Executive Summary

Recently Unilever triple tested 60 television commercials with three major copytesting systems and compared the results. One of the things they found out is that while all three, Ameritest®, Ipsos-ASI and Millward Brown, claim to measure “brand linkage”, each system is measuring something quite different from the other two.

What is the role of the brand in the little movie that is a television commercial? Is it the star of the movie driving the plot forward and occupying the center spotlight of consumer thoughts whether on screen or off? Is it a supporting actor carrying a scene or two in the service of the action and providing essential information to move the story along? Or perhaps it’s more like a director or producer, only appearing in the credits at the beginning or end of the film?

If one of the roles of advertising is to make ordinary products and services famous, then these casting questions are of paramount importance. From a research standpoint, this relates to the assessment of how well branded a television commercial is. Our intuitions about well-brandedness are similar to Supreme Court Justices’ views on obscenity: we know what it is when we see it. Why else, for example, do most adult US viewers have vivid memories of advertising for Aflac (in which the brand name becomes the creative idea), Snuggle (in which a character becomes a branding device), and the VW Beetle (in which the brand is well integrated with the story)? Are all of these well-branded commercials? How do we know?

Many commercial pre-testing systems report metrics with such labels as “branding” or “brand linkage.” Not surprisingly, each system defines this “branding” construct using its own proprietary approach. But we must be careful not to assume that just because the same “surface” label is being used, that the “deep” constructs being measured are equivalent. As you will see, the branding measures provided by three major pre-testing systems are each measuring something completely different.

Recently, Unilever, which produces hundreds of commercials a year in support of over a thousand brands, set us the task of some serious research-on-research to study the measures that were being produced by the three different research pre-testing suppliers that they were using: Ameritest®, Ipsos-ASI and Millward-Brown. In order to stimulate debate and advances in learning on advertising effectiveness, Unilever has decided to share some of these findings with the industry. (Unfortunately, the question of validation or what works best in-market is considered proprietary and will not be shared.)

The database we had to work with was unprecedented: 60 television commercials that were tested in all three systems—a multi-million dollar investment. The sample of commercials were all taken from the home and personal care product division, but represented a considerable range of creative work from over a half dozen ad agencies. Half the ads tested were for new products, and half were for established brands. Half were final film, and half were animatic rough productions. The conclusions we drew from each of these types of ads were similar.


In an initial project that we reported to the Advertising Research Foundation last year, we studied the variety of breakthrough or intrusiveness measures produced by the three systems. Our key conclusion was that Day After Recall lacks the ability to measure emotional advertising. This is not to say that recall is a bad or invalid measure, but only that it systematically rewards a certain kind of advertising. Ads that recall well tend to be linear and rational, boring and predictable. Not the kind of ads your average creative director or brand manager yearns to produce.

The current program of research extends this previous work by examining and offering insights into another key “report card” measure that is used to evaluate creative work—“brand linkage” or “branding”.

In this article we turn our attention to the second report card measure that, in combination with the first, is commonly used to produce a composite measure of advertising effectiveness: branded attention. These quantitative results were also presented to a second ARF conference on copytesting last year, and are based on the same database of 60 triple-tested commercials. But first, let’s look at the different philosophical approaches that are taken to measure branding.


Of the three pre-testing firms, the Ipsos-ASI measure of Brand Linkage is the only measure derived from two primary measures rather than measured directly. It is of particular theoretical interest because it tries to combine the two fundamentally different approaches that we have for tapping into memory—recall and recognition.

On a conceptual level, the Ipsos-ASI measure is quite attractive. It attempts to quantify the strength of the connection between two different memory systems where advertising content might be stored. Recall uses category and brand cues to retrieve memories from the semantic memory system. Recognition, by contrast, uses verbal descriptions of the ad to elicit the memory of the advertising from the episodic memory system which is then reported as an attention score. In metaphorical terms, therefore, we might think of this measurement construct as the bridge between two memory systems.

Millward Brown

Millward Brown’s measure of Branding is of major interest since it was specifically designed to help predict incremental gains in advertising awareness measured by their continuous tracking studies. Specifically, Millward Brown uses the mean of a five-point verbal scale ranging from “You couldn’t help but remember this commercial is for (brand)” all the way to “It could be a commercial for almost anything.”

The idea here is that advertising should avoid being generic. A poorly branded ad is one where both structure and/or style fail to direct the viewer beyond broad category cues; there is no unmistakable signpost for the viewer to identify the specific brand being advertised. In contrast, for a well-branded commercial, it is difficult for consumers to talk about the ad without also talking about the brand. This measure of branding, therefore, is characterized by strong product/narrative integration.

In metaphorical terms, we might think of this measurement construct as tailor-made. Advertising is like a suit that a brand wears; the well-dressed brand should wear a custom-tailored, “bespoke” suit rather than buying one ready-made right off the rack.



Ameritest’s measurement of Brand Linkage uses a method that is quite different from the first two systems. This measure of branding is similar to the unaided or top-of-mind brand awareness measure generated by a traditional advertising tracking study. At the beginning of the test interview, respondents are shown a clutter reel of five commercials—a test ad preceded and followed by two control ads—and are then asked the question “Which of these ads did you find interesting?” If they mention the test ad as one of the interesting ads, the open-ended response to this question is coded for whether or not the respondent used the brand name. Brand Linkage, then, is the ratio of respondents spontaneously referencing the brand to the total number of respondents who found the ad interesting (not the total test sample).


Since the respondent only has to retrieve the brand name from short-term memory, this measure is not really a memory test but a measure of brand salience. Consequently, an appropriate metaphor for understanding this branding construct is that of a handle which the consumer uses to hold on to the experience created by well-branded advertising.

Relationships Among the Branding Measures

In analyzing our database, the first thing we looked at was the correlations between the three measures of branding. Even though each system is approaching the problem differently, perhaps these measures are all tapping into the same thing in the mind?

As you can see in Table 1, the Ipsos-ASI measure of Brand Linkage is uncorrelated with the Millward Brown measure of Branding and has only a small positive correlation with the Ameritest® measure of Branding. The Millward Brown and Ameritest® measures of Branding are also uncorrelated. In short, each of these three measures of branding appear to represent three unrelated views of how well-branded an ad is—though, like the blind men describing the elephant, each of them may be a valid measure of commercial performance in the marketplace.

Branding and Breakthrough Measures of Attention and Recall

Next we examined the correlations between these three measures of branding and measures of breakthrough—recall and attention—produced by the three systems. As you can see in table 2, the Ipsos-ASI branding measure is highly correlated with recall and is only nominally correlated with the Ipsos-ASI measure of attention. The Ipsos-ASI branding measure is also uncorrelated with the Ameritest® measure of attention but negatively correlated with the Millward Brown measure of attention at a significant level.


Indeed, it appears that the Ipsos-ASI measure of branding is actually a recall score in disguise. One possible interpretation for why this is so is that the measure of attention used in this test is a verbal measure of advertising recognition—which is redundant with recall because it too taps into the verbal semantic memory system which may unfairly penalize visually interesting television commercials that are hard to describe in words. If this is true, then the apparent flaw in the Ipsos-ASI measure of branding may be in its operational definition rather than in its underlying construct.

Branding Measures and Commercial Diagnostics

The third set of relationships we investigated was the relationship between the three branding measures and diagnostic commercial ratings, shown in Table 3. The Ipsos-ASI measure is negatively correlated with commercial liking, entertainment value, executional uniqueness and the likeability of music used in these ads but positively correlated with the rating “boring.” This is not surprising given the measure’s high correlation with recall, which, as we reported earlier is linked to boring and predictable advertising.

The danger, evidently, in using this measure is that it appears to reinforce a belief held by some that there must be an inherent trade-off between creativity in advertising and branding—undoubtedly, a major source of frustration for many creative directors.

In sharp contrast, the Millward Brown measure suggests the opposite. There are significant positive correlations between their measure of branding and commercial liking, entertainment value and executional uniqueness, but significant negative correlations with boring and ordinary.

Ameritest® Brand Linkage is the Switzerland of branding measures, remaining neutral between the entertainment values of the execution and the persuasiveness of the selling proposition. The Ameritest® measure is neither positively nor negatively biased toward execution or strategy, that is, there are no significant correlations between it and any of these creative diagnostic ratings or purchase intent (not shown.)


Three Explanatory Models of Branding

To gain further insight into these measures of branding, we built three regression models to understand the conceptual basis of commercial well-brandedness using such executional variables as the timing and occurrence of audio category and brand cues, total audio brand name mentions, as well as visual diagnostics produced by the Ameritest Flow of Attention®. All three models showed a good fit with actual commercial performance on the branding measures, with each explaining over half of the variance in actual scores, as shown in Exhibit 1.

Five variables did the best job of explaining the Ipsos-ASI branding score. Liking enters the model with a negative beta, which is consistent with the authors’ earlier paper showing a negative relationship between liking and recall. The Ameritest® brand linkage measure also enters the model, which is not surprising given the small positive correlation reported above. Two audio variables also carry significant weight in the model. The first is the presence of an early audio category cue and the second is the total number of brand name mentions in the commercial. Both of these variables have frequently been commented on by Ipsos-ASI as drivers of related recall. Finally, a visual variable representing the rational information or product-related content in the commercial adds to the predictive power of this brand model. On the whole, therefore, this model further confirms that the Ipsos-ASI measure of branding is primarily a measure of related recall which rewards commercials that are constructed like sales presentations to the consumer: a linear ordering of product-related information.


Five variables do a good job of explaining the Millward Brown measure of branding. The first is five-point purchase intent while the second is the entertainment value of an execution. Thus, we see that the Millward Brown measure actually performs some of the “bridging” function claimed for the Ipsos-ASI measure. Next, we find two audio variables entering the model: the presence of an early brand cue and the total number of audio brand name mentions. This suggests that the presence of early audio brand cues and brand name repetition are not in themselves the reason why the Ipsos-ASI measure appears to force a trade-off between the attention-getting power of an ad and branding. Finally, the presence of an early brand visual in an attention peak also leads to higher Millward Brown branding scores.

The Ameritest® measure of branding can be explained in large part by four variables. The first is brand fit, which is the same question used in the Millward Brown branding measure, but weighted somewhat differently. The next two variables pertain to the level of viewer involvement in processing the visuals in the commercial: the overall average level of attention to the visuals and presence of a strong opening or visual hook to the ad. The fourth variable is the presence of a “single-minded branding moment” in an attention peak. A “single-minded branding moment” is an image in the commercial in which the identity of the brand is the focus of audience attention and does not have to compete for viewer attention with other executional elements within the frame.

Importantly, this last variable makes no reference to where in the overall flow of the ad—i.e., beginning, middle, or end—that the peak might be located. Of the three branding measures, therefore, this is the one most likely to respond to commercials that withhold the brand identity until the end of the ad.

Focus and Fit: The Key to Branding?

The multiple regression results reported above suggest that two of the three models of branding (i.e., those for Ipsos-ASI and Millward Brown), appear to be weighted somewhat more heavily toward the audio or verbal components of commercial executions while the Ameritest® model is driven more by visuals. This suggests that two of these measures may in some sense be complementary, balancing out both verbal and visual ways of branding television commercials.

On a higher conceptual level, we also see that the three models also have several components in common. Each contains a variable describing the opening of the commercial, whether it be the presence of category cues, brand name mentions, or the overall level of attention-getting power of the opening few seconds of the ad. In addition, each contains a variable which is determined from a peak moment in the attention flow.

Taken together, these two ideas might be thought of in terms of the ability of the execution to focus attention on the brand. The opening of the commercial is important insofar as it establishes the appropriate frame of reference for the audience to correctly interpret the advertising experience to follow. Peak visual moments are important because they tell us where audience attention is centered during various parts of the film. Both of these variables tell the advertising manager whether or not the brand is clearly the star in the “little movie” the consumer is processing.

Each of the three brand models also contains a variable which is related to the strategic content of the ad. The product information contained in the product-related peaks in the Ipsos-ASI model is where the much of the rational component of the strategic message content is conveyed. Similarly, the purchase intent variable in the Millward Brown model also highlights the strategic connection since consumer purchase intent is driven by selling propositions that are relevant to the consumer. Finally, we see that brand fit anchors the executional variables of the Ameritest® branding model. All three of these branding models, therefore, seem to be telling us that branding is not just a surface characteristic of effective advertising, but that the execution must also fit the brand at a deeper level in terms of what the brand stands for in the mind of the consumer.

Focus is a function of execution and fit is a function of strategy. With these two new metaphorical constructs, we can now move forward another step toward understanding the processes through which advertising executions and brand strategies come together to create a star.


Kastenholtz, J., & Young, C. E. Why day-after recall misses the emotion in advertising that builds brands. Paper presented at the 49th Annual Conference of the Advertising Research Foundation, April 2003.

Young, C. E. Brain waves, Picture Sorts® , and Branding Moments. Journal of Advertising Research, July/August 2002, 42 (4), 42-55.

P Senge: The fifth discipline: the art and practice of the learning organization. New York: Doubleday, 1990.

Young, C. E. and M. Robinson. “Video Rhythms and Recall,” Journal of Advertising Research, 29, 3 (June/July l989): 22–25.

About the Authors:

Charles Young is CEO of Ameritest.

John Kastenholz is VP Consumer Insights, Unilever North America Home & Personal Care.

Graham Kerr is VP Millward Brown in the Fairfield Conn. Office.