Skip to main content

Customizing Invocations

You can customize your forecast request by specifying the following parameters, valid for both Demand Prediction and Lead Time Prediction:

  • start_date: the first date to forecast;
  • horizon: the number of time periods you want to forecast;
  • granularity: the reference period for which the forecasts are to be calculated;
  • aggregation: the taxonomy level for which to receive forecasts.
tip

You can for example request a demand prediction for the next 3 months, starting from January 2022, with a monthly granularity and item-level aggregation.

Mantissa will provide you with forecasts for each item in your inventory for January, February, and March 2022. Each value forecasted will refer to the estimated demand for that item for the corresponding month.

Start Date​

The start date parameter specifies the first date for which you want to generate forecasts. It allows you to define the beginning of the forecast horizon and obtain predictions starting from the specified date.

warning

Due to the convention used to refer to each period with the last day of it, the start date you provide will be adapted: if you request a forecast with granularity M (month) and start date 2022-01-01, the first forecasted date will be 2022-01-31.

This is only a convention used to refer to the periods, and it does not affect the meaning of the forecasts: in the previous case, the platform interpreted your requested starting date as you requested a forecast for January 2022, and for convention it refers to January as the last day of the month.

Horizon​

The horizon parameter specifies the number of time periods for which you want to generate forecasts.

tip

Forecasts become less accurate as the forecast horizon increases. It is recommended maintain your historical data up-to-date and request forecasts for a reasonable horizon to ensure the accuracy of the predictions. It is preferable to request forecasts for shorter horizons and update them periodically as new data becomes available.

Granularity​

Granularity refers to the level of detail or the period length for which predictions are generated. It allows users to customize the time frame over which forecasts are made according to their specific needs.

Available granularity options:

  • Week: If you select W as the granularity parameter, the platform will provide predictions for each individual week. This option is ideal for users who require short-term forecasts with a high level of detail.
  • Month: Opting for M as the granularity parameter will generate predictions for each calendar month. This choice suits users who prefer mid-term forecasts and need insights aggregated at the monthly level.
  • Quarter: Selecting Q sets the granularity to generate predictions for each quarter of the year. This option is suitable for users who focus on long-term forecasting and prefer insights grouped into quarterly periods.

When selecting the granularity parameter, consider the level of detail you need in your forecasts and the time frame over which you want insights. Choosing the appropriate granularity ensures that the predictions align with your business objectives and provide actionable insights for decision-making.

Aggregation​

The aggregation level parameter determines the taxonomy level at which predictions are generated. It allows users to customize the level of detail in the forecasts based on their specific requirements.

Mantissa supports three aggregation levels: item, classification, and family. Each level provides a different perspective on the forecasted demand or lead times, allowing users to focus on specific items, product categories, or broader product families.

Available options for aggregation:

  • Item: Selecting item as the aggregation level parameter generates predictions at the most granular level, focusing on individual items. This option is ideal for users who require detailed insights into the demand for specific items in their inventory.
  • Classification: Opting for classification sets the aggregation level to generate predictions based on product classifications or categories.
  • Family: Choosing family as the aggregation level parameter aggregates predictions at the broadest taxonomy level, focusing on product families or groups.
warning

The ability to retrieve forecasts for a specific aggregation level depends on the availability of data at that level in your historical records. If you do not provide data for a specific taxonomy level (e.g., family), you will not be able to request forecasts at that level. Ensure that your data includes the necessary information to generate forecasts at the desired aggregation level.