Optimization - Price Guidance 1.1.0

1.1.0

2026-10-06

Added a Data Split column to the Transaction Data table. It shows whether each transaction is in the training, test, or calibration set.

PFPCS-11142

Combined the Feature Impact charts into one chart. It shows floor, target, and stretch impacts as lines over the transaction count bars.

PFPCS-11111

Added the Mean Estimate, Achieved Percentile, Absolute Error, and Pseudo R² metrics for Floor/Target/Stretch models to the Metrics portlet.

PFPCS-11059

Training outputs of the floor, target, and stretch models are now stored in shared tables, with a model key column that identifies each model.

PFPCS-11020

Added Mean Absolute Percentage Error (MAPE) to the Metrics portlet, for the training and test sets.

PFPCS-11004

The Summary of the features portlet now shows the minimum and maximum of each feature, and the cardinality of each categorical feature.

PFPCS-10976

Introduced the training_predictions table to support model training data storage.

PFPCS-10961

Added the Change Overview and Revenue and Profit Comparison charts to the Metrics tab of the Results step.

PFPCS-10954

Model evaluation returns floor, target, and stretch values for Floor/Target/Stretch models, in the API and on the Evaluation tab. The crossing guard keeps the floor and stretch on the correct side of the target.

PFPCS-10944

Floor/Target/Stretch models are trained with EBM quantile regression. Training fails for a price level when its achieved percentile differs from the configured percentile by more than coverage_tolerance.

PFPCS-10943

Added the Strategy Positioning section to the Feature Selection tab. Select the Floor/Target/Stretch strategy and set the Floor, Target, and Stretch percentiles.

PFPCS-10931

Implemented EBM Quantile Regression models to provide interpretable, calibrated pricing corridors within the production environment.

PFPCS-10911

The system fails to generate random samples for extended-value classification because the random index incorrectly rounds to a binary value.

PFPCS-11082

The system fails to store or export the learned contribution for null values in numerical features, causing a discrepancy between trained model predictions and evaluation results when null values are encountered.

PFPCS-11034

Model evaluation returns predictions that do not match the training_predictions table for items with high-cardinality categorical features or numerical pairwise interactions.

PFPCS-11033

The feature impact chart tooltip for numerical features fails to display the number of transactions.

PFPCS-10993

The system prevents users from selecting Quantity and Integer field types for the Weight field in the Definition step.

PFPCS-10985

The feature impact chart displays incorrect transaction density and inconsistent binning, causing data to cluster at the extremes rather than reflecting the actual distribution across the model's score bins.

PFPCS-10938