Beruflich Dokumente
Kultur Dokumente
This article examines the generalizability of convex-shaped advertising response functions. Using single-source data, we anaiyzed the response functions of brands in four consumer goods categories. This study supports the prior finding that convex response functions are typical, but not universal. While the convex response function is found to apply across a range of conditions, more work is needed to understand measurement issues, exceptions, and boundary conditions.
Ehrenberg-Bass Institute for Marketing Science University of South Australia Jennifer. taylor@ marketingscience.info
RACHEL KENNEDY
Ehrenberg-Bass Institute for Marketing Science University of South Australia rachel.kennedy@ marketingscience.info
BYRON SHARP
Ehrenberg-Bass Institute for Marketing Science university of South Australia byron.sharp@ marketingscience.info
ing two customers once, or one customer twice? It is an extremely important issue for media planners. Single-source analysis has delivered an empirical generalization (EG): the immediate advertising sa les-response function is convex, which means it is better to reach more consumers once. This empirical law appears to generalize for television advertising for established consumer goods brands in mature markets. It is such an important discovery, however, that it deserves further investigation. In this article, we present new findings from single-source analysis in four consumer goods categories. They support the generalization by showing a dominance of convex-response functions. The convex functions are not universal, however, and we have yet to identify the boundary conditions associated with deviations from the law. Much research needs to be done before this important law can be further refined.
HOW ADVERTISING EXPOSURES AFFECT BUYING RATES
cle, is on the immediate sales impact measured by longitudinal, individual-level single-source data. One of the most exciting and important EGs to come from single-source analysis is that the sales response to advertising exposures is convex (see Figure 1). Convex and S-shaped response curves imply that, after a certain level of exposure, consumers become saturated by advertising, and there is no further increase in their propensity to purchase a brand. And an S-shaped curve supposes that there is a threshold level of exposure, beneath which the advertising is ineffective. By contrast, a linear shape would imply that repeated advertising exposure produces a proportionately increasing propensity to purchase a brand (Roberts, 1996). The convex response function was summed up by Jones (1992) with his "once is enough" catch phrase: within a short window before category purchase, an advertiser gaias greatest return from reaching the consumer with a single advertising exposure. This pattern has significant implications for media planning, suggesting buying commercial spots with an emphasis on maximizing reach, and for any given level of frequency of advertising exposure. Although authors like Broadbent (1999), Roberts (1996), and Jones (1992) have found that there may be exceptions to convex-response functions, others (such as Vakratsas and Ambler, 1999) suggest that the weight of empirical findings makes it possible to generalize that the advertising-response function is convex.
DOI: 10.2501/S002184990909028X
The advertising-response function can be defined as "the quantitative relationship between some input of advertising and some output of presumed value" (Simon and Arndt, 1980). The outputs can he measured by salesthe change in an individual's propensity to buyor through such intermediate variables as advertising recall, attitudes, or intention to buy. Our focus, in this arti-
198
june 200g
EMPIRICAL GENERALIZATION
The advertising response curve is "convex"the greatest marginal response is from the first exposures. As the number of cunnulative exposures in a period increases, the marginal effect of the advertising drops.
o Convex o Linear
d Purchas ing
' S-Shaped
^ / ^
eft CD
Inert
7
0
1 2 3 4 5 6 7+ Increased OTS
Using TNS's TVSpan single-source panel data and examining all brands in each of four categories as a whole, the ShortTerm Advertising Strength (STAS) results displayed convex-response functions (see Figure 2). Within that generalization, however, there were brand-by-brand exceptions (see Table 1). Do these exceptions imply that there are conditions under which the law does not hold? Clearly there are measure-
ment issues (i.e., sampling errors) that can affect the apparent shape of the response function for a brand. In particular, STAS scores for small brands can be misleading because they represent small changes over a baseline. For example, Bertolli spread had frequency distributions of 687, 66, 44, 32, and 63. The STAS scores varied widely between one and four or more opportunities to see (OTS) (see Table 1) while the frequency distribution shows that, in fact, Bertolli's results were quite static. We may be able to generalize about the shape of the advertising-response function based on a large body of existing research work. But there still are unresolved issues in the accurate measurement of the response function on sales.
CONCLUSION
To be an empirical law, an empirical finding must hold approximately, but consistently, across some known conditions. Decades of single-source analysis on TV commercial sa I es-response functions keep reporting convex sa les-response fLinctions even though the analysis has been done in different ways by different researchers, in different countries, on data collected different ways, on different product categories and brands. The body of research would suggest that the convex sales-response function is law-like. But empirical laws are not merely discovered; there is always ongoing work on the generalizability of different laws. And this article demonstrates that there still is a great deal of research to be done to fully understand the convex sa les-response function. Nonetheless, we propose: EGl: For television advertising for established consumer goods brands in mature markets, when advertising has a positive sales
Figure 2 28-Day Advertising Response Functions, across Product Categories, for All Brands
Butter and Butter Substitutes 28-Day Response Functions of Brands with Positive STAS Scores: Ordered by Market Share
npcppnHint
Ehrenber^-Bass institute at the university of south Australia. The institutes fundamental research s
used and financially supported by many of the world's leading corporations including Coca-Cola, Kraft,
OTS N 1-' '.' 103144 '" 5 585 ]. 3 788 2' 2,744 3 8,069 A.
Kellogg's, British Airways, Procter & Gambie, Nielsen, TNS, Turner Broadcasting. Network Ten, Simplot, Mars, and many others. Dr. Sharp has published over 100 academic papers and is on the editorial board of
98 87 72
^,
five journals.
REFERENCES
126^ .??9
.,,
BROADBENT, S.
19.9
,
effect, the short-term advertsof OTS to correspond with an exposure. __ me response function 15 typi- fmi) ^ , cally, but not always, convex.
. . . . JENNIFEB TAYLOR is a research associate at the
The law as it stands today is important, , ,, , , , practical, useful knowledge. But there are , , , .
EhrenbergBass Institute for Marketing Science. University of South Australia. Her area of research ineludes media planning and advertising. She is currently completing a PhO examining the advertising J K & e. response function and its measurement.
JONES, ]. P. How Mucit is Enough? Getting the ., , ^ ,.. AJ ^- n ii M - V I Most from Your Advertising Dollar. New York: ^ , , ._
exceptions that require further investiga, ^ , tion to look for s u b p a t t e m s . ^ , , , , 1 To better u n d e r s t a n d the b o u n d a r y , ,. , conditions, the role that schedulmg h a s on the measurement of the sales effects of advertising should be investigated. with n e w RACHEL KENNEDY is an associate professor and the head of the Advertising Research Group at the Ehrenberg-Bass institute for Marketing Science, University of South Australia. She is involved in commercial market research, to understand buyer or social behavior, and marketing interventions iike advertising, as wetl as academic research into advertising and the dynamics of market structure.
SIMON, J. L., and j . ARNDT. "The Shape of the Advertising Response Function."/tnrnd/o/^vertising Research 20, 4 (1980): 11-28.
technologies that p u t viewers in control of what they watch and when? Research reexaniining the relationship
between OTS and actual exposures m a y also be needed to clarify whether adjustments need to be m a d e to the n u m b e r
VAKRATSAS, D., and T. AMBLER. "HOW Advertising Works: What Do We Really Know?" ournal of Marketing 63, 1 (1999):