Promotions forecasting
accuracy was low leading to inventory leftover or stock-out
positions for promoted SKUs. Forecast error was measured
at 22%-24% in the case of inventory leftover after promotions
and there was a 16% chance of stock-out situations during
promotions.
Promotions forecasting was a manual, highly labor
intensive and specialized function. The client received
electronic and manual spreadsheet transmissions detailing
upcoming promotions and merchandizing activity from
retailers and had to forecast demand for the items on
promotion based on past history of similar promotional
events and planned merchandize activity.
The primary objective was to redefine the promotions
forecasting process and improve forecasting accuracy
thereby realizing higher returns. To achieve this, a
promotions forecasting product solution needed to be
evaluated and implemented. This could utilize category
specific demand forecasting models to make promotions
forecasts thereby improving forecast accuracy. This,
in turn, would eliminate the use of highly error prone,
complex and labor intensive manual calculations using
traditional Excel based macros.
The promotions forecasting solution would need to:
 |
Forecast demand for
unique promotions every week across all retailers
totaling to 20,000 SKUs per week |
 |
Continually adjust forecasts,
with attributes (e.g., price, placement etc.) for
forecast of 300 to 500 SKUs changing constantly
prior to the launch of the promotion |
 |
Improve promotions forecasting
accuracy so as to improve customer (retailer) service
level |
 |
Reduce inventory leftover by
over-forecast of promotions |
 |
Reduce stock-out positions caused
by under-forecast of promotions |
The client would also expect a significant reduction
in leftover inventory which accounts for nearly USD
100 million in inventory investment per week. |