C Appendix: Rules to Populate Out-of-stock and Outlier Indicator Measures

This appendix describes how to create rules to populate an Out-of-stock (OOS) and Outlier Indicator measures.

Rules Overview

In most RDFCS implementations, the out-of-stock and outlier flags are interfaced into RDF. However, not all retailers keep track of outages and outliers, or they are not very exact.

The following are some rules that populate these indicators. Note that they are a point of view, and you are encouraged to further refine them to fit your business needs.

OOS Rules Logic

If we call OOS the out-of-stock indicator and outliers the outlier flag, the logic to populate the flags can be:

If the rate of sales of an item is below a threshold, OOS is false.

Otherwise, make sure the item has been selling for a while and also does not have extensive periods with zero sales. If it is a new item, I do not want to start correcting the demand until the patterns become stable/predictable. Also, if the sales history has many zeroes, this may be an indication of a bigger issue, and the user may want to correct/take action in the Source Measure Maintenance workspace.

If these conditions are fulfilled, and the sales are still considered low, then mark the week as out-of-stock.

OOS Pseudocode

The pseudocode can look like:

If threshold 1a < rate of sales < threshold 1b   
If minimum number of sales periods > threshold 2a 
                        &&       number of periods with zero sales < threshold 3a
                         &&      sales < percent 1a * rate of sales   then OOS = TRUE else if rate of sales > threshold 1b
   If minimum number of sales periods > threshold 2b
                         &&       number of periods with zero sales < threshold 3b
                         &&      sales < percent 1b * rate of sales   then OOS = TRUE

Outlier Flag Rules Logic

For weeks with regular demand (no event, discount, and so on, is active), if the sales are deemed to be too high then mark the period as an outlier.

Outlier Flag Pseudocode

The pseudocode can look like:

If sales > rate of sales * multiplier
                         &&   Period is not promotedThen outliers = TRUE