FGS - HomeDemand ForecastingInventory PlanningDatabase ManagementReplenishment PlanningDistribution Planning
The Finished Goods Series
from E/Step Software Inc.
  The Right Calendar for Each SKU
Forecast Module
FGS - Home
    Key features
    Industries served

About E/Step Software

        Introductory Workshop
        Intermediate Workshop



  Forecast Model Elements
   Tracking Signal
   Fourier Series

  Compare service
   Set service
   Component SS

   Modules & Interfaces

Replenishment Planning

Distribution Planning

Don't be tied down to one calendar

FGS is the only system that lets you set each SKU to the most appropriate calendar for that SKUCalendar ReviewIn this example, the water pump installation gasket kit has low, lumpy demand of 36 units per month.  There is no trend and no seasonality.  You consider a different calendar to see if it will generate a better forecast model.   On the left is the result of a "calendar competition" that shows the forecasts with the least relative error on the top and the greatest on the bottom.  The current 5-4-4 monthly calendar has a standard deviation of 83 adjusted for the lead time and is second worst.  If you use a quarterly calendar the standard deviation would be 63.4, or 24% less (which also means the safety stock will be 24% less).

Calendar SimulationIn the screen on the right you can see the simulation of the forecast using each of two different calendars.  The lower window, on a quarterly calendar, has a slightly decreasing trend.

Since the residual forecast error reduces, so does the safety stock, from 108 to 82 units.  At $5.82 per unit, the value of the safety stock reduction is $148.  The service stays the same at 99.1%.

Most businesses have a mix of both high- and low-volume SKUs.  To achieve the lowest possible forecast error, high-volume SKUs should typically be forecasted more often (monthly, weekly, or somewhere in between), and the low-volume SKUs less often (bimonthly, quarterly, annually, etc.).  Forcing all your SKUs onto the same calendar raises your forecast errors and thus your inventory as well.

Calendar ErrorThe error distribution on the monthly calendar isn't quite normal, falling somewhere between the normal and exponential distributions.

Notice that the standard deviation ("Std") on the monthly calendar is 31.0, and 40.0 for the quarterly calendar.  At first glance one might think that the quarterly forecast has more error, not less.  But remember that it is relative to its own calendar.  Adjusted by the lead time, the quarterly error is actually less.  Plus, instead of monthly forecast reviews, the quarterly model is reviewed only once per quarter—a 66% reduction in workload.  Think what this means for your operations—you get better results but with less effort!