Platform

1. Data load error messaging
   • Improved error messaging has been introduced for data uploads across all entity types. When uploads fail due to missing mappings, incompatible file formats, or incorrect files, users will now see clear, actionable error messages with optional context to help quickly identify and resolve issues, reducing dependency on support teams.
2. Data load – Flexible column sequencing: The data ingestion process now supports flexible column sequencing for all entity types. Files can be uploaded with columns in any order, as long as all required attribute names are present.
   • Example: A file with Product Description | Product ID | Product Price | Product Cost will now be accepted even if the configured order is different. Missing attributes will still cause failure, while extra columns are ignored unless validation is enforced.
3. Fuzzy logic-based column name matching is now supported during data load to reduce failures caused by minor typos or formatting errors.
   • Example: A column named Descrption in the uploaded file will now be automatically matched to Description, allowing the data load to proceed without error.
4. The application now supports partial failure tolerance during bulk data loads. If one file in a multi-file upload fails, the system continues processing the remaining valid files without interruption.
   • Example: In a batch of 10 files, if the 4th file fails, files 5 to 10 are still processed and loaded successfully.

Demand planning

1. Skipped periods in the Forecast: The forecast time series view now displays all time buckets—including skipped periods—for both historic and future horizons. Skipped periods will remain empty (for both History & future periods.
   1. Example: If the current period is May 2025 and the planner chooses to generate a forecast only from August 2025 to August 2027, the application leaves the buckets for May, June, and July 2025 empty. No forecast values will be generated or displayed for these skipped periods, even though they fall within the overall forecast horizon.
2. Publish NPI results: A new "Publish" workflow has been introduced for NPI plans, allowing planners to push the Final NPI Forecast to the configured demand planning time series.
3. Disaggregation of NPI Plan: The NPI publish process now supports disaggregating product-level forecasts into detailed product–customer–location combinations using a selected like item’s historical relationships.

• When publishing the Final NPI Forecast for “iPhone 16,” the system copies the values to “Demand Plan Units”. It disaggregates them proportionally based on the lower-level distribution of the selected like item (e.g., “iPhone 15”), ensuring total consistency across all levels.

Replenishment

1. Co-manufacture scenario: Replenishment logic has been enhanced to support a two-step flow for co-manufacturers (Comans) based on updated product sourcing network (PSN) configurations. When demand arises at a distribution center like DC 1, the solver first checks internal transfers (e.g., from DC 2) and then evaluates Coman fulfillment through a stock transfer from the Coman to the DC 1, followed by a production order at the Coman site.
   1. This requires two PSN entries: one for the stock transfer (L→L) and another for internal production (M→L or M→M).
   2. Example: For a demand of 2000 units at DC 1, 200 units may be fulfilled from DC2’s surplus. The remaining 1800 units are transferred from Coman to DC 1, and a production order is triggered at Coman to replenish those 1800 units, ensuring end-to-end traceability and supply continuity.