Calling forecast engine with SMP parameter enabled

This section provides an overview of the required conditions to perform the SMP (Slow Moving Product) test using forecast engine.

When calling the forecast engine:

• The engine must use a passed algorithm other than Events only and the Minimum Periods for SMP parameter (smpMinPeriods) greater than zero.
• The engine checks the number of non-zero history values for an item and location at the defined Period Level for the forecast engine, if Minimum Periods for SMP (smpMinPeriods) parameter > 0.
  Note: 

  • The history length is defined as the number of historical periods starting from the first non-zero history value.
  • The periodicity is set by the engine parameter, and the Periodicity value determines the number of forecast periods, in a year.
• When the item-location history length is greater than or equal to Periodicity:
  • If “Number of zeros in the most recent periodicity’’/ periodicity >= smpMinPeriods / periodicity; the combination is considered Sparse, SMP. Else, non-Sparse, non-SMP.
• When the item-location history length is less than Periodicity:
  • If “Number of zeros’ / ‘history length’ >= smpMinPeriods / periodicity; the combination is considered Sparse, SMP. Else, Not Sparse, non-SMP.
• If the item-location combination is deemed sparse, SMP:
  • The forecast engine processes with the BATS SMP technique when Algorithm = BATS
    • Algorithm is returned as BATS (7)
    • The "SMP Chance Event" and "SMP Average Event" output parameters are populated, if mapped.
    • SMP flag is set to 1, if mapped.
  • The forecast engine processes with CROSTONS technique (fitModelGivenAlgorithm, algorithm=CROSTONS) when Algorithm <> BATS. In this scenario the passed algorithm value or the algorithm specified in the forecast engine settings is not considered.
    • Algorithm is returned as CROSTONS (4)
    • The "Mean Inter-Arrival Time" and "SMP Average Event" output parameters are populated, if mapped.
    • SMP flag is set to 1, if mapped.
  • The Best Fit algorithm processes with the CROSTONS technique.
    Note: If other algorithm is used in the forecast engine, the value is not returned for the above-mentioned output parameters.
• If the item-location combination is deemed not sparse (non- SMP), the engine is processed based on passed algorithm value or algorithm specified, in the forecast engine setting.