Sales Spotter : An Algorithm to Identify Sale Prices in Point-of-Sale Data

This paper develops an algorithm, called the sales spotter, which identifies the sale prices in the transaction price series provided in point-of-sale data. The goal of the sales spotter is to identify the maximum number of sale prices while minimizing the incorrect attribution of non-sale price reductions to sale prices. The spotter is developed and the values of its parameters are selected by analysing around 7.5 million flagged sales in a US supermarket scanner data. At the optimal values of the parameters, the spotter identifies 84% of authentic flagged sale weeks in the data