Elasticity Demand Models are at the heart of the Retail Express solution which uses a demand model calibrating engine, capable of building reliable demand elasticity models from past sales transactions. The AMP2 demand engine uses daily/weekly data to automatically determine the best mathematical model based on goodness-of-fit and other statistical performance metrics.
Demand model equations are built for each SKU either for individual stores or a group of stores taking into account the following:
The pre-processing phase cleans and prepares the data before the latter is parameterised. Outliers are identified and may be corrected or in some instances discarded if no explanatory information can be associated to explain extreme variations.
During model build, AMP2 estimates regular price and promotional offers (price or other incentive e.g. BOGO) as well as direct and cross price/sales elasticity coefficients. These demand models identify regular and promotional complementary and substitution (cannibalisation) effects between certain products. The sensitivity in identifying cross effects can be adjusted using a filtering parameter.
Cross effects are systematically revised as the models are recursively updated when new sales data becomes available. AMP2 adjusts sales data with "lost sales" as long as out of stock information is systematically collected (start/end time) and made available during model build or update. AMP2 also differentiates the impact of different promotion types as long as historical information of such promotions is available to "train" the models.
Retail Express strongly believes that regular and promotional sales must be forecasted together as one thing impacts upon the other (e.g. promotion cross-effects on the sales of certain non-promoted products). The AMP2 modelling engine automatically estimates and updates complementary and substitution effects between products thanks to a recursive model update procedure.
The science and technology by which the data and algorithms combine to provide data models and forecasts for specific instances and conditions.
The provision of forecast on a real time and scheduled basis to support Merchandizing and supply chain users decision making and automated replenishment and ordering processes.
