Comment on measuring short‐ and long‐run promotional effectiveness on scanner data using persistence modelling

Abstract

In their paper, Dekimpe and colleagues support the view that a major step forward seems to have been accomplished since the time when Blattberg et al. [1] stated that long-term effect is 'probably the most debated issue in (the) promotional literature and one for which the jury is still out'.On the methodological front, I have few comments. First, persistence modelling offers an advantage over standard econometric modelling: impulse response function, associated with VARX modelling, avoids the need to predetermine the long term impact of such variables as promotion or advertising.A second remark is related to aggregation. Aggregation bias due to heterogeneity in promotional activity across stores can be an issue. This aspect has been discussed in econometric modelling but, to my knowledge, it has not been analyzed in persistence modelling. The authors are aware of this shortage and provide some comparisons which aim to demonstrate that the potential biases seem to be low [2]. Similar questions arise with the aggregation across SKUs up to the brand level. I agree with the conclusion that additional work needs to be done on this issue.Finally, an important methodological aspect is the fact that dependent variables other than sales or market share need to be investigated, e.g. own-price elasticity. The methodology used here does not make a distinction between base-price (regular price) elasticity versus temporary price cut (TPR) elasticity. A price variation whether it is temporary or not, if substantial, triggers an impulse shock. As a price-down repositioning can be considered as a TPR during the following few weeks, it seems fair to state that the authors measure a sensitivity to temporary price reduction. An important question is then: what additional key metrics need to be considered? Base price elasticity is possibly another one. We see below that this shift can lead to potential difficulties.As a practitioner, I consider that perhaps, the most important contribution of this research relies on the large number of product categories studied. This enables the authors to perform a meta-analysis and to provide benchmarks to practitioners. To summarize a long list of useful