What’s so Special about Leap Day?

Happy Leap Day!  February 29 seems weird.  It only happens once every four years, and we all tend to forget about it until it is upon us.   Went it arrives, we make jokes about “how old would you be if you were born on leap day?”  And, when we remember that the joke isn’t that funny, we realize that Leap Day is just another day, as it is today.

As this year’s Leap Day was approaching, a retailer asked me if it was a Special Day.  Meaning, should February 29 be marked in their workforce management system as a unique day that would be used to improve the accuracy of driver forecasts?

Fair question but the short answer is “no”. February 29 is only unusual because it happens once every four years but as I said a couple of paragraphs ago, today is just another Wednesday.  Looking at the National Retail Federation’s Retail Sales and 4-5-4 Merchandising Calendar, we can see that Leap Day is just the fourth day in the fourth week of a fiscal calendar.

While the Leap Day isn’t a Special Day, 2012 is a special year from a fiscal calendar perspective.  Looking again at the NRF calendar, you’ll notice that 2012 has 53 weeks in it rather than the standard 52 weeks.  This happens once every six years when the standard 364-day fiscal calendar needs to add a week to catch-up or synchronize with the Gregorian calendar.  Which leads to the obvious question:  how does your forecasting engine handle the 53^rd week? 

Of course, the answer depends upon the forecasting method(s) being used and how your WFM software vendor implemented that method.  But let’s consider a simple method that most systems use: trend of historic average.  This method looks back at a certain number of weeks and compares those weeks to the same fiscal weeks last year to establish a percentage that business has increased or decreased.  These percentages are then averaged and applied to the same fiscal week from last year that is being forecast.  The result of this calculation is the forecasted value. 

How your system handles the 53^rd week in this simple forecasting method depends upon how it was programmed.  When the system looks back to “same week, last year”, does it:

• look at a fiscal calendar table that says the fiscal weeks being looked at started on x date this year and y date last year?  If so, how does it look up the 53^rd week? The most likely scenario is that it uses either the 52^nd week of last year or the 1^st week of this year.
• simply look back 52 weeks (i.e., current week – 52 = last year)?  If so, the system will be compare different weeks (last year and this year).  When forecasting the 53^rd week for example, the system will actually look at the 1^st week of this year.  Making matters worse, this mistake may be present in every forecast for the next fiscal year (1^st week this year – 52 weeks = 2^nd week last year).  Again, whether or not it is present depends upon how the method was programmed.
• does it try to look back six years ago to the last time that a 53^rd week was forecast?  For many retailers, this is the desired result.  What happens if you don’t have six years of data in your system?  What does the system do then?

These are not hypothetical examples.  I’ve seen systems do all three.   

What does your system do?  You have plenty of time to plan for the 53^rd week but don’t wait until the 52^nd week to find out an answer because it