What is the Impact of a 53-Week Retail Calendar in RDF? (Doc ID 745534.1)

Goal

With regard to the RDF (Retail Demand Forecasting) application, what is the impact of having a 53rd week in the retail calendar?  What special arrangements must be made in the configuration in order to properly generate forecasts in a 53-week year?

Background information for issue clarity:

The "53rd week" issue results from the manner in which most retailers set up their fiscal calendars. The basic idea is that the full 52-week year is broken up into four 13-week periods (quarters), and these 13 week periods are further broken down into what we might consider "months". The most common way to arrange these weeks is a "4-4-5" format, which states that for a particular 13-week period, the first "month" contains 4 weeks, the second "month" contains 4 weeks, and the third "month" contains 5 weeks. This pattern would then repeat for the remaining 13-week periods in the year. Note that this arrangement may or may not correspond exactly to how the days/weeks actually fall within the Gregorian calendar, as certain weeks in the 4-4-5 scheme may actually straddle two Gregorian months (e.g., the fourth week in "Month 1" may actually contain days from both "January" and "February", depending on how these days fall). Also note that some retailers use a "4-5-4" scheme; basically the same thing, but just a slightly different placement/order of the "months" containing 5 weeks instead of 4.

The big advantage for arranging the calendar in a way similar to this is that each week is treated the same, in that the end date of each period is always the same day of the week. This makes retail planning much easier.

The problem with using a system such as this is that it only accounts for 364 days in a year (7 days * 13 weeks * 4 quarters). This means that as the calendar rolls forward, one day in every 365-day year is left unaccounted for (and more in leap years). So every 5 to 6 years or so, an extra week (the "53rd week") must be added to the retail calendar to make up for this.

Forecasting and planning applications can be affected by this, as this "53rd week" by default has no corresponding week in the previous year of history to which to map cleanly.

Solution