| Summary: | Managing the advertising function for established brands requires an understanding of the nature of the advertising-sales relationship. Historically, both experimental and non-experimental approaches have been used to investigate this relationship, but the impressive amount of literature in this area seems to have identified only a number of broad generalizations. In part, this is due to the inadequacies of the different methodologies and data sources that have been used, which make difficult a comparison of the reported studies for the purpose of establishing guidelines for strategic advertising management. Continuous panel-based experimental research seems to offer greater potential for providing further insights into the nature of the advertising-sales relationship. The research first investigates the appropriateness and sensitivity of a number of models in identifying and quantifying the effect of changes in advertising strategy on sales, using The Test Marketing Group's (TMG) consumer diary and scanner panel data. It is shown that the ability to identify an advertising effect, referred to as the system's sensitivity, is significantly influenced by a number of factors, and that it can be predicted from the number of purchase transactions of the test brand. By using one specific model, thirty-five advertising strategy tests are analyzed at the aggregate, panel level, in order to estimate the probability of causing an advertising effect on all panelists, and to identify factors that influence the effect. Application of this methodology represents the first consistent analysis of a collection of historical data with the objective of developing a knowledge base regarding advertising strategy making and testing. It is found the probability of causing an advertising effect does not differ between copy and weight tests, but that a change in copy carries a significant risk of causing a negative effect. Increases in weight are particularly effective in causing a positive effect for small share brands. among the tests that are analyzed there is a 37.1% probability of observing an advertising effect at the panel level, which is lower than the probability observed in the literature. Subsequent analysis of the same tests examines the effect of a change in advertising strategy at the disaggregate level, that is, on certain segments of panelists. The results of this analysis show that significant advertising effects are observed more often, thereby increasing the probability of observing an advertising effect to 60%. Thus, by applying one methodology consistently across a set of panel-based advertising strategy tests, it is possible to identify a number of empirical norms that can aid managers in determining effective advertising strategies for their established brands. This so far has been difficult to derive from reported advertising studies. It is also suggested that further insights into the advertising-sales relationship can be obtained by increasing TMG's ability to specify advertising exposure. An experimental data collection system developed and tested on the basis of this further research is presented and evaluated.
|
|---|
