, but they cannot be unified as individuals or queried as standalone time-series event streams.
One common point of confusion is where to place customer orders or financial transactions. Even though people informally call orders "transactional data," an order placed by a customer on a specific date is an action that happened at a point in time. In Data Cloud's standard Customer 360 Data Model, Sales Order is modeled as an Engagement DMO so marketers and analysts can filter by purchase recency and order frequency. Reserve Other for static or slowly changing lookup attributes—like the product SKU details linked to each order line item.
Data Categories (Defined at Data Stream Creation / Planning Time)
├─ Profile ──── People, businesses, and identifiers (Individual, Contact Point Email, Phone).
│ Required for Identity Resolution and Segment Target selection.
├─ Engagement ─ Time-series events and interactions (Web clicks, email opens, Sales Orders).
│ Requires an Event Time field; powers time-windowed behavioral filters.
└─ Other ────── Reference and lookup datasets (Product catalogs, store directories, lookup tables).
Enriches Profile/Engagement DMOs via relationships; no Event Time or unification.
⚙ How Categories Govern Identity Resolution and Segmentation
Before creating a Data Stream in a real implementation, evaluate how your downstream consumers will query that object. Data Cloud strictly enforces category boundaries across its core engines:
- Data Mapping Compatibility: When you map fields from a DLO to a DMO, Data Cloud filters available target DMOs based on the DLO's category. A Profile DLO maps to Profile DMOs, an Engagement DLO maps to Engagement DMOs, and an Other DLO maps to Other DMOs.
- Identity Resolution Scope: Unification rulesets link fragmented source records—such as a CRM Contact, a Marketing Cloud Subscriber, and an e-commerce customer profile—into a single
Unified Individual. Identity Resolution only reads DMOs in the Profile category. If you ingest customer registration records asOther, they will never participate in match rules. - Segment Target vs. Related Attributes: When building a segment, the primary audience entity you "Segment On" must be a Profile DMO (for example,
Unified IndividualorIndividual). Once that target is selected, Engagement DMOs appear in the attribute library with time-window and aggregation controls (such as counting events within the last 14 days), while Other DMOs appear only as standard relational attributes when joined through a valid foreign-key relationship.
- Core Rule: Every ingested object belongs to one of three categories: Profile (entities and contact points), Engagement (time-stamped events), or Other (reference lookups).
- Key Benefit: Enables Data Cloud to automatically enforce schema rules—such as requiring an Event Time on events—so Identity Resolution and Segmentation work reliably out of the box.
- Primary Tradeoff: The category is locked when the Data Stream is created; fixing a misclassification later requires recreating the stream, mappings, and dependent segments or Calculated Insights.
- Boundaries and Limits: Only Profile DMOs can be unified or selected as a segment's target population, and only Engagement DMOs support native relative time-window filters.
- Anti-Pattern Mirror (Defaulting to "Other"): Data Cloud will happily ingest an
Otherstream on day one without errors, but teams later discover they cannot unify profiles or build recency-based segments from it. - Future Proofing: Keep Data Cloud's P-E-O Data Categories mentally separate from Service Cloud Knowledge Data Categories, which control article visibility and taxonomy.
⚠ Common Mapping Mistakes and How to Avoid Them
When reviewing a Data Cloud data model before go-live, watch out for these three frequent mistakes and verify each stream early:
- Ingesting Behavioral Logs as "Other": Teams sometimes select
Otherwhen rushing through a custom S3 or Ingestion API setup because they do not want to configure anEvent Timefield immediately. Always verify that any log containing customer activity has a clean ISO-8601 timestamp column and is created as Engagement. - Mapping Event Streams as "Profile": Putting high-volume clickstream or order rows into the Profile category treats every event row as a distinct individual record, bloating profile counts and corrupting Identity Resolution match results. Keep person/account attributes in Profile and move their actions into Engagement linked via a foreign key (such as
Individual Id). - Skipping Pre-Ingestion Validation: Before deploying a stream into production, open the Data Streams tab in Data Cloud, inspect the Category column on each stream, and confirm that every Engagement stream displays the intended
Event Timefield in its schema details. Catching a category mismatch before building Calculated Insights and Segments saves hours of rework.
🎓 Core Q&A and Interview Prep
Q: When setting up a Data Stream in Salesforce Data Cloud, you must choose a Category for each object. What are the three options, and why does getting this decision right early matter?
A: Choosing the category tells Data Cloud the structural role of the incoming dataset:
- Profile: Captures people, businesses, and their identifiers (such as
Individual,Account,Contact Point Email, andContact Point Phone). It is mandatory for customer records because Identity Resolution exclusively unifies Profile DMOs. - Engagement: Captures time-series interactions and events (such as website clicks, email opens, and orders). It requires a designated
Event Timefield and powers behavioral, time-windowed segmentation. - Other: Captures static or lookup reference data (such as a product catalog or store list) that enriches Profile and Engagement records without acting as a person or an event stream.
Getting this right during planning prevents costly rework: you cannot toggle a Data Stream from Other to Engagement after deployment, so a wrong choice requires creating a new stream and re-creating any Calculated Insights or Segments built on top of it.
❓ Scenario-Based Follow-Ups
Q1: Marketing reports they cannot build a segment for "users who browsed the site but did not buy anything in the last 7 days." What mapping mistake likely caused this?
A1: Evaluating a relative window like "in the last 7 days" requires Data Cloud to query the browse records as a chronological time series. That capability is only available when the website browse stream is configured under the Engagement category with a valid Event Time field. If the stream was accidentally created under Other (or Profile), Data Cloud does not treat the rows as time-stamped events, and the time-window filter will not work. The solution is to ingest the browse data through a new Engagement Data Stream, map the timestamp to Event Time, link the customer identifier to Individual, and rebuild the segment.
Q2: Are Data Categories in Data Cloud the same thing as Salesforce Knowledge Data Categories?
A2: No—they share a similar name, but they belong to completely different Salesforce features. Data Cloud's categories (Profile, Engagement, Other) define the architectural data type of ingested objects for identity unification and segmentation. Service Cloud Knowledge Data Categories are hierarchical classification trees used to organize Knowledge articles and control which user roles or portals can view them. In technical discussions or interviews, always specify Data Cloud or Knowledge upfront so there is no ambiguity.
Q3: Follow-Up Question — Once you choose the Profile category for customer data, how does Data Cloud distinguish between the person record itself and their email addresses or phone numbers during Identity Resolution?
A3: Inside the Profile category, standard Data Model Objects follow specific structural patterns—often referred to as Profile Category Types—such as Individual (the core person record), Contact Point objects (Contact Point Email, Contact Point Phone, Contact Point Address), and Party Identification (external loyalty IDs or driver's license numbers). During Data Mapping, you link the Contact Point and Party Identification DMOs back to the Individual DMO via a Party foreign key. Identity Resolution then traverses those relationships so match rules can compare emails, phones, and external IDs belonging to the same Individual without duplicating the core person record.
Comments
Questions, corrections and your own experience are all welcome.
Post a Comment