While it is widely believed that advertising media spending may not be optimal or efficient, it is not easy to measure the inefficiency, or the potential loss of sales due to inefficient ad spending. Advertising media spending efficiency needs to be measured and benchmarked. Here two approaches are presented and compared to evaluate top 100 advertisers' media spending inefficiency. The two methods are a nonparametric approach called Data Envelopment Analysis (DEA) and a parametric approach called Stochastic Frontier (SF). Results show that top 100 marketers' advertising spending in print, broadcast, and outdoor media are not efficient and could bring in 20% more sales. Results also show that the two methods may not always produce the same results. Therefore, it is suggested that both approaches be used in all applications.
It has long been noted that advertising practice might not be as efficient as it has been theorized (e.g., Aaker and Carman, 1982, Bass, 1979, MacNiven, 1980, Simon and Arndt, 1980 and Tull et al., 1986). The widely quoted statement by the first U.S. postmaster general John Wanamaker “I know half of my advertising is wasted, I just don't know which half” reflects the reality for many firms. Bass (1979) observed that advertising spending waste might be as high as 407% of the net income for some companies. High level of advertising inefficiency, defined here as the potential loss of sales due to inefficient ad spending, plagues businesses and frustrates managers Aaker and Carman, 1982 and Smith and Park, 1992 as inefficient media spending and misallocated resources contribute to lower profit margins and hurdle a company's ability to sustain a healthy growth Danaher and Rust, 1994 and Stankey, 1988. Given the critical importance of this practical issue, it is surprising that no study has empirically measured it, let alone compare and benchmark advertising inefficiency using multiple methodologies.
To fill in this void, this present study presents and compares two frontier or “best practice” methodologies that have been widely used to benchmark inefficiency. One is a nonparametric method called Data Envelopment Analysis (DEA), the other is a parametric technique called Stochastic Frontier (SF) method. DEA and SF have been extensively used to assess efficiency in various applications such as universities, banks, insurance companies, hospitals, salespeople, and countries (see Seiford, 1996, for a bibliography of more than 800 articles on DEA applications, and see Kumbhakar and Lovell, 2000, for an excellent review of the literature on SF models).
Advertising is generally undertaken to increase company sales and/or profits. However, a number of marketing scholars have theoretically addressed that there might be inefficiency in advertising spending (e.g., Aaker and Carman, 1982, Bass, 1979 and Simon and Arndt, 1980). This study was intended to apply two frontier methodologies (a parametric approach called SF and a nonparametric approach called DEA) to benchmark advertising spending inefficiency. Results of top 100 advertiser media spending inefficiency show that about 20% of the top 100 marketers' advertising spending in print, broadcast, and outdoor media could have been saved if they had advertised efficiently. The results also show that the two methods may not always produce the same results. Therefore, it is suggested that both approaches be used in all applications.
DEA methodology, by simultaneously dealing with multiple media spending inputs and outputs, is directed to evaluate each advertiser's media spending inefficiency. DEA identifies and estimates inefficiencies in each output and each input for each advertiser. An advantage of DEA is that it does not enforce an explicit functional relationship between media spending inputs and single output with fixed weights.
Unlike DEA, SF measures media spending inefficiency by averaging overall advertisers. In SF model, the error term is decomposed into random error and managerial inefficiency, removing much of the bias due to measurement error. Doing so may help improve the accuracy of all parameter estimates including inefficiency estimators. In SF analysis, a production function has to be subjectively specified to measure advertising inefficiency.
