Individual and Contact Point Email are Profile DMOs—they define who. Website clicks are an Engagement DMO—they define what happened and when. A product catalog is Other. If you mis-categorize engagement data as a profile, the system will break down trying to "unify" click events as if they were people.
The Ask: Marketing requested a simple audience segment: "People who browsed the website but did not buy anything in the last seven days."
The Failure: The tech team couldn't build it natively. They tried clunky workarounds—creating calculated insights and eventually doing a nightly export to a spreadsheet.
The Reason (The Old/Bad Way): The website browse events had been mapped to the Other category instead of Engagement. Only Engagement data has the time-series capability required to build a time-windowed segment (like "last seven days"). Because categories are set during the initial mapping phase and are hard to change later, every workaround was just a desperate attempt to recreate a feature that was accidentally switched off months ago by a simple dropdown choice.
The Fix (The New/Good Way): The team remapped the browse stream to the Engagement category, ensured there was a proper event timestamp field, and rebuilt the mappings. Once the category was correct, building the segment took exactly ten minutes.
The Core Concept: P-E-O
When you map a data stream in Salesforce Data Cloud, you must assign each object a Data Category. This tells the platform what role the data plays and drives how it is modeled, unified, and utilized.
- Profile (The "Who"): Describes a person or a business. This includes entity records and their identifiers (e.g., Individual, Account, Contact, Email, Phone). The Profile category is mandatory for any object representing people, because Identity Resolution and unified profiles are built exclusively from this data.
- Engagement (The "What & When"): Represents time-series, event-oriented data. This includes web clicks, email opens, bookings, orders, and customer service cases. Engagement data must have an event time field. It powers behavioral segmentation and real-time signals.
- Other (The "Reference"): This is reference or transactional data that is neither a person nor a time-series event. Think of lookup tables, product catalogs, or store location lists. It enriches your data but is not a profile or an engagement event on its own.
Decide your categories during initial planning. If you treat engagement events as profile data, Identity Resolution fails. If you map engagement as "Other," behavioral segments break. This is why category selection is a foundational modeling decision, not just a setup form field.
Data Categories (Set at Mapping / Planning Time) ├─ Profile ──── The person/business + identifiers (Individual, email, phone). MANDATORY for people. ├─ Engagement ─ Time-series events (clicks, opens, orders). Requires an event-time field. └─ Other ────── Reference/transactional data (catalogs, lookup tables, store lists).
The category is chosen when you map the stream. "Hard to change" has a very real meaning here: you cannot simply edit the category on a mapped field later. You have to remap the field, which means any insights or segments built on top of it must be rebuilt as well.
- Rule: There are three categories. Profile is who. Engagement is what happened and when. Other is everything else.
- Gain: The category tells the platform exactly what the data is used for, allowing automated matching and audience building to function smoothly.
- Price: It is set at mapping time and requires significant rework to change later.
- Limits: Profile data carries the person and is mandatory for Identity Resolution. Engagement is time-series and is strictly required for time-windowed segments. Get it wrong, and behavioral segments simply cannot be built.
- Mirror (Tagging everything as Other): The system won't complain on day one. But later, nothing will be able to be matched or segmented.
- Future Proofing: Do not confuse these with Salesforce Knowledge Data Categories, which organize articles for visibility. They share a name but do completely different jobs.
Core Q&A & Interview Prep
Q: When setting up a stream in Data Cloud, you are asked for a Category on each object. What are the choices, and why does getting it right early matter?
A: Each category tells Data Cloud the fundamental role of the data:
- Profile: Represents the person or business (e.g., Individual, Contact) and their identifiers (Email, Phone). It is mandatory for any people data because Identity Resolution relies exclusively on Profile data to create unified individuals.
- Engagement: Represents time-series, event-oriented data (clicks, orders). It requires an event-time field and is the only category that can power behavioral segmentation and real-time triggers.
- Other: Represents reference data like product catalogs. It enriches data but isn't a profile or an event.
It matters early because tagging engagement events as Profile will corrupt your identity matching, and tagging them as Other will make behavioral segmentation impossible. It is a planning-phase decision, not a runtime afterthought.
Scenario-Based Follow-ups
Q1: Marketing complains they can't build a segment for "users who browsed but didn't buy in the last 7 days." What mapping mistake likely caused this?
A1: This is a classic behavioral, time-windowed segment. It requires the browse events to be ingested as Engagement data with a valid event-time field. If the events were mistakenly mapped as Profile or Other—or if they lacked a usable timestamp—Data Cloud cannot process them as a time series. Because it isn't recognized as time-series data, the "last 7 days" parameter cannot be evaluated. The fix is to remap the browse events to the Engagement category.
Q2: Are Data Categories in Data Cloud the same thing as Salesforce Knowledge Data Categories?
A2: No, this is strictly a name collision. Data Cloud's categories (Profile, Engagement, Other) classify the architectural type of data in your model for resolution and segmentation. Knowledge Data Categories are used to organize and control the visibility of articles in Service Cloud. In an interview, always clarify which product you are talking about before using the term to ensure you are answering the right question.