Media Guru (Michael J. Napples)

'Bu

rst' ve

 

Effective frequency:

then and now

 

by Michael J. Naples

In 1979 I authored the Association of National Advertisers (ANA) book which I titled Effective Frequency: The Relationship Between Frequency and Advertising Effectiveness in order to bring the industry up-to-date with regard to what research was telling us about television advertising exposure frequency. Specifically, what we were all interested in was advertising response and the effects of frequency, and in zeroing in (for more efficient and effective media planning purposes) on the effects of one or more advertising messages on a target customer for a brand in terms of realized sales potential.

This was an important subject in 1979 because of the rapid escalation in media costs in the mid to late 1970s - most especially in television - and the perception that those brands with smaller market shares were losing their ability to maintain advertising momentum.

Although Effective Frequency emerged as a strong and widely embraced media planning concept worldwide once the 1979 book was written, until its publication the industry was struggling to bring the concept into clear focus. To complete the book I reviewed over 500 public and private sources, more than 50 of which are referenced in the text. The proprietary studies were obtained by way of an appeal from the ANA Research Policy Committee of which I was a member. (Without those proprietary studies there would have been no book - a lesson there.)

Nevertheless, there were surprising few well designed studies that were germane, all of which I summarized in the book. Moreover, all of these involved television and frequently purchased packaged goods. Upon completion of the task I stated that we were really only beginning to scratch the surface of what would be possible once the advertising and research communities had embraced the subject in a serious and productive way.

Here are my very words at the time of completion of the book:

I am certain that when Advertising Media and Research people have carefully studied the contents of this book and the case studies - when they have gotten familiar with what we already know - then I am convinced they will lead the way to much more and relevant inferences and conclusions regarding advertising frequency, and that further useful long-term marketplace experimentation will take place in the 1980's. This, in turn, will help to more firmly establish the relationship between frequency and the advertising effectiveness, and will lead to our view of the media planning and advertising process as one that is as much science as it is art.

My approach to the book was to make the information both tightly focused and at the same time understandable to media planners. This uncompromising and pragmatic orientation helped the concept to emerge clearly as, "the avoidance of both too little brand advertising to be effective and too much to be inefficient." Nothing has changed that would lead me to modify this view today. After all, not many advertisers can adopt this view of the world's largest.

Procter & Gamble, in a given year, has to sell 400 million boxes of Tide - and to do that, we have to reach our customers over and over throughout the year. Frequency of the depth of sale in advertising are critical to preserving loyalty to frequently purchased brands like ours. For example, in any given month, P&G brands like Tide and Crest and Pantene will reach more than 90% of their target audience six or seven times (Edwin Artz, Former Chairman, Procter & Gamble).

Moreover, the view persists today, as illustrated in this viewpoint forum quotation from Jim Surmanek (1995) of the International Communications Group.

For nearly all products there is a magic range of advertising pressure below which the sales effect is minimal, and above which advertising simply does not pay out. The same appears with reach and frequency.

Also inherent in the Effective Frequency Concept is a time frame for target audience delivery. This means scheduling the right amount of GRPs (weight) within a purchase cycle for a brand to produce a short-term effect in the marketplace.

Here is the way I organized and expressed that short-term purchase cycle to bring focus to what I ultimately called Effective Frequency. In fact, the title did not emerge clearly in my mind until I was well into the final conclusions section. In hindsight, I believe it was an appropriate, and as it turned out, compelling choice.

The 1979 book covered:

* Psychological learning theory

* Small-scale laboratory advertising exposure experiments

* Practitioners' theories

* Four frequency-specific marketplace studies conducted to zero in on television exposure levels (three were proprietary)

* Other marketplace studies of various media strategies

Beyond insights from early learning theorists, a few small-scale laboratory experiments and practitioners' theories, there were only a few studies (four, in fact) that I concluded were specifically well designed enough to be directly relevant to the issue of television advertising exposure frequency.

As an overall observation, everything was inferential regarding specific exposure effects except for the following four marketplace studies I came to rely on to form conclusions about the Effective Frequency Concept.

The four studies were:

1965: Ogilvy & Mather Research Department. "An Experimental Study of the Relative Effectiveness of Three Television Day Parts"

Criterion measure: Brand preference change

1966 and 1971: Colin McDonald. "What Is the Short Term Effect of Advertising?"

Criterion measure: Brand switching

1974: Major advertiser Adtel scheduling study

Criteria measure: Purchase probability

1975: Major advertiser study (38 brands)

Criterion measure: Unaided brand awareness change

The following are my conclusions as set forth in 1979. These four studies remain, even today, as the four best reference sources we have. This will remain true until we can properly unlock the single-source data resources that currently exist or are now emerging internationally.

CONCLUSION 1: One exposure of an advertisement to a target group consumer within a purchase cycle has little or no effect in all but a minority of circumstances.

CONCLUSION 2: Since one exposure is usually ineffective, the central goal of productive media planning should be to place emphasis on enhancing frequency rather than reach.

CONCLUSION 3: The weight of evidence suggests strongly that an exposure frequency of two within a purchase cycle is probably an effective level.

CONCLUSION 4: By and large, optimal exposure frequency appears to be at least three exposures within a purchase cycle.

CONCLUSION 5: Beyond three exposures within a brand purchase cycle, or over a period of four or even eight weeks (as in the Ogilvy & Mather study), increasing frequency continues to build advertising effectiveness at a decreasing rate, but with no evidence of a decline.

CONCLUSION 7: Very large and well-known brands - and/or those with dominant market shares in their categories and dominate shares of category advertising weight - appear to differ markedly in response to frequency of exposure from smaller or more average brands.

CONCLUSION 11: Although there are general principles with respect to frequency of exposure and its relationship to advertising effectiveness, differential effects by brand are equally important. There were other conclusions as well, but I believe these are most relevant for my discussion here.

Now let's fast forward to November 1994 when the Advertising Research Foundation held an Effective Frequency Research Day in New York City. Fifteen papers were delivered, including an update by Colin McDonald on his progress toward updating my ANA 1979 book Effective Frequency: The Relationship Between Frequency and Advertising Effectiveness and John Philip Jones with a review of his forthcoming 1995 book, When Ads Work: New Proof that Advertising Triggers Sales. In addition, a media planners discussants panel at the end of the day gave their views on effective frequency and what they had heard discussed during the course of the day.

After summarizing the research evidence behind what I called the concept of effective frequency in 1979, from that point on the concept gained substantial acceptance among media planners while at the same time generating considerable debate both among practitioners and academicians on how supportable it was based on the evidence set forth.

Over the years, the effective frequency concept has been used and interpreted in many ways, and to many achieved a 3+ "rule of thumb" status. In other words, in practice it was both popular and oversimplified, and naturally as it gained influence it also attracted debate. Nevertheless, effective frequency has managed to withstand all attempts at revision simply because new high quality research evidence has been difficult to achieve.

The problem has been that this type of empirical market study has been the exception rather than the rule. Plus, whatever tests were conducted have remained proprietary - so there is no recent marketplace experience body of knowledge of any worth to guide media planners. Obviously such studies would have been useful to test effective frequency media strategies (after the fact, i.e., to test out the best effective frequency concept applications for a brand). But this did not happen.

Moreover, relevant new research input has been slow in coming because the matching of exposure information and buying patterns from the same individuals or households has proven to be so difficult to obtain and prohibitively expensive, even from the scanning research laboratories which emerged as successors to Adtel, and from other panels as well.

The major problem has been to pin down consumers' opportunities to see (OTS) advertising messages because of the necessity of having a record of all advertising messages scheduled (sent) that can be matched with consumer media consumption (received). The process is extremely expensive to implement on a scale large enough to allow analysis of individual brands in many categories over a period of time.

Information Resources Incorporated (IRI) managed to collect such data for all their category brands for a time in the 1980s, and in the past few years, a Nielsen household panel service has also more or less put all of the single-source pieces together by using a combination of exposure meters in the home - Monitor Plus to electronically capture all schedules as sent, and UPC scanned buying records for all store purchases brought into the home. Recently, however, the expense of this effort has also led to its discontinuance.

Also, since 1979 the BehaviorScan marketplace laboratories have been in operation in the United States, but for all of those years, no proprietary studies of media strategy effectiveness - let alone effective frequency - have been shared with the industry.

Thus, while useful data has been collected, the difficulty and expense of obtaining and analyzing single-source data needed to unlock media strategy information has not resulted in any new learning to modify the frequency conclusions I first put forward in 1979.

This was reflected in what our panel of media planners concluded at the finale of the 1994 ARF Effective Frequency Research Day. Their consensus was:

* Although there is no "rosetta stone" to define exactly what a general principle of effective frequency means, it is important that the industry keep the subject alive by discussion, debate, and new research.

* To planners, effective frequency is not an end in and of itself; rather it is a tool, one of many used by advertisers and their agencies to help define the most intelligent way to spend ad dollars for different brands.

* They cautioned against use of a magic number in using effective frequency and pointed out that many media planners had ignored conclusion #11 in the 1979 book which said that each brand should experiment to find the right answer for its own frequency equation. They also voiced concern that from the emerging new research that a new number would be picked and used incorrectly as a general principle.

* Finally, they noted that all such analyses are only as good as the data on which they are based.

As it happens, the 1994 ARF Effective Frequency Research Day was the event which signaled the reawakening of interest and investigation into the issues of advertising reach and frequency. In their presentations, both Colin McDonald and John Philip Jones argued for one exposure effectiveness based on evidence contained in their then forthcoming books, and their challenge to established frequency thinking has set off a wave of renewed interest and investigation.

My view of their work is that together they provide a useful up-to-date picture of what we have and what we don't have by way of research evidence either in support of or to modify the concept of effective frequency. As you will see in the course of this paper, I have a few differences with both McDonald's interpretations and Jones's data analysis, but there is no denying that the substantial efforts of both have had the positive effect of also springboarding new interest and investigations into the subject by others after so long a period of dormancy.

You will see much of this new work contained in this special issue of the Journal of Advertising Research, along with the latest views of both McDonald and Jones. It also contains the commentaries of Simon Broadbent (Leo Burnett), Gerard Tellis (USC), and Erwin Ephron (Ephron, Papazian & Ephron Inc.) - and the analyses of Walter Reichel and Leslie Wood (A-to-S LINK), Ken Longman (Longman-Moran Analytics Inc.), Mike von Gonten and Jim Donius (Michael von Gonten, Inc. and Marketplace Measurement Worldwide) and Gary Schroeder, Bruce Richardson, and Avu Sankaralingam (IRI). All provide fascinating points of view on the evidence for and against our ability to concretely settle on an approach to advertising frequency. Taken together, we now have a more positive basis for solid progress on the subject than at any time since the completion of my book in 1979. In other words, we have finally begun to add to the body of knowledge once again.

Now, having set the stage, I believe it is important to set forth my own views on this new work that you will be encountering as you study this special compendium. Let me begin with McDonald's second edition. Since my position as Advertising Research Foundation President left me with little time for such a task, I turned to Colin McDonald whose earlier work had been so essential to the Effective Frequency concept. His second edition has brought a good deal more clarity to the discussion from the benefit of 17 additional years of observation. In addition, McDonald has added a section on print frequency, since there have been new studies to provide insight. He has been able to reveal a new conclusion with regard to "recency of exposure" which amounts to an important new media planning principle.

Those of you who take the time to study the two texts will see that we agree on most conclusions, but there are some obvious differences which I suspect can be a natural consequence where two different analysts are involved. My sense is that McDonald gave more emphasis to the academic literature than I, plus he was influenced by more recent work, particularly that of John Philip Jones which also led him to reanalyze his classic 1966 study which had appeared in the 1979 book edition.

Our key difference, and it is an important one, is in regard to one exposure. McDonald differs from my 1979 conclusion by explaining that one exposure can be effective in a majority of circumstances and not a minority of circumstances as I stated in 1979.

McDonald's conclusions are:

CONCLUSION 1: There is convincing evidence that advertising, when it is good enough to work at all, may have short-term effects, including effects of purchase probability.

CONCLUSION 5: . . . The major effect will often be at one exposure/OTS, especially if we are looking at large and familiar brands . . .

The bottom line is that McDonald now believes that one exposure opportunity to see (OTS) within a purchase cycle can potentially have a beneficial sales effect for most typical brands.

One of the reasons for this changed view is John Philip Jones's single-source data analysis put forth in his 1995 book. Jones tells us that his basic building-block measure is termed STAS (Short-Term Advertising Strength) and is derived as shown in Figure 1. He states that, "STAS is based on market share or more precisely a brand's share of all purchase occasions." Jones develops his STAS analyses from single-source data derived from 12 product categories of the Nielsen Household Panel of 2,000 houses for all of 1991. Jones describes the STAS differential, the main indexed measure he uses for analysis of all brands as follows:

. . . for brand AL (a brand of Packaged Detergents), the Stimulated STAS is 60 percent above the Baseline STAS. (3.0 = 100; 4.8 = 160). This strong STAS Differential is evidence of the immediate sales effectiveness of AL's advertising (Jones, 1995).

McDonald's 1996 description of the data and the analytic method are thus:

The Nielsen Household Panel provides true single-source data, the purchase data stream from a handheld bar-code scanner, and the television data stream from people meters. Identification of commercials comes from the Monitor Plus system, which is based on a chain of monitoring stations that keep a running log of the advertising appearing at 15-second intervals in the 23 largest Areas of Dominant Influence (ADIs) in the United States. In each region, the information is collected from eight stations: the four broadcast networks, plus the leading cable station watched in the region. On average, 2.5 million commercials are logged every month.

Jones has developed a measure called STAS, short for Short-Term Advertising Strength. STAS is simply the brand's share of all purchasing occasions. This is compared between those cases when no television advertising for the brand had been received by the household during the previous seven days (the BASELINE STAS) and cases where advertising had been received (STIMULATED STAS - naturally different amounts of advertising can also be compared) (McDonald, 1996).

Analyzing the brands in 12 categories over a year's period is clearly an ambitious undertaking and Jones is to be congratulated for striving to put the power of single-source data under the microscope for advertising effects knowledge building. His intent is to do no less for advertising than what scanner data did for promotions in the 1980s - thus the name Short-Term Advertising Strength.

To my way of thinking, Jones's methodology of examining advertising exposures only during the seven days before purchase and not change over the entire purchase interval renders his findings and observations - at least insofar as specific levels of advertising exposure effectiveness are concerned - more inferential than definitive. That is because those who were exposed to advertising probably:

* purchased more of the advertised brand (were more loyal)

* were exposed to the brand's advertising more frequently than the nonexposed group (perhaps due to schedule on which brand advertising appears)

* were more receptive to ad messages because of reinforcement of their brand choice

These in-going differences or predispositions are not taken into account by STAS as currently constituted.

Across the 78 brands and the 12 categories analyzed, Jones shows the average STAS for the O, one exposure, and Total exposure levels in the last seven days (see Table 1).

Jones states that "More than 70% of the sales effect comes from the first exposure in a flight of advertisements." In fact, if we look at the true picture of the way he analyzed his data, i.e., looking back for exposures that occurred up to seven days before purchase, what he is really looking at is the last exposure in a flight, i.e., the last exposure before purchase.

We have no way of knowing if the one exposure group (or the O exposure group) received other exposures prior to the seven days but still within the average brand's purchase interval, but that is likely. From my experience, the purchase intervals of categories I am familiar with - such as detergents and bar soaps - have purchase intervals more like four to six weeks.

In other words, I believe that because of the limitations cited it's difficult to conclude or even infer enough about exposure effects to come to a finding that one exposure alone within the purchase interval or even within the last seven days is sufficient.

Having said this, however, I fully acknowledge that Jones's STAS measure has succeeded in opening the industry's eyes to the analysis potential of single-source data for far better advertising exposure to sales understanding, and I refer you to the Schroeder, Richardson, and Sankaralingam paper in this issue for a well-rounded review in this regard. Their rigorous analytical treatment, if pursued, will help us to definitively pin down number of exposure effects short of conducting a great many single variable marketplace experiments, a situation that has little likelihood of happening.

In conclusion, in addition to studying this entire special issue of the JAR, I refer you back to the full text of Colin McDonald's second edition as the most comprehensive review of our effective frequency knowledge to date, and the most sensible exposition of both what we can now conclude as well as the steps that must be taken to resolve those issues that remain. With each passing year, the subject has continued to grow in importance, and the present renewed interest of serious researchers should be most welcome to all.