Limitations of Multivalue Fields in CRM Analytics

Limitations of Multivalue Fields in CRM Analytics

CRM Analytics does not fully support multivalue fields across all operations. The following limitations are documented to help you plan your data architecture and avoid unexpected behavior. Where possible, restructure your data to avoid relying on multivalue fields for joins, groupings, or compute transformations.
For technical background on SAQL (Salesforce Analytics Query Language) performance with multivalue fields, see SAQL Performance Considerations for Multi-Value Dimensions.

Filter and Security Predicate Support

• Supported in Filters — Multivalue fields work in filters. When defining a filter on a multivalue field in the UI, searching within the values for filter criteria may return unexpected results. This does not impact the functionality of the filter itself.
• Supported in Security Predicate — Multivalue fields work in Security Predicates. Multivalue fields have a limit of 32,000 characters. When used in a security predicate and unable to find a value, the value is likely truncated due to this character limit. To verify, use the array_join() function in a recipe transform node to convert the multivalue field to a string and check whether the value is truncated.

Unsupported Operations

• Not fully supported in UI and code — Faceting, Interactions (Bindings), and SAQL queries involving multivalue fields will likely fail or produce unexpected results.
• Not supported in XMD (Extended Metadata) — The "isMultiValue" parameter fails to validate if included in XMD 2.0.
• Not fully supported in ValuesTable — A ValuesTable displays only one value from the array rather than the full list of values.
• Not fully supported in Record Action Display Fields — The record action disambiguation screen displays one value rather than the full list if a multivalue field is included in the Display Fields. Records to disambiguate are shown in table format, which does not render multiple values.
• Cannot be used as Augment or Join keys — The left or right key of an Augment (in a dataflow) or a Join (in a recipe) cannot be a multivalue field.
• Not supported in computeExpression and computeRelative — Multivalue fields are not supported in computeExpression and computeRelative nodes in a dataflow. When ingesting a datasource that includes a multivalue field and a compute node is added downstream, the multivalue field may lose its multivalue attribute and be converted to a Text field instead.
• Not supported with Sharing Inheritance — If an object's Id is used in a multivalue field, Sharing Inheritance does not work for that object.
• Not supported in Einstein Discovery — Multivalue fields are not supported in Einstein Discovery analyses.

Performance Considerations

• Negative impact to Augment performance — Including multivalue fields in the "right_select" parameter of an Augment may cause Augment processing time to increase significantly.
• Negative impact to sfdcRegister performance — Registering a dataset that includes one or more "operation": "LookupMultiValue" Augments results in a significant increase to processing time.
• Negative impact to Grouping performance — Multivalue fields do work when used as a grouping in a Lens or Query, but may cause negative impact to query performance. Do not use multiple multivalue fields in nested groupings — this produces a Cartesian product that is likely to cause query timeouts.
• Negative impact to List widget performance — Including a multivalue field in the query for a List widget may cause negative impact to query performance.

Data Prep Limitations

Multivalue fields may not display or be available in Data Prep recipes. Specific limitations include:

• Join Nodes cannot use multivalue fields as keys.
• Transform Nodes do not support multivalue fields, with the exception of multivalue formula functions.
• Append Nodes do not support multivalue fields.

Other Known Limitations

• Dates lose date part associations — Attempting to create a multivalue field from a Date field disassociates the individual date components and results in meaningless data. A dataflow warning appears if this is attempted.

Recommendations

Augments

Restructuring the grain direction to avoid multiple matches is the optimal approach when working with Augments.

Multiselect Picklists

• Multiselect Picklist field values are ingested as a single, delimited string. Values may be truncated or missing if the Precision of the field is not set high enough. Ensure that the Precision is set high enough in the sfdcDigest metadata to allow the full set of included values to be ingested.
• If you want to see all field values in a table (rather than a single value from the array), digest the field as a single value field: set the "isMultiValue" parameter to false and remove the "multiValueSeparator" parameter from the sfdcDigest JSON or Dataflow Editor. This ingests the field contents as a single string rather than separate values. While this prevents the field from being effective in groupings, a "contains" filter allows for proper filtering.

External Data

Set the "isMultiValue" parameter in the external metadata schema to false. This ingests the field contents as a single string. Using "contains" filters allows for proper filtering. If grouping is needed, duplicating the field as both a multivalue field (for grouping) and a single value field (for other functions) may be the optimal approach.

General

It is not recommended to rely on multivalue fields for necessary data operations in CRM Analytics.