Optimize & Improve the Computed Attributes Value Distribution Minimum Success Rate

Hi ​@sessoms25 

The answer depends on what is meant by "improve":

• Lowering the required minimum below 30%: There is no documented customer-facing setting to change this threshold. If the 30% rule is enforced for the current organization or feature path, it is a service-side guardrail and cannot be adjusted from the Computed Attributes UI.
• Increasing the observed success rate above 30%: This may be possible by improving the source data and Computed Attribute definition, but that does not change the minimum threshold itself.

The following items should be validated:

1. The source must be a Profile-enabled Experience Event dataset.
2. The event identities must resolve to the intended profiles through the configured identity and merge-policy setup.
3. The event filter should not be more restrictive than the business requirement.
4. The lookback period must include the events required for the calculation.
5. The aggregated field must be populated consistently and have a data type supported by the selected function.
6. Check the Last evaluated timestamp and Last evaluation status before judging the result. Only events available before the last successful evaluation are included in that run.

Also, Value Distribution and Sample Profiles should not be treated as the only indicator of whether the Computed Attribute populated successfully. Value Distribution is generated from sampled data, while Sample Profiles show values for sampled profiles. These views can therefore differ from an individual profile lookup.

If the use case naturally applies to fewer than 30% of profiles, changing the event rule merely to meet the threshold may produce incorrect business results. In that situation, a better design may be to calculate the derived value with Query Service or Data Distiller, write the result to a dataset, and make that data available to Real-Time Customer Profile where appropriate. Query Service supports creating query results as a dataset for ingestion into Profile.

Therefore, the practical conclusion is: the 30% minimum itself is not customer-configurable, but the qualifying population and observed success rate can sometimes be improved through data quality, identity resolution, lookback, and rule-definition changes.