emotive advertising

Measuring the hidden power ofemotive advertising

Robert HeathBath School of Management

Pam HyderMarketing Consultant

Winner: 2004 David Winton Award for Best Technical PaperWinner: 2004 ISBA Award for Best Paper on Advertising Research

This paper is about advertising that works on our emotions without necessarily

achieving high levels of attention or recall. We compare the most popular recall-

based metric – claimed ad awareness – against an approach that deduces

effectiveness from recognition, and find claimed ad awareness seriously

underestimates the effectiveness of the advertising tested.

Introduction

In 1961, in response to Vance Packard’s famous polemic The HiddenPersuaders, Rosser Reeves (1961) declared: ‘There are no hidden persuaders.Advertising works openly, in the bare and pitiless sunlight.’ Doubtlessthere are some who believe that this is still the case, and that the wayadvertising works is totally transparent. But we know a lot more abouthow our minds and our brains work than we did 40 years ago, and whatwe have learned confirms that advertising, indeed communication ingeneral, is a far more complex process than we used to think it was.

What complicates everything is not claims or brands or products, butemotions — specifically, our emotions as consumers. When Rosser Reevesmade his pronouncement it was believed that emotions were aconsequence of our thoughts, and that if we understood what we werethinking then we understood everything. But pioneers in psychology likeRobert Zajonc and Robert Bornstein shattered this illusion in the 1980s.They showed that feelings and emotions have primacy over thoughts, andthat emotional responses can be created even when we have no awareness

International Journal of Market Research Vol. 47 Issue 5

© 2005 The Market Research Society 467

of the stimulus that causes them. More recently, Antonio Damasio (1994)has proved that our emotions are critical to decision-making; and inanother area of psychology, work by Daniel Schacter and others hasdemonstrated that learning can take place even when we pay no attentionwhatsoever, and that this learning can interact with our emotional memorystores (Schacter 1996).

All this points to advertising having far more power than we think ithas. In general we do not think it affects us very much, if at all. A poll ofAmerican consumers conducted by Cap Gemini Ernst & Young in 2003found 82% of US consumers did not believe that advertising influencedtheir decision to buy a car, which CGE&Y used as evidence that carmanufacturers are wasting money on ads (Financial Times 2003).However, 13 years of IPA Advertising Effectiveness Awards have provedbeyond doubt that advertising affects us, whether we believe it or not. Sohow is it that consultancies like CGE&Y can make such a patently naïveclaim, suggesting that what consumers believe represents the sum of truthabout how advertising works?

We think the explanation arises from the way in which advertising isevaluated. Although our knowledge of how advertising works has changedconsiderably, the measures we generally use to measure it have not. We stillrely on survey data that ask people their opinions of advertising. We stilluse questions that invite people to recall things they have no reason at allto remember. And in many cases we evaluate success using metrics whoseorigins can be tracked back to the early part of the 20th century.

We believe it is time to initiate an investigation into the current approachto advertising evaluation. We focus in this paper on recall measures, andwithin these we investigate in detail one of the most commonly usedmetrics: claimed ad awareness. We test this against a different approach tomeasuring the emotional influence of advertising, and unearth someinteresting discrepancies. We would like to stress that neither of us has anyconnection with any company in the research industry, and we present thispaper as research buyers not research sellers. We hope you will agree thatour investigation has produced some important conclusions.

A brief history of advertising measurement

The remark, attributed in the UK to Lord Leverhulme, that ‘50% of myadvertising is wasted, but I don’t know which 50%’ is well known. Lesswell known is an observation by Niall Fitzgerald (1998), current chairmanof Unilever, who said in an interview in 1998, ‘If someone asked me, rather

Measuring the hidden power of emotive advertising

468

than one of my distinguished predecessors, which half of my advertisingwas wasted I would probably say 90% is wasted but I don’t know which90%.’ Given the skills and technology we now have at our fingertips, it isnothing short of astonishing that the chairman of a major world advertiseris less sure nowadays of the power of advertising than half a century ago.

Partly this situation arises because of the complexity of advertising. BillBernbach’s famous statement in 1980 that ‘Advertising is fundamentallypersuasion and persuasion happens to be not a science, but an art’ gives aclue to the myriad alternative opinions that can be ventured about themerit of even the simplest and smallest piece of marketing communication.And alternative opinions equate to harder decisions and longer decisiontimes. Little wonder that discussions about how well or poorly a campaignhas performed start almost as soon as the first ad appears, because if anadvertising campaign is found not to be working then it may be monthsbefore it can be adjusted, and years before a replacement is available.

So how can it be discovered at this early stage if an advertising campaignis working? In the IPA Awards, advertising effectiveness is generallydeemed to have been proved only when shifts in attitudes or increments insales or margin can be linked directly to advertising activity (Broadbent2000). But we all know that sales are subject to many influences otherthan advertising and can take a long time to respond; likewise we havebecome used to image metrics showing few if any shifts in the short term.As Gordon Brown observed, ‘reality rarely co-operates’ and ‘the majorityof advertisers have to be content with determining the probableeffectiveness’ (Brown 1986).

Generally the advertising industry has favoured fairly simple hierarchi-cal models, so ‘probable effectiveness’ can be determined by collectingdata relating to one of the intermediate stages of whichever model youhave based your advertising on. Thus, in the case of the earliest model,AIDA (Attention → Interest → Decision → Action), the probableeffectiveness of advertising could have been determined by measuring theattention or interest in the ad, were they themselves not almost impossibleto measure.

As it turned out, the first practical intermediate measures were devisedin the 1920s by ‘two young and entrepreneurial college professors fromthe Midwest’ (Feldwick 2002) called Daniel Starch and George Gallup.Starch’s model, ‘Noticed → Read → Understood → Desired → Action’(Starch 1923), makes nothing like as catchy an acronym as AIDA, but itallowed the probable effectiveness to be assessed by the level of ‘noticingand reading’ [sic] that took place. This led him to devise the Starch Test, a


469

recognition-based system for the measurement of ‘Reading and Noting’ ofpress advertising.

George Gallup took things a step further. The Starch Test took noaccount of the decay in memory resulting from the time gap that elapsesbetween exposure to advertising and purchase of product. Gallup deviseda different intermediate measure – spontaneous recall – which allowed forthis memory decay, and his system was introduced as a measure of pressadvertising effectiveness in the 1930s. Later, when TV advertising arrivedin the 1940s, he pioneered with Claude Robinson the famous ‘day-after-recall’ technique. By the 1960s, according to William Barclay andcolleagues, ‘Recall of commercial content was the principal measure ofcommunication effectiveness’ (Barclay et al. 1965).

But it was Gordon Brown, described by Admap in 1994 as ‘probably themost influential single voice in contemporary advertising research’, whowas responsible for turning recall into the potent metric it is today. Brownrealised that genuine spontaneous recall was going to be of less and lessvalue as advertising became increasingly commonplace, so he devised amore sensitive question: ‘We show a list of brands and ask “which of thesebrands … have you seen advertised on television recently”’ (Brown 1985).The usefulness of the resulting metric – claimed ad awareness – wasenhanced by the development of an awareness index, which provided aone-number score for the amount a campaign had increased claimed adawareness. And publication of a study run with Stephen Colman ofCadburys, which showed evidence of a link to sales effect, lent even furtherweight to this metric (Colman & Brown 1986). By 1991 the awarenessindex was described by Feldwick as having ‘gripped the imagination ofboth advertisers and agencies as a simple figure … a measure of advertisingeffectiveness’ (Feldwick et al. 1991).

Measures such as brand name prompted ad awareness and detailedrecall have appeared in almost every major advertising tracking study forthe past 20 years, and this ubiquity has encouraged a simplistic view ofadvertising effectiveness to develop. Gordon Brown described an ad thatfailed to achieve high recall as ‘a disaster’ and one that did achieve highrecall as ‘a triumph’ (Brown 1985). Haley and Baldinger (1991), writingof an experimental study into advertising copy testing, observe that‘Persuasion and recall … are likely to remain primary evaluative measuresin the foreseeable future’. Even notable academics like John Rossiter andLarry Percy accept the importance of recall without question, elevatingbrand-prompted ad recall to the status of ‘a check on advertising’scausality in influencing brand attitudes’ (Rossiter & Percy 1998).


470

But has anyone actually checked how valid these recall-based metricsare? Common sense suggests that advertising will work best if weremember it, but does this mean that advertising works less well if wedon’t remember it? Evidence that we may be wrong to place so much faithin these types of metrics comes from Leonard Lodish’s massive meta-analysis of 389 split cable TV advertising experiments using theBehaviorscan® panel. Lodish and Abraham concluded that, ‘It is unlikelythere is a strong relationship between standard measures of TVcommercial recall … and sales impact of the copy’ (Lodish & Abraham1995).

So what is the case against the use of recall-based metrics? And, morespecifically, how well do they evaluate emotive advertising?

The case against recall-based metrics

The belief underlying recall-based metrics is that advertising has to bepersuasive in order to be effective. This is particularly the case in the USA,where, as Tim Ambler eloquently puts it, ‘The assumption that advertisingequals persuasion is so ingrained … that a challenge elicits much the samereaction as questioning your partner’s parentage’ (Ambler 2000).Persuasion, defined in the Longman Dictionary as ‘to move by argument,reasoning, or pleading to a belief, position, or course of action’, clearlyworks best if the person you are trying to argue, reason or plead with paysattention to and recalls what you are saying, otherwise your efforts will belargely wasted.

This is almost certainly why marketing universally assumes that highattention equates to advertising effectiveness. For example, Philip Kotlerand colleagues, authors of the world’s most popular marketing textbook,state unhesitatingly that, ‘The advertiser has to turn the “big idea” into anactual ad execution that will capture the target market’s attention andtheir interest’ (Kotler et al. 1999). And even the UK’s most celebratedmarketing academic, the late Peter Doyle, wrote ‘For an advertisement …to be effective it must achieve first exposure and then attention’ (Doyle1994). But here we encounter an interesting non sequitur. Although theytalk about the need for advertising to get attention and for the message tobe remembered, neither Kotler nor Doyle says anything about the need forthe advertising itself to be remembered – yet both include advertising recallin their list of recommended metrics. This suggests that their list ofevaluative metrics is derived not from their own criteria for advertisingeffectiveness, but from the measures that are widely used.


471

If remembering advertising has no theoretical role in advertising success,just how important or useful a metric is advertising recall? In fact, the ideathat advertising can work without ads being attended to or recalled goesback a very long way. In 1903, just five years after the AIDA model wasconceived, the importance of attention was questioned by Walter DillScott, who quoted a subject who claimed never to have looked at any ofthe ads in the tramcars she travelled in each day, yet ‘knew them all byheart and … held the products they advertised in her highest esteem’ (Scott1903). The schism between high and low attention models widened in the1960s, when Herb Krugman coined the phrase ‘low involvement’ tocontrast with the persuasion-shift high-involvement models thatdominated ad pre-testing at the time, later validating the incidence of lowinvolvement by measuring the brainwaves of a subject while watching TVadvertising (Krugman 1971). More recently, a team led by psychologistStewart Shapiro has shown by using controlled attention experiments thatads can influence product consideration even when processing was entirelyperipheral. Their conclusion is that advertising ‘has the potential to affectfuture buying decisions even when subjects … do not process the adattentively and … do not recollect ever having seen the ad’ (Shapiro et al.1997).

It is difficult to source data from real life to support the idea that ads canwork without being remembered. Partly this is because marketers whoregard ad awareness as important will usually change their advertising if itperforms badly on this metric. What little evidence there is comes frombrands whose marketers regards ad awareness as unimportant andtherefore not worth measuring. One such example is Stella Artois, nowacknowledged as the most successful beer brand in the UK. Despite havingrun for eight years, Stella’s initial press campaign (as measured by acompetitor’s tracking study) had only achieved claimed ad awareness of4% in 1990, compared with 29% for the leading TV-advertised brandCastlemaine XXXX. Yet Stella’s rating for quality on the same survey was45% compared to just 19% for Castlemaine. A rigorous analysis of allother factors indicated it could only have been the advertising that gave thebrand its exceptionally high repute, thereby confirming that advertisingcan build strong brand values without necessarily performing well onmemory-based evaluative measures (Heath 1993).

So there is evidence to suggest that recall-based metrics may not bereliable indicators of advertising effectiveness. What is the explanation forthis? How can advertising work without being explicitly remembered? Aclue comes from a paper presented at Esomar in 2000, which found


472

evidence that UK advertising simply does not conform to the traditionalpersuasion norm so popular in the USA. In a study of 36 UK adver-tisements the authors noted that the majority fitted the non-persuasivecategory, with 86% being seen as having told viewers nothing they did notalready know. In their conclusions the writers demurred from the task ofexplaining how these ads might work, but in the text they admit that theirmeasures were rather rational and were flawed by ‘not picking up allemotional facets fully’ (Mills et al. 2000). And it is in this area of emotionthat we believe advertising has the ‘hidden power’ that enables it to workwithout attention or recall.

Non-persuasive advertising models

Krugman developed the first recorded non-memory dependent advertisingmodel. In his 1977 paper he concludes that ‘recall and attitude effects arenot necessary for advertising to do its job of aiding in-store purchasing’and that ‘quick and/or faint perceptions of product advertising, evenunremembered, do their job in most cases’ (Krugman 1977). ButKrugman’s explanation focused on left- and right-brain processing, anddid not touch on the role of feelings and emotions. This is no surprisebecause it was another three years before Zajonc published his revelationabout the primacy of feelings over thinking (Zajonc 1980). More recently,Damasio and others have confirmed that, when it comes to decision-making, affect (i.e. feelings and emotions) always dominate cognition (i.e.thinking) (Damasio 2000).

This new learning about the power of affect led Tim Ambler of LondonBusiness School to develop what he calls his MAC (Memory-Affect-Cognition) model (Ambler 2000). In this model he concludes that it is theaffective content that drives advertising effectiveness and that most of our‘thinking’ is merely supportive of the decision that you ‘feel’ is the rightone. In his words, ‘Memory dominates Affect, which in turn dominatesCognition’. Furthermore, Ambler recognises the problem such a modelposes for advertising evaluation, stating that ‘The difficulty of measuringAffect needs to be considered, not ignored’.

Ambler’s MAC model does not discuss the role of attention or recall inadvertising. One model that does is Heath’s Low Attention ProcessingModel (Heath 2001), which specifically describes how advertising canwork without high levels of attention being paid, and without beingrecalled. Note that this was formerly known as the Low Involvement


473

Processing Model, which caused confusion in the USA with models thatuse involvement to refer to product or category involvement.

The Low Attention Processing Model can be summarised as follows.

1. Because brands match each other’s performance so swiftly, andconsumers exist in a time-poor environment, considered choice tendsto give way to intuitive choice, in which emotions are more influential.

2. This situation inhibits the consumer’s desire to seek out informationabout brands, and minimises the need for them to pay attention toadvertising. Brand information can, however, be ‘acquired’ at low andeven zero attention levels, using two distinct mental processes. Thefirst process is passive learning, which is a low-attention cognitiveprocess. Passive learning has been shown to be poor at changingopinions and attitudes (Petty & Cacioppo 1996) but is able to recordand link together brand names and other elements in an ad.

3. The second process is implicit learning, which is a fully automatic non-cognitive process that has been shown to be independent of attention.Implicit learning cannot analyse or reinterpret anything: all it is ableto do is to store what is perceived, along with any simple conceptualmeanings we attach to these perceptions.

4. Because of this limitation, implicit learning does not establish strongrational brand benefits in the consumer’s mind. Instead it builds and reinforces associations over time and these associations becomelinked to the brand by passive learning. These associations areextraordinarily enduring, and can trigger emotional markers, which inturn influence intuitive decision-making (Damasio 1994).

5. Passive and implicit learning are semi-automatic and fully automaticmental processes. As such they will be used every time an ad is seen orheard, regardless of how little attention is being paid. Becauseattention to advertising tends to diminish over time, the occasions onwhich an ad is processed attentively will be outnumbered many timesby the occasions on which it is processed at lower attention and itscontent is learned passively and implicitly.

So advertising that exploits low-attention processing will work betterwhen seen several times at low attention than if seen once or twice at high


474

attention. The implication for research is that brand associationsreinforced by this sort of repetition may remain in memory long after thead has been forgotten. If this is the case then metrics based on memory ofthe ad itself, such as claimed ad awareness, are likely to underestimate theeffectiveness of advertising that has a strong affective content.

Support for the Low Attention Processing Model comes from the mostrecent work published by Damasio, which has shown that affect isprocessed instantaneously and irrespective of how much attention is paid.Partly this is because of the way our memory operates. Memory has twodifferent types of store: a permanent store called long-term memory and atransient store called working memory. Working memory is the systemthat enables us to organise (i.e. cognitively analyse and interpret) themillions of perceptions we receive every moment of the day. All ourattentive processing operates through our working memory, thuseverything that drives recall of advertising has first to pass throughworking memory. What Damasio has discovered is that feelings andemotions are processed without the use of working memory, and so bydefinition must be processed automatically and implicitly (Damasio 2000).

All this suggests that metrics based on recall are likely to work even lesswell when it comes to measuring advertising with a strong emotivecontent. In order to examine this hypothesis we need to identify anintermediate metric that will be capable of measuring advertising thatworks affectively and inattentively. This is discussed in the next section.

Metrics based on implicit learning

Implicit learning is known to interact directly with a separate memorysystem called implicit memory (Schacter 1996; Heath 2001), so the way tofind a metric that measures implicit learning is by finding one that iscapable of tapping into implicit memory. We know from an experiment byGardiner and Parkin that recall is mainly a test of explicit memory,because it diminishes in divided-attention situations and increases whenfull attention is paid (Eysenck & Keane 2000). But there is another test ofmemory that is used extensively in psychology because it is more powerfulthan recall and much less dependent upon attention. That test isrecognition. Recognition memory has been shown in experiments to beeffectively inexhaustible. Lionel Standing, whose experiment showedrespondents able to recognise up to 10,000 pictures without difficulty,concluded that ‘recognition memory of the subjects seems unsaturatable’(Standing 1973).


475

The value of using recognition is that it taps into both explicit andimplicit memory, which means we can get a much better idea of the actuallevel of advertising exposure that has taken place. By identifying thosepeople who have seen the advertising and subsequently forgotten it weshould be able to test directly how effective recall is at evaluating emotiveadvertising.

A controlled test of recall vs recognition metrics

The critical area of debate in this paper concerns the implication that theprocessing of affect has for ad tracking research. If, as is predicted by theLow Attention Processing Model, brand associations and their emotivelinks endure in memory beyond the point at which conscious recollectionof the ad itself disappears, then measures such as claimed ad awarenessand detailed recall are likely to underestimate the effectiveness ofadvertising that has a high affective content. In simple terms, significantnumbers of people who have been exposed to the ad and influenced by itwill not actively remember it and will therefore not believe the brand hasbeen advertised recently.

This hypothesis can be tested by collecting both claimed ad awarenessand recognition, and cross-tabulating them against a dependent measurethat links to sales. Sales intention is one possible dependent measure, butBarnard has found that this is more influenced by what was bought lasttime than by what is likely to be bought next time (Barnard 1994). Abetter measure is favourability, which has been shown in longitudinal teststo predict future brand choice (Hofmeyr & Rice 2000).

An initial examination of these two metrics was run in 2002, and apaper by Robert Heath describing this has been accepted for publicationby the Journal of Advertising Research. The findings from this study arebriefly summarised in the next section.

Butchers Dog pilot study

A research study was run in the UK on a new TV advertising campaign forthe Butchers Dog brand of dog food, which had not advertised for over ayear. The campaign was chosen because it offered no persuasive evidencethat the brand made dogs fit (the brand claim) but instead featured simplevisual playlets of dogs doing things that they physically cannot do. Thiscreative approach was deemed likely, after initial exposure, to encourageenjoyable but essentially low-attention processing.


476

A single wave of research was run on the Taylor Nelson CAPI Omnibusthree weeks after national transmission of the advertising had started.Favourability towards the brand was measured using a ten-point semanticscale ranging from ‘very unfavourable’ to ‘very favourable’. Claimed adawareness was asked using the brand-name prompted question alluded toearlier: ‘Have you seen any advertising on TV recently for (brand)?’ Thosewho answered in the affirmative were then asked to describe anything theycould recall about the advertising. Recognition was measured by playingthe ad in full to respondents. In order to minimise overclaim throughexpectation of the brand advertising, the ad was adapted to remove allreference to the brand, and respondents were asked if the advertising hadbeen seen by them several times, once or twice, or not at all.

It should be noted that this definition of recognition differs from thatwhich is often quoted by ad research companies. For example, MaxSutherland defines recognition as ‘People … shown photo-stills and askedwhether they have seen that ad’, an approach that can result in poorrecognition if a key element in the advertising is sound or music or actionof some sort (Sutherland 2000). Recognition can be accurately assessedonly by showing the entire ad as stimulus, and in our test we removed onlythe brand references, which was necessary in order to preclude responsesbased on an expectation that the advertising would have been seen.

Results

Because brand users rate their brands higher than non-users, analysis wasrestricted to users of dog food who were non-users of the Butchers Dogbrand. This avoids errors due to different proportions of users and non-users in each part of the sample. Thus all results shown are from users ofother brands.

The first step was to examine the exposure levels as predicted by the twometrics, and a cross-tabulation of the sample of 318 respondents is shownin Table 1.


477

Table 1 Users of dog food who were non-users of the brand

Claimed seen brand Claimed not seen brand Total

advertising on TV recently advertising on TV recently (column %)

Recognised one or other execution 70 ( 152 ( 222 (70%)

Recognised neither execution 10 ( 86 ( 96 (30%)

Total (row %) 80 (25%) 238 (75%) 318 (100%)

As we predicted, claimed advertising awareness indicates a lot lessexposure than recognition: just 80 respondents (25% of the total sample)claimed they had seen the brand advertising on TV, but 222 respondents(70% of the total sample) recognised one or other of the executions. Andwhen the 80 people who claimed the brand was advertising were askedwhat detail they recalled, only one-quarter (6% of the total sample) wereable to identify anything from the new campaign. These findings confirmthe expected poor performance of explicit memory in respect ofadvertising with a strong emotive content.

A 25% claimed ad awareness might be considered by some to be anacceptable level just three weeks after advertising has started, but one hasto ask what exactly this metric is measuring. The received wisdom is thatclaimed ad awareness is an indication of those respondents who have seenthe advertising and lodged it in their mind. But evidently it is not a verygood measure of this, because 10 of the 80 people who claimed the brandhad been advertising did not recognise either ad, from which it can bededuced that they had either not seen the advertising at all and wereguessing, or their appreciation of time was so muddled that they thoughtthe campaign that had been on air over a year earlier had been on air recently.

If claimed ad awareness is indeed a measure of those who have seen andremembered the advertising, then it makes sense that these people shouldmanifest the maximum effect of the advertising. In this study we measuredadvertising effectiveness by the shift in favourability towards the brand. Soif the advertising is effective and claimed ad awareness is representative ofthose people who have seen and remembered the ad, then those whoclaimed they had seen the brand advertised on TV recently should havehigher brand favourability than those who did not. The results of a cross-tabulation of these two metrics is shown in Table 2.

These results show there is no difference in favourability between thosewho claimed to have seen the brand advertised on TV and those whoclaimed not to have seen it advertised on TV. Both groups rated the brandat 4.33. So if claimed ad awareness truly measures the optimum type of adexposure then this campaign is clearly ineffective, because there was noshift in brand favourability.


478


Claimed recently Claimed not recently

Total advertised on TV advertised on TV

Sample size (%) 318 (100%) 80 (25%) 238 (75%)

Favourability mean (10-point scale) 4.33 4.33 4.33

But what happens if we now look at the brand favourability amongthose who recognised one or other of the ads, and can therefore beassumed to have really seen them? Also, bearing in mind that the LowAttention Processing Model suggests that repeated viewing is moreeffective than one or two exposures, what is the brand favourability ofthose who saw an ad several times compared with those who saw it justonce or twice? These results are given in Table 3.

Here a very different picture emerges. Those who recognised one orother of the ads are more favourably disposed to the brand than those whodid not recognise either ad (and bear in mind that these are all non-usersof the brand). And an even higher level of brand favourability is observedamong those who said they had seen an ad several times.

So there is a positive favourability shift among those who recognised anad, but no shift among those who claimed the brand had been advertisedon TV. What explains this disparity? The best way to find out is to look atthose who claimed the brand had been advertised and see if theirfavourability scores vary across different levels of actual exposure. Table 4shows the results.

What is at once evident is that those who saw the ad several times(column C) have the highest rating for the brand, higher than those who


479


Did not Recognised Recognised an Recognised an

recognise one or execution and execution and

either other seen once seen several

execution execution or twice times

Sample size (%) (n = 318) 96 (30%) 222 (70%) 90 (28%) 132 (42%)

Favourability mean (10-point scale) 4.14 4.41 4.20 4.56


Claimed Claimed Claimed

Claimed advertised advertised advertised and

Total: claimed advertised and and saw ads and saw an ad saw ads once,

advertised not seen ad once or twice several times twice or not at all

on TV (A) (B) (C) (A) + (B)

Sample size (%) 80 (25%) 10 (3%) 21 (7%) 49 (15%) 31 (10%)

Favourability mean

(10-point scale) 4.33 3.60 3.38 4.90** 3.45**

**Significant @ 99%

saw the ad only once or twice (column B), or those who did not see the adat all (column A). Unfortunately the sample size of the cells A, B and C istoo small to allow significance to be tested. But if we combine columns Aand B we can compare those who did not see an ad or saw one only onceor twice with those who saw an ad several times. The differences are nowsignificant: an ANOVA reveals the difference in mean favourabilitybetween these two groups (columns A + B and C) is significant at 99%[F (468) = 7.24, p < 0.009].

The explanation is now clear. Those who claimed they had seen thebrand advertised comprised two different groups: one had been exposed tothe advertising once or twice or not at all, and had not been affected by it;the other had been exposed to the advertising several times, and had beenaffected by it. In summary, as the Low Attention Processing Modelpredicts, claimed ad awareness produced a sample whose consistency inrespect of their actual exposure and response to the advertising wasseriously flawed.

This study was the first one conducted. In order to validate the findingswe decided to repeat the study, this time with advertising specificallydesigned to exploit the Low Attention Processing Model in the way itinfluenced consumers’ feelings.

Standard Life test

In 2002 Standard Life decided to replace its advertising campaign.Working with The Leith Agency and The Value Creation Company, itdeveloped a new campaign, the purpose of which was to increasefavourability towards Standard Life. This was achieved by illustratingthings in life that people like and linking this positive feeling to theStandard Life brand using the line ‘I like Standard Life’.

Primary media used were outdoor and TV, and the campaigncommenced on 6 October 2003. The advertising was tracked by theNunwood Consultancy in two independent surveys: a continuous CATI(computer-assisted telephone interview) survey collected data on brandawareness, brand favourability and claimed ad awareness; a monthly face-to-face CAPI (computer-assisted personal interview) survey collected dataon these measures also, in addition exposing the debranded ads andcollecting recognition, estimated frequency of exposure and brandattribution. Question design corresponded with that used in the ButchersDog pilot described above, except that detailed recall was not asked in thisstudy.


480

CAPI results

The first wave of CAPI research was run 11 days after the start ofadvertising. As in the pilot research our first step is to examine the sample,which is shown in Table 5.

Once again it is evident that claimed ad awareness under-represents theactual exposure of the campaign. Some 27% claimed that Standard Lifehad been advertised on TV recently, compared with 65% who recognisedand can therefore be deemed to have actually seen the advertising. And37% of those who claimed Standard Life had been advertising (18 out of48 respondents) did not recognise the new campaign at all.

Next we analyse favourability among those who claim the brand has andhas not been advertised on TV recently. These results are shown in Table 6.

This time we see that the favourability score among those who claim thebrand had advertised recently is actually lower than among those whoclaim the brand has not advertised recently. Again, the clear implication isthat the advertising has not worked.

Now compare these results to the analysis of recognition, in Table 7.


481

Table 5 Non-users of Standard Life

Claimed seen Claimed not seen

TV advertising TV advertising Total

for brand recently for brand recently (column %)

Recognised one or other execution 30 ( 87 ( 117 (65%)

Recognised neither execution 18 ( 45 ( 63 (35%)

Total (row %) 48 (27%) 132 (73%) 180 (100%)


Claimed recently Claimed not recently

Total advertised on TV advertised on TV

Sample size (%) 180 (100%) 48 (27%) 132 (73%)



Did not Recognised Recognised

recognise one or other an execution and

Total either execution execution seen several times

Sample size (%) 180 (100%) 63 (35%) 117 (65%) 49 (27%)

Favourability mean (10-point scale) 5.98 5.65 6.15 6.16

Again we see that those non-users who recognise the advertising aremore favourably disposed towards Standard Life than those who do not.In this case, however, the favourability among those who have seen the adseveral times is not higher. The conclusion is that this advertising eitherworks very quickly, or perhaps needs a longer exposure period than thedog food advertising. We address this point later on.

The final analysis of the CAPI data is a cross-tabulation of those whoclaim the brand was advertised recently against recognition, and theresults are shown in Table 8.

Again we see that the low favourability score among those who claimedthat the brand had been advertising on TV recently is caused by the 37%who had not actually seen the advertising. Their mean favourability is4.99, compared with a mean favourability of 6.21 among those whorecognised the ad.

CATI results

By running a telephone survey alongside the CAPI survey we were able tosee how the recognition, claimed ad awareness and favourability changedover time. A graph of the results is shown in Figure 1.

Recognition and claimed ad awareness both rise during the initial heavyburst of advertising. However, when the advertising weight drops at theend of week five we see claimed ad awareness starts to fall. In contrast, therecognition measures stay at roughly the same level.

This timing differential is important. The trend in claimed ad awarenessis suggesting that the advertising’s effectiveness starts immediately, risesevenly, peaks around the end of the heaviest burst, then declines when thelighter drip campaign starts. The conclusion that would be drawn fromthis trend is that the advertising has little long-term effect, and needs to beheld at the higher weight in order to have any effect. But when we look atfavourability we find a different trend. Favourability does not move untilthe end of the burst, at which point it rises and stays permanently higher.


482


Total claimed Claimed advertised Claimed advertised

advertised recently and recently and

recently on TV not seen ad seen ads

Sample size (%) 48 (100%) 18 (37%) 30 (63%)


This mimics the movement of the recognition measures, and suggests acampaign that requires a sustained period of repeated exposure beforeconsumer attitudes start to be influenced, but once this happens then theadvertising can safely drop to a lower weight.

Discussion of results

The results support the theory that advertising which relies upon affect(i.e. that mainly influences emotions and feelings) is likely to benefit fromlonger periods of repeat exposure. It also supports the hypothesis thatrecall-based metrics such as claimed ad awareness are likely to seriouslyunderestimate the effectiveness of advertising that relies on affect. Whatdoes this imply for the research industry?

The question ‘Have you seen brand X advertised on TV recently?’ is onethat we who work in the industry instinctively like and understand. Wecan answer it easily and accurately, because we pay attention to adverti-sing: it is, after all, what pays our wages. The results above suggest thatthe clarity with which we can answer this and similar questions is notalways shared by consumers. Ordinary people, who watch TV as part oftheir day-to-day routine, almost certainly pay a lot less attention to TV


483

0

10

20

30

40

50

60

70

80

1 2 3 4 5 6 7 8 9 10

%

4.00

4.20

4.40

4.60

4.80

5.00

5.20

5.40

5.60

5.80

6.00

TV

65 %

75 %

27 %

44 %

74 %

48 %

Figure 1 Rolling four-week trends

Standard life non-users

Weeks

Claimedad awareness

Recognise Meanfavourability

Seenseveral times

Mean favourability

advertising than we do and remember a lot less of it than we do. So whenasked if they have seen a brand advertised on TV recently they can get theanswer wrong; and when asked what they can recall about the advertisingfor a brand, they sometimes recall nothing. But this does not necessarilymean that advertising has had no influence on their attitudes towards thebrand.

The effect that advertising has on consumers’ emotions and feelings (i.e.affect) is largely unknown to them, which is why we use the term ‘thehidden power of advertising’ to describe it. Advertising that relies on affectis increasingly common: indeed, the majority of advertising employs affectto a greater or lesser degree nowadays. Yet the systems we use to evaluateadvertising are primarily designed to evaluate advertising that is trying todeliver a rational persuasive message. The results of our research suggestthese systems need to be revised to accommodate the new paradigmswithin which modern advertising works.

Conclusions

We believe the findings of this paper call into question the value both ofrecall metrics and continuous tracking research. The continuous collectionof data in tracking research is needed primarily in order to collect recall-based metrics such as claimed ad awareness, but this type of metric hasbeen shown to provide confusing and misleading findings for advertisingwith a high emotive content. Under the circumstances we conclude thatcontinuous research is unnecessarily costly: the findings derived from theperiodic studies we used were more accurate and informative, andconsiderably less expensive.

Overall, we conclude that methods based upon the periodic measure-ment and cross-analysis of recognition and brand metrics represent abetter future direction for advertising research, because they are uniquelyable to evaluate the hidden emotional power that advertising undoubtedlyhas.

However, there is one caveat. It should be noted that these tests wereconducted on advertising selected for having a high affective content.Although low-attention processing is applied by the consumer to alladvertising, it is important to stress that the model does not necessarilyapply to all advertising. It will not be as relevant to advertising whosemain purpose is to get a direct response, or advertising that is trying toimpart ‘new news’, as both of these types of advertising will benefit byachieving the highest level of attention they can. It is possible that the use


484

of recall-based metrics can be justified on such advertising, and weencourage this to be investigated.

References

Ambler, T. (2000) Persuasion, pride and prejudice: how ads work. InternationalJournal of Advertising, 19, p. 299.

Barclay, W.D., Doub, M.R. & McMurtrey, L.T. (1965) Recall of TV commercials bytime and program slot. Journal of Advertising Research, 5, 2, p. 41.

Barnard, N. (1994) What can you do with tracking studies and what are theirlimitations? Admap, April, pp. 1–26.

Broadbent, T. (2000) Advertising Works 11. Henley-on-Thames: World AdvertisingResearch Center.

Brown, G. (1985) Tracking sudies and sales effects: a UK perspective. Journal ofAdvertising Research, Feb./March, pp. 56, 57.

Brown, G. (1986) Modelling advertising awareness. The Statistician, 35, p. 286.Colman, S. & Brown, G. (1986) Advertising tracking studies and sales effects.

Journal of the Market Research Society, 25, 2.Damasio, A.R. (1994) Descartes’ Error. New York: G.P. Putnam.Damasio, A.R. (2000) The Feeling of What Happens. Heinemann.Doyle, P. (1994) Marketing Management and Strategy. Hemel Hempstead: Prentice-

Hall, p. 240.Eysenck, M.W. & Keane, M.T. (2000) Cognitive Psychology, 4th edn. Hove:

Psychology Press, p. 181.Feldwick, P. (2002) What is brand equity, Admap, 430, July.Feldwick, P., Cook, L. & Carter, S. (1991) Ad effectiveness and ad awareness.

Admap, March, p. 22.Financial Times (2003) 14 October.Fitzgerald, N. (1998) A man with a mission. Interview by Judie Lannon. Market

Leader, 2, Autumn, p. 22.Haley, R.I. & Baldinger, A.L. (1991) The ARF Copy Research Validity Project.

Journal of Advertising Research, 31, 2, p. 30.Heath, R.G. (1993) Reassuringly expensive – a case history of the Stella Artois press

campaign, in Advertising Works 7, Baker, C. (ed.). Henley-on-Thames: NTCPublications.

Heath, R.G. (2001) The Hidden Power of Advertising. Admap Monograph No. 7.Henley-on-Thames: World Advertising Research Center.

Hofmeyr, J. & Rice, B. (2000) Commitment-led Marketing. Chichester: Wiley.Kotler, P., Armstrong, G., Saunders, J. & Wong, V. (1999) Principles of Marketing,

2nd European edn. Prentice-Hall, p. 668.Krugman, H.E. (1971) Brain wave measurement of media involvement. Journal of

Advertising Research, 11, 1, Feb., pp. 3–9.Krugman, H.E. (1977) Memory without recall, exposure without perception.

Journal of Advertising Research, 17, pp. 52–53.


485

Lodish, L.M. & Abraham, M. (1995) How TV advertising works: a meta-analysis of389 real world split cable TV advertising experiments. Journal of MarketingResearch, 32, 2, p. 138.

Mills, P., Kennedy, R., Ehrenberg, A. & Schlaeppi, A. (2000) The forms that adstake. Paper presented at the joint Esomar ARF Conference, May.

Petty, R.E. & Cacioppo, J.T. (1996) Attitudes and Persuasion: Classic andContemporary Approaches. Boulder, Colorado: Westview Press.

Reeves, R. (1961) Reality in Advertising. New York: Knopf, p. 70.Rossiter, J. & Percy, L. (1998) Advertising, Communications, and Promotion

Management. Singapore: McGraw-Hill, International Editions, p. 587.Schacter, D.L. (1996) Searching for Memory. USA: Perseus Books.Scott, W.D. (1903) The Psychology of Advertising. Boston: Small, Maynard & Co.Shapiro, S., MacInnis, D.J. & Heckler, S.E. (1997) The effects of incidental ad

exposure on the formation of consideration sets. Journal of Consumer Research,24, June, p. 102.

Standing, L. (1973) Learning 10,000 pictures. Quarterly Journal of ExperimentalPsychology: Learning, Memory, and Cognition, 19, pp. 582–602, cited in Rose, S.(1992) The Making of Memory. Uxbridge: Bantam Books, p. 117.

Starch, D. (1923) Principles of Advertising. Chicago: A.W. Shaw.Sutherland, M. (2000) Do you model ad awareness or advertising awareness?

Journal of Advertising Research, 40, 5, Sept./Oct., pp. 32–36.Zajonc, R.B. (1980) Feeling and thinking: preferences need no inferences. American

Psychologist, 35, pp. 151–175.


486

emotive advertising

Documents