Impact of Shipping Days

The number of shipping days in a company calendar is sometimes ignored.  Usually Wall Street analysts will reference it when referring to car sales and say something like sales are up x% but this year had y days more than the prior year.  Some companies are on a 5-4-4 calendar or 13 4-week months and then the number of shipping days is less impactful.  But a month such as August can have 21 shipping days if August 1 is a Saturday and 23 shipping days if August 1 is a Monday.  All things being equal that means sales in one August can be 10% higher than another August even if sales/day are constant over a number of years.  It also means sales in June or July could be 10% lower than in prior years compounding the error by distorting the seasonality.  Other months may have a number of holidays so everything being equal will ship less in those months.  This is less impactful at retail so when analyzing scanner data this is less of an issue since monthly forecasting is rarely something a retailer is concerned with.
Considering the number shipping days in a week or month is critical for ex-factory shipments as shipping facilities will often be closed for holidays so the amount of sales (which occurs when product is shipped) will be lower.  This is less critical for scanner data since retailers are usually open the same number of days per week and there is usually not a lot of value in comparing the impact of promotions across time periods.  When evaluating trade promotions at the retail level using scanner data, there is usually not any interest in tying back to overall retail corporate financials which is not the case with CPG companies.

The best way to handle differences in shipping days is to normalize all months to 20 shipping days and all weeks to 5 shipping days (or 7 days in some cases) and then pass the normalized sales data to the different models that are being used.  For forecasting or completing the analysis of the trade promotions it is necessary to reverse the process converting the normalized data back to actual or forecasted shipping days.  Review the DATES file used by SPAR to get an idea of how SPAR handles the data to address this issue which in my experience is better than alternatives that have been used.

In summary, a failure to take into consideration the number of shipping days can lead to inaccurate results as Trade Promotions run during periods with more shipping days will appear to be more successful and Trade Promotions run during periods with less shipping days can appear to be less successful.