&Quot;Liking&Quot; Through Moment-To-Moment Evaluation; Identifying Key Selling Segments in Advertising

Mark Polsfuss, Viewfacts, Inc.
Mike Hess, Viewfacts, Inc.
[ to cite ]:
Mark Polsfuss and Mike Hess (1991) ,"&Quot;Liking&Quot; Through Moment-To-Moment Evaluation; Identifying Key Selling Segments in Advertising", in NA - Advances in Consumer Research Volume 18, eds. Rebecca H. Holman and Michael R. Solomon, Provo, UT : Association for Consumer Research, Pages: 540-544.

Advances in Consumer Research Volume 18, 1991      Pages 540-544


Mark Polsfuss, Viewfacts, Inc.

Mike Hess, Viewfacts, Inc.

Historically the advertising industry has used two primary approaches to judge television advertising. These have been termed "evaluative" and "diagnostic." The evaluative techniques of recall and persuasion normally employ some sort of established category norm. Depending on the technique, commercials that obtain a recall or persuasion score exceeding the norm usually go to air, while those that fall short are either reworked or abandoned. Over the past several years the industry has largely relied on these two measures as their ultimate decision criteria For this reason "evaluative" measures have also sometimes been referred to as "criterion" measures.

In contrast the "diagnostic" approaches have tried to understand how advertising works. Primarily this has been done through one-on-one interviews or focus groups. The diagnostics go beyond recall and persuasion to examine issues such as the main message of the commercial, what people liked or disliked, whether there was confusion about anything, etc.

In this paper we will discuss how "liking," in the past generally used as a diagnostic variable, is currently in the process of becoming a criterion variable as a result of research reported in recent papers. In addition, we will examine how a comprehensive understanding of liking through moment-to-moment analysis allows us to pinpoint key selling seconds in the commercial that lead to more effective advertising.

Let's first examine what Viewfacts means by moment-to-moment analysis and how to measure it.

During the late '70's, technological applications to the entire market/media research industry began to appear with increasing frequency. One such application was a small hand-held microcomputer developed by PEAC Media Research, Inc. (PMRI) in Toronto, Canada. The unit was designed to track a respondent's moment-to-moment (M-T-M) response to any audio/visual material.

The original motivation for developing the technique was to facilitate understanding of childrens' reactions to programming. In conjunction with the Children's Television Workshop and TV Ontario, PMRI began using their "PEAC System" (Program Evaluation and Analysis Computer) with children. The basic idea was to use the data to enhance a qualitative exploration. This approach worked so well with children that it also was easily adapted for use with adults. Over the years a variety of improvements have been made to the technique, which is now in use worldwide. Here is how it works today.

Respondents continuously enter their reactions on the wireless hand-units by pressing keys (A-E) which correspond to a 5-point scale from very positive to very negative. While they are giving us their anonymous, spontaneous reactions, their responses are instantly aggregated and displayed for both the moderator as well as for the clients in the viewing room. We have the option of displaying the data by total sample, by sub-groups i.e. age, sex, etc., by the number of negative key pressers, or deviations from the group average and so on. The group moderator subsequently utilizes the movement in the line as a discussion tool, probing for the reasoning behind those reactions. What we get as a result, then, is diagnostic information explaining what's working, what's not working, and most importantly - why. Clients have considered this approach a significant improvement over standard focus groups because it gives the moderator structured feedback from the audience itself to use in the ensuing group discussion.

As mentioned, this data was used in a structured way to elicit qualitative responses. The M-T-M line has proven itself to be an excellent tool for examining a commercial's ability to involve viewers, to determine its level of appeal and to find out what the commercial was communicating. Since we frequently have used a positive to negative scale, we at Viewfacts often felt we were getting a good read on what was "likeable" about the commercial as well. How useful this knowledge was in terms of creating "successful or persuasive advertising" has always been somewhat controversial, because likability,--while considered an excellent "diagnostic" variable, has not generally been considered a "decision criterion" variable, in contrast to recall and persuasion.

Alexander Biel, Executive Director of the Center for Research and Development and an authority in the area of "commercial liking," was a pioneer in tackling the likability issue by conducting a large-scale study in 1985 which investigated whether liking a commercial had anything to do with persuading consumers to buy the advertised brand. In a recent discussion of that research he and his co-author Carol Bridgwater (1990), found that "people who liked a commercial a lot were twice as likely to be persuaded by it than people who simply felt neutral towards the advertising." The study went on to define what likeable advertising was and was not. Without getting into that issue in depth, its overall conclusions were that "commercial liking went far beyond mere entertainment. Viewer involvement and perceived relevance are factors that link commercial liking to persuasion in the first place. People like commercials which they feel are relevant and worth remembering." The study also found that "liking was a function of product category..." That is, the way in which liking works in one category, is different than how it works in another, in terms of being persuasive.




In our own continuing effort at Viewfacts to understand the moment-to-moment response data and its relationship to such measures as likability and persuasion, we had always believed that there was more to the measure in terms of its quantifiable aspects. Some of our clients had also felt this way. Intuitively recognizing a relationship between persuasion and the moment-to-moment data? our regular clients have, over the years, built their own norms for acceptable performance of their own advertising. During the last two years Viewfacts has learned how to statistically transform the M-T-M data into measures that are directly correlated to internal measures of persuasion, and even to actual product purchase (Polsfuss and Hess, 1989; Spaeth, Hess and Tang, 1990). We would now like to discuss how this has come about.

In 1989, Viewfacts, along with our client Time, Inc., undertook the challenge to determine if the moment-to-moment data could be mathematically related to commercial success or failure. By obtaining M-T-M data for ten direct response television commercials (Call 1-800...) and a measure of actual marketplace performance for each commercial (based on cost-per order) Time, Inc. and Viewfacts divided the ten cases into two groups: S successful and 5 unsuccessful commercials.

With the M-T-M data on each commercial, an overall liking score and additional standard scalar attribute data, we set out to develop a multivariate model that would hopefully reveal relationships between these variables and commercial success.

One conclusion was that there was a strong inverse correlation between overall liking and cost per order. That is, the more the commercial was liked, the lower the cost per order, in general, and of course, the more likely that the commercial was a success. In the end, the resulting discriminant model was able to predict which of the commercials in each pair would be the superior sales producer 100 percent of the time. As mentioned, this model employed overall liking as one variable, several additional closed-end attribute variables, as well as several very specific moment-to-moment variables. Time, Inc. has agreed that we may share these basic modeling outcomes. We are not at liberty to reveal the exact identity of the variables themselves for proprietary reasons.





As indicated in the above summary chart, it is important to note that when simpler models were tried, the results were good, but did not explain 100% of the commercial winners and losers correctly. For example, a model consisting only of closed-end attributes was able to classify 7 of the 10 commercials correctly, while a model that employed just M-T-M variables got 9 of 10 right.

The primary conclusion of this project, which was reported at the 1989 ARF Copytesting with our client's permission (Polsfuss and Hess, 1989), was that models using both overall and moment-to-moment viewer reactions can be used successfully to predict a commercial's in-market sales performance. Both types of variables were important to the success of the modeling effort. An interesting notion which flowed from these findings was the idea that the moment-to-moment variables enabled us to dissect the overall liking of a commercial into the specific liking for each commercial clement. In fact we found that the overall liking measure correlated with the mean moment-to-moment score with a .96 coefficient. Clearly the moment-to-moment measure of liking and the overall measure of liking, were measuring essentially the same thing. The difference is that the moment-to-moment measure can be broken down into its constituent elements, whereas the overall measure is a simple pass/fail grade in the form of just one number.

This past summer at the annual ARF Copytesting Workshop, Russ Haley, Professor Emeritus of Marketing, University of New Hampshire delivered the results of the ARF Copy Research Validity Study (Haley, 1990). This study, some eight years in the making, was based on comparing 5 pairs of packaged goods television commercials that were run in BehaviorScan split cable tests. The only variable that differentiated each member of a commercial pair from the other was a difference in advertising copy. All other variables, ad weight, in-store promotion, etc. were held constant.

The primary conclusion that one can draw from the findings, is that copy tests do work. They can be predictive of a commercial's in-market sales effectiveness. We've had pre-test measures of copy for a long time, but the industry has rarely had such a solid demonstration of the relationship between those copy pre-test measures and the eventual sales effectiveness of the advertising. A second conclusion is that with few exceptions, all of the various types of copy testing measures tried, worked in one form or another. This means that recall, persuasion, commercial reaction, etc. all were effective measures at differentiating winning from losing commercials.

The new ground here is that of all the measures tested, the one that ranked first in terms of predicting sales effectiveness was "commercial reaction," consumer reaction to the commercials. But what exactly was this reaction measurement? It was a simple five point liking scale as configured below.

The simplicity of this measure is its great power. We're all familiar with the difficulty of getting a reasonably valid measure of recall, and the even greater difficulty of getting a useful and valid measurement of persuasion. Measuring liking, however, is more direct.

Based on this "commercial reaction" finding, Haley concluded, "Commercials that are liked, sell better than those that are not liked." Couple this with Biel's conclusion that "people who liked a commercial 'a lot' were twice as likely to be persuaded by it..." and we have some solid motivation for now proceeding to determine what elements contribute to making a commercial likeable. Is it the overall impact, or are there segments that may be doing most of the work? We at Viewfacts were motivated to address this critical issue for three reasons, as shown below.






Viewfacts' own 1989 research for Time, Inc. linking "liking" to the M-T-M line and the M-T-M line. in turn. to sales.

The recent studies reported by Haley and by Biel and Bridgwater clearly indicated liking was predictive of commercial effectiveness .

The development of the Viewfacts M-T-M database, which would allow us to dissect the time periods in a commercial for over 1,000 commercials.

In an extensive analysis of that database during the current year, we looked at a broad range of product categories, as shown below.



Soft Drinks


In each of these categories we found, as we had previously learned in the Time, Inc. study, a high correlation between the M-T-M data and a separate measure of liking, or commercial reaction. In the computer category, for example we found that 86% of the variance in commercial reaction for 9 IBM commercials could be explained by a model consisting primarily of variables created from the M-T-M line. What variables were these?

The most important variables were those segments of the M-T-M line that were generated during the specific few seconds of the commercial in which the "key selling points" were presented. In fact, the specific correlation with overall liking of this M-T-M response to just a few seconds of the commercial, when the product attributes were presented, ranged from .73 to .89 across the 9 commercials tested.

When we turned our attention to the soft drink and automobile categories, we found a similar pattern, as shown in the chart below. For both 7-Up and Volkswagen, a model consisting only of MT-M responses when product attributes were shown, was almost as good at predicting liking than a more complete M-T-M model containing additional time periods.

By now, you may be wondering exactly what the contents of those key selling segments are and also, why they are so effective at predicting liking. Well, for proprietary reasons I can't answer the first question because to answer it would be to tell you what that firm's key selling variables are. The second question, I'll try to answer with the help of another diagram. The diagram indicates a conceptual scheme that might be helpful at understanding the relationships that may be at work here. It has been compiled by synthesizing the findings from the Biel, Haley and Viewfacts research studies referred to throughout the course of this paper. The analysis goes like this:

Commercials that are liked, lead viewers to believe the brand attributes that are presented in the commercial.

When these same attributes are compelling and persuasive, they provide good reasons to buy the product.

If we also assume that marketing management and the ad agency have done their jobs prior to development of the commercial, then the proper product attributes for the category and brand have already been identified.

Therefore, if the commercial is presented in an essentially believable fashion and it "penetrates your consciousness," then those same attributes will help sell the product.

The role of the commercial, in this sense, is to credibly and pleasantly present persuasive brand attributes.

In turn the viewer response to that portion of the commercial is critical to subsequent sales effectiveness.

Put another way, if a commercial has done its job well, pleasantly and effectively, then it is liked. After that, it is up to those few key seconds during which the commercial delivers its message, to "make the sale." So much for theory and analysis; where is Viewfacts now with respect to practical application of these findings? We are currently giving our clients the following summary of our thinking on this topic.

1. The ARF and Viewfacts have now both reported findings that indicate commercial liking is linked to sales.

2. Viewfacts has found that its M-T-M PEACline and liking are highly correlated. Therefore, the M-T-M line can be linked to sales as well.

3. Although an understanding Of the entire M-T-M line taken as a whole is necessary to adequately model liking mathematically. . .

4. Within the M-T-M line there exist certain "key selling segments" that appear to be more predictive of commercial liking than are other segments Of the line.

5. These "hot spots" of selling power tend to be both brand-benefit oriented and positional (beginning/end of commercial).

6. Key selling segments, as well as the meaning of likability appear to wary from category to category. Therefore, it will be necessary to-create separate models for each advertised category.

7. These models will be similar to each other in that each will depend on the shape of the M-T-M line; critical positional effects; creative elements; and brand benefits. But they will differ from each other with respect to the exact location of positional effects and to the nature of relevant brand benefits.

Overall liking of commercials, we have found, is driven by consumer's moment-to-moment response to key commercial segments. We are continuing our efforts to build an expert system that identifies those key positive selling segments for our clients. The marketplace is dynamic, and-what is "hot or not" is always changing. By understanding advertising on a moment-to-moment basis as we have outlined in this paper, one can optimize the selling power of a commercial. The blending of this scientific approach to understanding the art of advertising is useful when trying to stay with or ahead of an ever-changing world.


Biel, Alexander L., and Bridgwater, Carol A. (1990). Attributes of Likable Television Commercials. Journal of Advertising Research, June/July, 38-44.

Haley, Russell I. (1990). The ARF Copy Research Validity Project. Transcript Proceedings, Seventh Annual ARF Copy Research Workshop, New York, July 11-12.

Polsfuss, Mark and Hess, Michael (1989). The Relationship Between Second-To-Second Response and Direct Response: What Is The Link? Transcript Proceedings, Sixth Annual ARF Copy Research Workshop, New York, May 22-23.

Spaeth, Jim; Hess, Michael and Tang, Sidney (19903. The Anatomy of Liking. Transcript Proceedings, Seventh Annual ARF Copy Research Workshop, New York, July 11-12.