by Byron Sharp
There is solid empirical evidence that brand advertising does produce sales. Yet the effects of advertising are very difficult to see in sales trends. The reasons for this are not well understood, and this ignorance means that some people who study sales effects (e.g. econometric modelling consultancies and academics) often get their estimates wrong.
There are two reasons why advertising's sales effects are hard to see in weekly and monthly sales figures (i.e. sales neither jump when advertising starts or increases, nor collapse when it is reduced). The proper effect of advertising on sales is often missed because:
•The aim of most advertising is to maintain market share, that is, to prevent sales from not occurring. Few companies spend enough (or create good enough advertising) to increase market share or accelerate an upwards trend. Mostly, advertising is used to prevent what (without advertising) would be a very gentle decline in sales. Broadbent (1989) provides an apt analogy of this: “The sales of a brand are like the height at which an airplane flies. Advertising spend is like its engines: while the engines are running everything is fine, but when the engines stop, the descent eventually starts”. Much advertising is aimed at preventing competitors' advertising from stealing future sales. This prevention means a brand enjoys sales, over a long period of time, that it would not have otherwise obtained if it had not advertised. So advertising causes sales even if a brand's sales figures are flat.
•Advertising's sales effects are spread out in time. This means that the effect of today's advertising is layered very thinly across weekly sales figures over a long time period. It typically takes dramatic bursts of expenditure to cause a week's figures to detectably increase. This is particularly the case for large brands, because their advertising expenditure is small in comparison to their other marketing efforts, such as their ongoing word of mouth, their sheer physical presence (shelf space is advertising too) and the already established mental availability (past advertising effects) in many consumers' heads. Changes in advertising weight are more likely to make weekly sales figures of a small brand move up and down because it doesn't have a large base of sales that are supported by existing market-based assets.
Consequently, statistical models of changes in advertising spend and changes in overall sales report a weak association. But if we look below the surface and divide the consumers who were exposed to the advertising from those who are not, then we can see sales effects that were previously masked. The value of ‘single source’ individual-level data is illustrated in Figure 9.1, where the data shows that despite declining sales (because of competitive pressure) the brand's advertising really was generating purchases among those consumers who saw the ad.
Figure 9.1: Brand purchasing households, per dollars/1000 households
Source: Flaherty 2007.
When more rigorous controls of other causal influences on sales are in place (e.g. Danaher, Bonfrer & Dhar 2008), advertising's effects on aggregate data appear greater, as we would expect; however, with aggregate data this is fiendishly difficult to do, making it astonishingly hard to see advertising's proper sales effects.
This is a difficult concept to grasp because it seems counterintuitive – advertising causes sales even when weekly sales figures don't change.
When sales figures do change, this still gives an incorrect indication of the true effect advertising has on sales. Movements in this week's or month's sales figures don't reveal the full sales impact of advertising – they are like the tip of the iceberg but we don't know what proportion of the total iceberg the tip represents.
This isn't always the case; for some advertising, the effect on sales occurs immediately. An example of this is the advertising that informs its audience of a time-dependent event, for example, 'Sale ends this Thursday'.69 But most advertising affects people who won't buy the brand for weeks, so the sales effects are spread far into the future. Hence, the effect on sales is spread so thinly over a week's sales figures, that a change in weight of advertising delivers no visible change in sales trajectory. A little bit of your advertising today works today, but much is to prevent people from being less likely to buy you when they make their next category purchase in months/years time.
How advertising and price promotions cause sales in different ways
Chapter 4 discussed how all brands, even very large ones, have mostly very light (i.e. very occasional) buyers (see Figures 4.1, 4.2 and 4.3). We saw that most Coca-Cola buyers only buy themselves a Coke once or twice a year. At the other end of the spectrum, there is a tiny portion of consumers who buy Coke every morning, noon and night. Consider what happens to the buying behaviour of a person who drinks Coke several times a day after seeing a Coca-Cola ad – nothing changes. So who is Coke's advertising aimed at? It's aimed at the millions of people who occasionally buy Coca-Cola, who scarcely ever think of Coke and who seldom buy it. These consumers could easily forget about it and not make their semiannual purchases. This is why established brands need to advertise: to hold onto their buyers (in the face of substantial competitor advertising) and to give themselves a chance to grow.
Coke's advertising tries to grab a tiny window of our attention and remind us that Coke is fun, that we've had it before and we like it – Coke's advertisements are mostly reminding us of things we already know. By doing this, Coke is trying to increase our (very low) probability of buying it tomorrow. If it works, it changes our probability of buying tomorrow from almost nothing (say 1 chance in 400) to slightly more than almost nothing (say 2 chances in 400). This is a change in probability that is so slight we'd hardly notice, which is why people say that advertising doesn't affect them, because it's too subtle for them to notice. But if every person who is exposed to a single Coca-Cola ad increases his or her buying propensity from 1 chance in 400 to 2 chances in 400, Coke's sales among this group double! 70 Therefore, even advertising that says nothing persuasive and gives no new reasons to buy can have a dramatic impact on sales – without causing people to rethink their opinion of a brand, and largely without them even noticing.
This explains why even established brands have to spend considerable sums on advertising – the bigger the brand the more they have to spend, even if they are old and well known. It also explains why consumers often say they don't believe they are affected by advertising. In addition, it accounts for why advertising's sales effects show up in sales figures in a very different way than price promotions do. Put simply, the entire sales effect of a price promotion shows up in a week's sales figures, while the sales effect of advertising is spread very thinly into the future; this makes changes in advertising weight very difficult to see and extraordinarily hard to measure/model reliably.
Any marketing intervention, if it succeeds, works by increasing the probability that customers will purchase a product. Successful advertising reaches and affects many consumers. It reaches all types of consumers, not just the ones who are easy to reach (i.e. a brand's heavy, regular buyers, who easily notice the brand's advertising because their brand-related mental structures are better developed and they are less likely to screen out the advertising).
Successful advertising, in particular, reaches the millions of consumers who have a very low probability of buying a brand next week or month. If a brand's advertising reaches them, if they notice it, if it refreshes or builds the memory structures that make the brand more likely to be noticed or come to mind in a buying situation (i.e. enhancing mental availability), then it nudges their propensity to buy the brand. This is the sales effect. But most of these additional sales won't show up in this week's sales figures because hardly any of these consumers will buy this week; they simply don't buy that often even with their newly enhanced propensity. Indeed, most of these extra sales will never show up as part of a lift in overall sales because, before the consumer even buys from the category, he or she will be hit with competitor advertising (or other marketing activity) that nullifies the effect of the advertising exposure. But this doesn't mean that the sales effect did
n't happen; that nudge in propensity protected the brand's sales from the effect of competitor activity. Instead of the competitor's ad winning extra sales it merely got the brand back to where it had been before (see Danaher, Bonfrer & Dhar 2008).
Consider people who, after being exposed to a Coke ad, increase their propensity to buy Coke tomorrow from 1-in-300 to 2-in-300. This means they will now make a purchase every 150 days, rather than once every 300 days. So these extra purchases still take a long while to occur. This is the meaning of advertising's sales effects being spread thinly over time.
The reach of price promotions is very different. Their reach is very restricted as it's skewed to heavier, more regular buyers of a brand. Also, the sales effects of price promotions aren't spread out in time. In-store promotions only affect those people buying from the category in the week of the promotion. So the entire sales effect shows up in that week's sales figures71. That's why price promotions show clear effects: sales increase immediately when price promotions are turned on, and sales decrease immediately when they are turned off. 72 There might be a longer lasting effect if purchasing and using the discounted brand leads to enhanced mental availability, but the evidence is that longer term sales effects of price promotions are essentially zero (see Chapter 10). Advertising's effect on purchase probability occurs through the effect on memories, and memories have some ability to last.73 Whereas price discounts affect purchase probability because we like a better deal, if one of the brands that we use and notice is on sale then the chance leaps that we choose it over the other brands in our repertoire – but our purchase probabilities return to normal immediately afterwards. So the price discount only works while it is there.
That's great news for those seeking to measure the sales impact of price promotions (the full sales effect can be seen and measured in a week's sales figures), but it's sobering news for those seeking to measure the sales impact of advertising. It means you should never mix price promotion effects and advertising effects into a single econometric modelling attempt to quantify the relative sales effects of advertising and price changes.
However, the good news is that the sales response to advertising can be seen even when a market is stable and there is no aggregate level change in sales for the advertised brand. This requires what is known as 'single-source data': individual-level data that records continuously over time what each individual buys and what advertising they are exposed to. To explain the advantage of such data, consider how a direct marketer would assess the impact of a direct mail campaign to selected households in Chicago. Would they use national sales data? No. Would they use Chicago sales data? No. They would look at the buying patterns of the households they sent the letter to. No one expects sales to change in households that didn't get the letter. Also, to correctly estimate the real sales impact, the marketer would compare sales among these households against a benchmark, such as the sales of similar households who were deliberately not sent the letter. Single-source data lets us employ the same logical approach to assessing the impact of television, print and other advertising exposures (Kennedy, McDonald & Sharp 2008; McDonald & Sharp 2005).
Forty years of single-source-based analysis has delivered solid empirical evidence that advertising drives sales among those who are exposed to it (and that some advertisements are vastly better than others). These findings have held across a range of brands, categories, countries and data sets (i.e. McDonald 1969; Jones 1995a & b; Roberts 1994; Roberts 1996; Roberts 1998; Newstead et al 2009; Taylor et al 2009; Wood 2009). This is good news for those who advertise. It is also good news for advertisers who want to measure the sales effects of their advertising so that they can work out what creative executions and what media strategies work better.
In a review of his single-source data analysis of sales effective advertising, Professor John Philip Jones (1997) noted:
“The European styles of advertising ... understated, softness, quirkiness, indirectness, unusual visual effects, and bizarre humour ... are often surprisingly effective. ... The advertising shown to have greatest effect ... were certainly not hard-selling in the conventional sense ... they were likeable ... visual rather than verbal.”
Rather than examine in depth the findings of single-source research on advertising sales effects here is a very short summary of the important findings that are supported by findings from other reliable methodologies, such as experiments:
•advertising generates sales
•some advertising copy is much more sales effective than others; some barely works at all
•creative copy without a persuasive message can still be very sales effective
•media strategy that achieves greater reach is particularly effective; reach is more important than frequency of exposure; continuous advertising is more effective than bursts followed by long gaps, because it counteracts memory decay.
All of this fits with the negative binomial distribution of buying rates (see Chapter 4). It makes sense given the importance of reaching all category buyers (see natural monopoly law), with solid coverage over time (i.e. no long gaps).
Advertising works by reaching brains and nudging buying propensities. This mostly happens without us noticing; occasionally an ad produces the reaction 'I should buy that' but even this intention only weakly nudges our buying propensities, because we often forget or are deflected from our intentions. This is a reminder that advertising works through its effect on memories.
Without noticing (and processing) there is nothing
An advertisement cannot build memory structures if it is not processed; memory structures cannot generate sales if they are not associated with the brand that is being advertised. Most advertising exposures fail these two hurdles, so the money spent is wasted, or worse, the ad refreshes memories for competitor brands. As discussed in Chapter 1, less than 20% of television advertisements are both noticed and correctly branded (i.e. there is over 80% wastage). This is shocking; it's alarming enough to think that large corporations would have a special task force, perhaps within marketing, that was charged with eradicating this problem. This task force would conduct research and development, set guidelines and cultivate techniques for developing and evaluating advertising to ensure that all a brand's advertising passed these two hurdles. Yet the reality in most organisations is that these metrics are sometimes not even measured, no-one's job depends on them, and no-one is responsible for fixing the problem. Therefore, no serious research and development is allocated to the issue – think how different this would be if the issue were wastage in the factory or product failures.
It's true that advertising can work without us paying it much attention. We are able to notice things at a very low, even subconscious, level; for example, while in conversation at a crowded party we can still hear our name being called in the hubbub of background conversations. But just because part of our brain is monitoring our environment at a very low level, it doesn't mean that it lets this information impinge much on our long-term memories. It's better if advertising can generate more conscious attention and processing.
As people screen out so much advertising, the challenge is to get past the brain's screening mechanisms and to generate that little emotional reaction in the direction of acceptance: 'I will pay attention to this'. Therefore, the primary task of advertising agencies is to generate outstanding creative ideas that viewers will notice and will be willing to process over and over. This processing must be brand-centric; it must refresh the memory structures that relate to the brand. This is a difficult task, which is why most advertising fails. It's also why some of the new findings from neuroscience and psychology are so important – we need to understand how attention and memories work.
We now know that very little thinking, if any, can be described as purely rational. Emotion is a primary source of human motivation, and exerts substantial influence on attention, memory and behaviour; it is no wonder emotion is heavily used in advertising. For example, emotion can be seen in adver
tising when the audience sheds a tear watching an ad for cancer research, or laughs with the characters in a beer ad, or shows shock, fear and relief watching an insurance ad. People watch movies, listen to music and read books largely for an emotional ride. They enjoy gaining the same from advertising and when they get it they pay more attention.
Memory is everything
Apart from a very small amount of direct-response advertising (including online search engine advertising), advertising must work through memories. This is an uncontroversial statement, yet it is common for marketers and academics to forget the essential role of memory; instead, they think advertising works largely through persuasive arguments or by creating emotional feelings about a brand.
Memory is the link between an ad and brand choice. Even a frequently purchased item (e.g. margarine) is only bought, on average, eight times a year, and any single brand is only typically bought once or twice a year (Nielsen 2007). There can be many months between exposure to an ad and when the viewer is in a shopping situation with a relevant opportunity to recall brand memories (which are possibly influenced by the ad). To influence behaviour, advertising must work with people's memories.
Advertising builds mental availability
The dominant way that advertising works is by refreshing, and occasionally building, memory structures. These structures improve the chance of a brand being noticed and/or recalled in buying situations; this in turn increases the chance of a brand being bought. Memory structures that relate to a brand include what the brand does, what it looks like, where it is available, when and where it is consumed, by who and with whom. Memories are associations with cues that can bring a brand to mind. They can also remind consumers to take it out of their pantry and eat it.