When creating a new data stream in Data Cloud for data ingestion, you need to follow a six-step process. In the third step, you will reach the Schema Review dialog page, where you define the data stream “Category” property on the left-hand panel (alongside selecting fields, and optionally creating formula fields on the right-hand side).
The data category defines how the data from this data stream will be used in the Data Cloud data model. Beware, as this is not a trivial task – once you make your selection, this data stream’s category cannot be changed! (In that case, you would need to delete the data stream and start over again.)
Data Stream Categories
There are three data stream types to choose from: Profile, Engagement, and Other. Let’s take a look at when to use each.
Note: The term “attributes” will be mentioned in this guide. These are essentially the fields where Data Cloud stores data. When creating a data stream, you’re creating a Data Stream object (DSO). When the stream is active, data from the stream will arrive in a Data Lake object (DLO).
Once ingested, you will model data to your needs by mapping Data Stream objects (DSOs) to Data Model objects (DMOs) – in other words, mapping attributes into a relational data model.
When you reach the data stream schema review a dialog page, Profile will be the default selection.
Here are ‘rules of thumb’ for keeping Profile selected:
- The data is demographic, i.e. it contains contact points (e.g. email, phone, address, etc.) and/or profile attributes (e.g. shoe size, gender).
- The data set contains individuals with an ID that enables you to uniquely identify each row of data, e.g. CRM record ID. This will become the individual ID value, which will be used to link Data Model objects (DMOs) to other DMOs at a later stage.
- By its nature, Profile category data will make a good jumping-off point for profile unification (compiling a ‘golden record’), and later, segmentation as you will be working with ‘people’ records whose data can be blended with Engagement (and Other) data.
Engagement data is behavioral and is generated by individuals as they interact with your brand – think website activity, online purchases, etc. Once ingested and modeled, you can stitch together this series of engagement points to determine an individual’s behavior over time – what they like, and how likely they are to purchase a product or service.
Being behavioral means that the engagement happens at a certain point in time. A DateTime field in the dataset represents this.
The characteristics of Engagement data streams can be summarized as:
- A DateTime field exists in the dataset. This will be defined as the Event Date field in your data stream schema – essentially making this field the identifier for the engagement (i.e. the value that distinguishes one row in the data table from another).
- The Event Date field can’t be changed after you set up the data stream. Instead, you will need to delete the data stream and start over.
- For a DateTime field to be a valid Event Date value, you must include the time part and the time zone. Failing to include the time zone will mean it will default to the UTC zone, which could be very inaccurate.
However, there is one concept that is key above all else: the Event Date value is immutable – meaning that this DateTime value won’t change and is frozen in time. Immutable values represent the point of the engagement happening.
- Examples of immutable values: Transaction date, subscription sign-up date.
- Examples of mutable values: Last updated date, file extract date.
Knowing the difference between immutable and mutable values will help you understand the purpose of the ‘Other’ data category.
The Other category caters to the following scenarios:
- Mix and match of data types: You may have a data set with a combination of profile and engagement data
- Mutable engagement data: You have engagement data that contains mutable date fields (i.e. DateTime fields that can change).
- Miscellaneous data: Data that’s neither demographic nor behavioral, e.g. airport locations and codes.
Additional Facts and Considerations
- Once you make your selection, the data stream’s category cannot be changed (you would need to delete the data stream and start over again).
- The same applies when deciding on an Engagement data stream’s Event Date (delete and start over).
- Once you have chosen your category on the schema review dialog page, the next step presents a list of fields with their suggested data types for the Data Source object (DSO), including where you will select the Event Date for Engagement data streams. Remember that the DSO acts as a foundation for the data stream.
- When it comes to modeling data (i.e. working with Data Model objects), the DMO’s category will be inherited from the first DSO that’s mapped to it.
When setting up a new data stream in Data Cloud, it’s crucial to carefully consider the data category you select during the Schema Review process. The Profile category is ideal for descriptive data about individuals or accounts, while the Engagement category is perfect for recording interactions and behaviors. If the data doesn’t fit into either of these categories, the Other option is available.
Keep in mind that once chosen, the data stream’s category cannot be changed, so it’s important to choose wisely. By selecting the appropriate category, you can effectively structure your data and leverage the full potential of Salesforce Data Cloud for enhanced insights and personalized customer experiences.