ℹ️ Internal article. This action is an internal CRS operations job and is not intended for customer use. Contact support if the functionality is of interest to you.
Price suggestions. Trains an AI demand model on sales history in the BI database and writes guardrail-checked sales price suggestions to the ml_pricing_suggestions table for Power BI, without touching Sapera.
Price-suggestion evaluation. Closes the loop on price suggestions by detecting whether each one was applied, ignored or changed differently, and backfilling realized quantity and contribution margin from later sales in the BI database.
The action is an internal operations job that runs on a schedule at CRS. It requires administrator rights and direct database access and is not part of a normal customer setup.
Parameter | Explanation |
| Connection string to the BI/Power BI database that holds the sales history and receives the price suggestions. |
| System name of the sales price type holding competitor prices, used as extra model features for price sensitivity. |
| Maximum model mean absolute error allowed for a SKU to be considered trustworthy enough to suggest a price. |
| Guardrail capping how far a suggested price may move from the current price, as a fraction. |
| Guardrail requiring a minimum absolute profit gain in kroner before a suggestion is kept. |
| Guardrail requiring a minimum gross margin before a price suggestion is emitted. |
| Guardrail requiring a minimum relative profit gain before a suggestion is kept. |
| Minimum number of historical sales rows a SKU must have before it is eligible for a suggestion. |
| Optional organizational unit id to optimize for a single OU; leave empty to process every OU that has enough data. |
| System name of the protected ordinal price type; products currently using it are skipped because Sapera auto-overrides their price. |
| Half-life in days for the recency weight applied during training; set to 0 to weight all rows equally. |
| Optional date (yyyy-MM-dd) to predict prices for; defaults to tomorrow. |
| How many months of sales history to train on; older rows are down-weighted so recent patterns dominate. |
Parameter | Explanation |
| Connection string to the BI database holding the price suggestions to evaluate. |
| Number of days after the target date over which realized sales are measured. |
| Minimum age in days a suggestion must reach before it is evaluated, giving time to apply the change and gather sales. |
| Price match tolerance in kroner used to classify a suggestion as applied or not applied despite commercial rounding. |
See also the overview SyncTool – actions.
