AN ADAPTIVE FORECASTING MODEL

Newspapers are considered the fastest moving consumer goods in the country,with a shelf life of only a couple of hours. Since the publishers carry theresponsibility for over or under stocking newsprint products in the various retailstores, they employ a number of ‘allocation’ staff who manually modify asimplified forecasting technique to ensure that the optimal number ofnewspapers are delivered to each retailer in Gauteng. As it is common for thecost of newspaper production to equal or sometimes exceed the newspaperretail price, the correct allocation of newspaper copies carries a financial benefitto the publishers. To achieve a more accurate copy allocation, this researchemploys the Holt-Winter forecasting methodology, coupled with an adaptiveclosed loop algorithm. The closed loop algorithm automatically adjusts the Holt-Winter forecast to ensure that the number of copies allocated to each of theretailers is optimised. This is done to maximise the newspaper sale volumewhile reducing the volume produced, thus minimising the number of unsoldnewspaper copies and the associated financial losses attributed to overproduction.The adaptive forecasts were produced on an aggregated and disaggregatedlevel. The aggregated method was used to forecast on a macro level, prior tothe allocation of the individual copies at each retailer. Whilst this methodperformed well in comparison to the existing heuristic forecasting method, theforecast accuracy actually lays in the correct allocation of the aggregatedforecasted result. The disaggregated method was used to forecast the numbernewspapers to be supplied at a retailer level. The adaptive disaggregatedforecasting method not only achieved a reduction in the volume of unsoldnewspaper copies, but also a reduction of sales volume. Therefore, neither theadaptive aggregated or disaggregated forecast models achieved the objectiveof maximising newspaper sales while simultaneously reducing the volume ofunsold newspapers.