You are at a party. You see someone smiling and coming toward you, and you know that you’ve met that person before — you clearly recognize the face — but you just can’t recall the person’s name! It’s a simple test that reminds us that the human mind stores memories in more than one way. For advertising researchers, it triggers the debate about the relative merits of recognition vs. recall as a measure of advertising awareness.
The logic of much of the research that is conducted on advertising is that for advertising to be effective, it must leave some kind of trace in the memory of the consumer. But what kind of trace? Think about our hypothetical party for a moment. Which is proof that you have or have not met the person before: the fact that you recognize the face or the fact that you can’t recall their name?
It doesn’t seem surprising that in the case of in-market tracking, the measure of awareness of a specific advertising campaign using recognition can be substantially different from one based on recall. It is sometimes higher by as much as a fact or of two! Which measure should you believe?
Like most people, I trust my recognition memory system more than I trust my ability to recall things such as names. While some people might have minds as orderly and indexed as a library card catalog, my own mind is better described by a French word gallimaufry — a hodgepodge or jumble. It’s like the junk drawer in the kitchen of the mind. Retrieving a memory from my mental gallimaufry requires careful searching.
This helps explain why in response to recall questions, consumers will sometimes playback images from ads that haven’t aired in 20 years and yet completely fail to mention a campaign that our client just spent $20 million airing. Or they will sometimes describe a competitor’s ad as one of our client’s because they filed it away in memory under the wrong index card. Further complications arise when a large pool of commercials is airing. It is extremely difficult if not impossible using recall to tell exactly how much or how little of a campaign the consumer has already seen and how exactly that depth of exposure impacts sales. Recall is an all-or-nothing kind of exposure.
The problem is compounded when it comes to “experiential” or “emotional” advertising. There is substantial evidence (e.g., Zielski, 1982) that day-after recall — which relies on verbal descriptions of the advertising — unfairly penalizes emotional advertising. Emotional advertising tends to be more difficult than rational advertising to describe in words. This is particularly unfortunate in light of recent research (Young, 2002) that suggests that it is precisely the emotional branding moments that largely contribute to brand equity.
And yet, in the United States, recall measures of advertising have been the dominant approach used for at least the past quarter century. Why? Because historically, telephone surveys conducted by large cost-efficient WATS centers were the most economical way of interviewing a projectable sample of consumers. The Internet now provides a viable alternative.
One of the main advantages of using the Internet for studying advertising is that it allows you to put something, a stimulus, in front of the eyes of consumers, to test whether or not they recognize the “face” of the advertising. While some telephone trackers have attempted to produce a “recognition” measure by reading a verbal description of advertising over the phone, it’s quite obvious that this substitute approach is no more reliable than it would be for you to help me find a particular face in a crowd if all you had to go on was a one or two sentence written description of what that face looks like. Getting a visual recognition through a verbal cue is simply not as effective as using a visual cue.
The principal barrier to conducting television advertising research on the Internet is bandwidth — the connection speed needed for playing full-motion, full-screen video. While this remains a problem for pre-testing, there is a relatively simple solution for using the Internet to track TV ad awareness. Bandwidth is not a problem if all you want to show the consumer is the “face” of the ad.
To illustrate, I would like to report the results of a simple experiment we conducted in malls to see how the recognition measurement is affected when different types of stimulus are used. With one group we showed respondents the full-length video of several test commercials; with a matched group of respondents, instead of the video we showed them a photo-board with only four to six key frames taken from each commercial. The experiment involved five commercials from an ad campaign that had been on air for a few months with fairly heavy media weight. Each group consisted of 100 respondents screened to be in the target audience for the advertising.
For each commercial, we asked respondents whether or not they had ever seen the ad on television before. We also asked them how sure they were that they had seen it. The results are shown in Table 1. While claimed awareness for the full video is quite a bit higher than for the photoboard, the results are much more similar if we compare those who were “very sure” they had seen the advertising based on the video cue against those who claimed to have seen the advertising based on the photoboard cue.
The averages for those two sets of ads were statistically the same. In three of the five cases the results were virtually identical. In the last two cases the results for the full video were somewhat higher — though in all five cases the rank order of the ad recognition scores were the same.
The results of this experiment were quite encouraging for online tracking. The approach appears to work well using only a small subset of the total information contained in a television commercial.
As we took the photoboard concept online, both in the U.S. and internationally for several clients, we also made two improvements to increase the accuracy of an Internet ad-tracking system.
First, in examining the two cases where the video and photoboard scores were farthest apart we found that music was an important component of those creative executions. Consequently, to add a “voice” to the “face” of the advertising, we added 10 seconds of streaming audio to accompany the photoboard cue for ad recognition. Streaming audio has much less of a bandwidth constraint than video and can easily be used in Internet research today.
Obviously, six pictures represent only a small part of the visual information content of the average TV commercial. And yet just as in the TV quiz show Wheel of Fortune, where a minimal, if well-chosen, set of information can be enough to trigger recognition, the same principle seems to be operating with the photoboard. The second and more important improvement to our recognition measurement approach, therefore, was to identify a good strategy for choosing the small number of pictures which would be used in constructing the photoboard.
Fortunately, the means for developing such a strategy was already at hand. Ameritest’s Flow of Attention method, which we use in pre-testing commercials (Young, 2001), provides a way of choosing the four to six visuals which are used to construct the photoboard stimulus for online tracking.
The Flow of Attention, which is constructed from the Picture Sorts method, helps our clients make a paradigm shift. You must stop thinking of the human eye as a camera, that is, as a recording device, and start to think of it more as an intelligent search engine that actively sorts through the visual information that is continuously streaming toward it. The process of seeing is quite complex. Indeed, half of the human brain is devoted to the process of constructing visual perception in the mind. Visual perception is, in fact, the act of visual selection. The Flow of Attention is a tool for studying the cognitive search process engaged in by an “intelligent” eye when watching television.
Used as a diagnostic for pre-testing it helps explain why some ads recall well (see Young and Robinson, 1987) and why emotional ads do not. In addition, because it measures how the eye filters visual information, it also tells us which images remain in long-term memory and, therefore, which are the best images to include in the photoboard to be used in the ad-awareness recognition test.
An example of a Flow of Attention graph is shown in Figure 1. In this kind of graph, the pictures are plotted in the order in which they actually appear in the commercial. The height of each picture on the Y-axis shows the percentage of the audience that actually recalls that image only a few minutes after viewing the ad. You will note that the pattern is rhythmic and wavelike, with a high degree of variation in the percentage of recall between different images in the ad. This is because the human eye does not record information, rather it consumes information, as it systematically searches for content that is relevant or meaningful to it. In Figure 1 you are looking at a map of selective perception in action.
The fact that there is a high degree of variation in the recall of the component images in a 30-second commercial has important implications for our photoboard test. In our experience, the range of recall scores for these individual pictures will usually fall between 40 per-cent and 80 percent, and sometimes even more. Moreover, we typically find that it is necessary to use between 20 and 30 pictures to describe the visual content of the average 30-second television commercial. Therefore, if you were to arbitrarily choose six pictures from an ad to use in a photoboard for an online test, the amount of error or noise that would be introduced into your measurement process could be quite large.
If you were to pick six pictures that fall in the valleys or low spots of the Flow of Attention and use those in your photoboard, you would likely find that the test ad has low in-market awareness. However, the pictures that fall on the peaks of our curve — typically only four to six pictures in a 30-second ad — are the ones that the gatekeeper eye of the consumer has already decided are the important ones to store away in long-term memory. So these are the most appropriate ones to choose for a reliable, well-defined measurement of the awareness of television advertising.
In a sense we are applying recognition measurement on multiple levels in a kind of fractal scaling of the analytic process. On a micro level we use the picture sor t to deconstruct a commercial into individual images that are used in a recognition test to analyze the experience of the individual commercial. On a macro level, we are now using this same approach to deconstruct an entire advertising campaign into audience awareness of the individual ads that make up the campaign.
As a result of this deconstruction, there are analytic benefits to this online, recognition-based approach to measuring ad awareness. For instance, with a reliable measure of the awareness of individual ads in the campaign, it is possible to identify the stronger and weaker members of a campaign pool — which can be useful for optimizing the media plan. This can be done by adjusting the awareness level with the media spend behind the individual ads to calculate a measure of ad efficiency, as shown in Figure 2. In this example, ads C and D have roughly the same awareness level, but, after adjusting for differences in media spend, we see that ad C is a strong ad and D is a weak ad because it was necessary to spend 50 percent more media dollars behind ad D in order to generate the same awareness level.
Since in dividual ads wear out much faster than campaigns do (the Marlboro campaign has been running for 50 years) measurements made at the level of the individual ad are key to understanding when a commercial is worn out and needs to be replaced. Asking respondents directly if they’re “tired” of seeing a particular ad may be preferable to the current practice of using theoretical media wear-out models to determine when it is time to rotate in new advertising. This is a second advantage of online, recognition-based tracking.
A third advantage is that it is possible to measure depth of awareness of an ad campaign by looking at segments of the audience who have only seen one ad in a pool, or two of the ads in a pool, or three of the ads, etc., in order to assess the synergy between multiple executions in terms of their cumulative impact on consumer perceptions and behavior. With traditional telephone recall studies you simply know whether or not consumers are aware of a campaign as a whole, but you do not know their degree of awareness of the individual components.
So, the Internet has created a new opportunity for improving the way in which TV commercials are tracked. Recognizing the faces of advertising in that noisy advertising party called the marketplace is more reliable than recall. At first glance, it appears that bandwidth limits the promise of the Internet. However, the missing piece of the Internet puzzle is to use, as a substitute for video, a photoboard constructed with a small set of pictures — peak visuals previously selected by the original search engine, the consumer’s eye.
Young, Charles, “Researcher as Teacher”, Quirk’s Marketing Research Review, March 2001. (Enterarticle Quick Link number 671 in the online Article Archive at www.quirks.com to read the article.)
Young, C. and Robinson, M., “Video Rhythms and Recall”, Journal of Advertising Research, vol. 29, No. 3, June/July 1989.
Zielski, Hubert, “Does Recall Penalize Emotional Ads”, Journal of Advertising Research, Vol. 22, No 1, Feb/Mar 1982.