Dcsn Optimization Option Summary DMO

Stores per-arm performance analytics for decision optimizers (contextual bandits). Each record summarizes one arm’s behavior within a single evaluation cycle, including IPS-corrected reward rates, exploration/exploitation counts, credible intervals, and arm classification status.

Object API Name: std__DcsnOptimizationOptionSummaryDmo__dlm
Category: Unassigned
Availability: Available in 262 and later versions
Primary Key Field: Id

• cdp_sys_record_currency
• CredibleIntervalLowerNumber
• CredibleIntervalUpperNumber
• CumulativeRegretNumber
• CumulativeRewardNumber
• DataSourceId
• DataSourceObjectId
• DcsnOptmztnIdentifier
• DcsnOptmztnOptionIdentifier
• EmpiricalProbToBeatAllPct
• EmpiricalRewardRateNbr
• EvaluationCycleIdentifier
• EvaluationDateTm
• ExploitCount
• ExplorationRatePct
• ExplorerCount
• Id
• InternalOrganizationId
• MetricName
• ModelCalibrationScoreNbr
• PersonalizedLiftNumber
• PullCount
• RewardStdDevNumber
• RewardVarianceNumber
• SelectionRatePct

• Field API Name: std__cdp_sys_record_currency__c
• Data Type: TEXT
• Description: System-generated metadata field that stores a 3-letter ISO currency code.

• Field API Name: std__CredibleIntervalLowerNumber__c
• Data Type: DOUBLE
• Description: Lower bound of the 95% credible interval on the empirical reward rate (rate minus 1.96 times standard error)

• Field API Name: std__CredibleIntervalUpperNumber__c
• Data Type: DOUBLE
• Description: Upper bound of the 95% credible interval on the empirical reward rate (rate plus 1.96 times standard error)

• Field API Name: std__CumulativeRegretNumber__c
• Data Type: DOUBLE
• Description: Per-arm contribution to cycle-level pseudo-regret: PullCount times (best_rate minus arm_rate); measures reward left on the table by routing traffic to this arm instead of the globally best-performing arm

• Field API Name: std__CumulativeRewardNumber__c
• Data Type: DOUBLE
• Description: Total attributed engagement reward accumulated by this arm in the cycle, summed across all users with no IPS correction

• Field API Name: std__DataSourceId__c
• Data Type: TEXT
• Description: A unique, system-generated identifier for this record.

• Field API Name: std__DataSourceObjectId__c
• Data Type: TEXT
• Description: A unique, system-generated identifier for this record.

• Field API Name: std__DcsnOptmztnIdentifier__c
• Data Type: TEXT
• Description: Identifier of the decision option configuration this arm belongs to

• Field API Name: std__DcsnOptmztnOptionIdentifier__c
• Data Type: TEXT
• Description: Identifier of the personalization decision (option) being summarized

• Field API Name: std__EmpiricalProbToBeatAllPct__c
• Data Type: PERCENT
• Description: Monte Carlo probability that this arm has the highest reward rate among all arms in the decision option; primary win metric that drives arm classification

• Field API Name: std__EmpiricalRewardRateNbr__c
• Data Type: DOUBLE
• Description: IPS-corrected (SNIPS) average observed reward rate for this arm, de-biased to estimate true performance across all users

• Field API Name: std__EvaluationCycleIdentifier__c
• Data Type: TEXT
• Description: UUID identifying the specific evaluation cycle (Spark job run) that produced this record

• Field API Name: std__EvaluationDateTm__c
• Data Type: DATETIME
• Description: Timestamp when this evaluation cycle was computed

• Field API Name: std__ExploitCount__c
• Data Type: DOUBLE
• Description: Number of pulls where this arm was selected via exploitation (model’s top pick)

• Field API Name: std__ExplorationRatePct__c
• Data Type: PERCENT
• Description: Percentage of this arm’s total pulls that were exploration (ExplorerCount / PullCount)

• Field API Name: std__ExplorerCount__c
• Data Type: DOUBLE
• Description: Number of pulls where this arm was selected via exploration (random/epsilon selection)

• Field API Name: std__Id__c
• Data Type: TEXT
• Description: Primary key composed from DcsnOptimizerConfigurationId + DcsnOptionId + EvaluationCycleId

• Field API Name: std__InternalOrganizationId__c
• Data Type: TEXT
• Description: A unique, system-generated identifier for this record.

• Field API Name: std__MetricName__c
• Data Type: TEXT
• Description: Goal metric this summary is computed for taken from the recommender configuration

• Field API Name: std__ModelCalibrationScoreNbr__c
• Data Type: DOUBLE
• Description: IPS-weighted error between the ML model’s predicted reward and observed reward for this arm; IPS-weighted Brier score for Conversion mode or IPS-weighted MAE for Attainment mode; lower values indicate better calibration

• Field API Name: std__PersonalizedLiftNumber__c
• Data Type: DOUBLE
• Description: Ratio of average attributed reward under exploitation to average attributed reward under exploration for this arm; values above 1.0 indicate the model’s targeting adds value over random selection

• Field API Name: std__PullCount__c
• Data Type: DOUBLE
• Description: Total number of requests where this arm was selected (recommended) in this evaluation cycle

• Field API Name: std__RewardStdDevNumber__c
• Data Type: DOUBLE
• Description: Standard deviation of observed rewards (square root of RewardVarianceNumber), used to derive the standard error for credible intervals

• Field API Name: std__RewardVarianceNumber__c
• Data Type: DOUBLE
• Description: Cluster-robust variance of observed rewards for this arm, computed by collapsing per-(user, arm) before across-user aggregation to break per-user dependence

• Field API Name: std__SelectionRatePct__c
• Data Type: PERCENT
• Description: Percentage of total decision option traffic this arm received (PullCount / total PullCount across all arms)