Farm Road, Henley-on-Thames, Oxon RG9 1EJ, UK
Vol 19 No 2 (2000)
Tel: +44 (0)1491 411000 Fax: +44 (0)1491 571188
What Do Advertisements Really Do for brands? INTRODUCTION
What advertisements ‘really do for brands’ is to influence behaviour. Specifically, when a shopper sees our ads, this should improve the odds of our brand being chosen.
Before we can speculate about how these effects are produced, we need to know what they are. The better writers on this subject have analysed some scores or hundreds of brand histories. Prominent in this small group are Andrew Ehrenberg and his team, John Philip Jones, Andrew Roberts and the organisations at Information Resources Inc and Millward Brown. This paper introduces another collection, called here 113 Cases.
‘How’, or what goes on in the shopper’s head, is the subject of White’s (1999) review of how most practitioners think about advertising. Essential for those writing the advertisements, ‘how’ is not in the language of the accountant; it does not unlock budgets. Nor does it convince until the behavioural evidence is produced.
Another approach to the question is similar to, though wider than, media planning. It discusses the routes by which it has been suggested advertising can ultimately affect sales. These of course include direct effects on shoppers. They broaden out to the improvement of the results of other marketing activities (packaging, promotions, price), relations with retailers and consequent shelf space allocations and other store activity, the advertiser’s staff, City opinion and other media coverage and so on. There is no space to discuss these routes here.
Nor does this paper move outside fast-moving packaged goods, since this is where the large databases are. It first summarises how grocery shoppers’ purchase choices are related to the advertisements they have seen in traditional media. If we can agree on methods of analysis, and produce numerical generalisations about behaviour, we have helped marketing decision-makers and we have also established a sound basis for discussing ‘what ads really do’.
Now that generalisation has been mentioned, this too requires comment. It is of course the purpose of this paper to generalise. But over-generalisation is a constant danger. Great care must be taken that we do not give it priority over variety (see below).
In this paper, the primary source is 113 Cases. In addition, the conclusions are based on various single source analyses (Broadbent etal. 1997), on 20 years of modelling experience and on other publi cations, especially various papers by Ehrenberg et al. (1997), Hall (1998), Jones (1997) and Roberts (1999).
A familiar lesson, drawn again here, is about the variety of the sizes and shapes of the effects of advertising. Following Jeremy Bullmore, this paper is not titled: ‘What does advertising do?’ It does not conclude, for example, whether advertising is a ‘weak’ or a ‘strong’ force; some campaigns are weak and others strong.
No one asks, ‘How do vegetables grow best?’ You get one answer about a good baking potato, but it does not apply to asparagus. In the same way, do not expect the same answer about financial services and retailers, a new TV channel and an automobile.
There is variety in every collection of case histories. See, for example, Tellis (1988) on price elasticities, with a
reported range of at least 5:1. Assmus et al. (1984) have reported an even wider scatter for advertising elasticities; there the range is over 10:1.
FINDINGS FROM ‘113 CASES’
This section concentrates on how advertisements may affect the shopper’s choice of brands in the following ways:
Are there short-term effects? How short? How large? If not always, when do we most expect them? Are there diminishing returns? Are there longer term effects? What exactly does this mean? For how long should we expect effects to last? How big are they? When do we expect them?
The approach to these questions is econometric, and the following structure generally applies. For each of the five following paragraphs there exists a descriptive parameter.
1. Short-term ad effects have a half life (in weeks) determined by the brand situation and environment, and
by the copy used and choice of media. Adstock modelling is used to measure half life; conventionally, ‘short’ means under ten weeks.
2. These ad effects may be large, small or too small to be measured reliably. Their size is determined by the
brand situation and the short-term sales effectiveness of the copy.
3. Depending on the scheduling, shoppers overall may have recently seen a few or many ads. To have seen
many does not mean that the chance of buying the brand is a linear multiple of the chance if there were few: there are diminishing returns.
4. Adstock with a half life of ten weeks or longer may also help explain sales. In adstock terms these effects
are ‘long’, but with a wider perspective they are medium-term effects. The half life is again determined by the brand situation and environment, and by the copy used.
5. These ad effects may be large or small. Their size is determined by the brand situation and the medium-
Given a history of ad exposures to shoppers, these five parameters together determine the effects of the advertisements on brand choice for up to one or two years. These parameters vary across brands. As they change, the effect of advertisements on sales can take many different shapes, and can assume more or less importance.
For a particular brand and campaign, a single number can be estimated for each of the five dimensions in the majority of cases. It is important to know what fits in my case, with my brand situation and my copy. This is the way advertisers think, and they are right to do so. But there is no one way that advertising works. There is a single and flexible model. There is not a single set of parameters.
The data source was a meta-analysis of case histories from eight countries: Australia, Germany, Hong Kong, Italy, Japan, Thailand, the UK and the US. For laundry detergents, haircare and ready-to-eat cereals, the study was repeated in several countries. For the following thirteen categories, only one country was involved: baby wipes, canned beer, canned fish, carbonated soft drinks, cereal bars, cough and cold cures, nappies, digestive aids, fruit spreads, tampons, toaster pastries, waffles and yellow fats. In a couple of cases, the years covered began in 1994, but most began in 1996. The most recent data were for early 1999. A hundred and thirteen brands were selected for analysis. This is a convenience sample. It represents the categories and brands of most interest to Starcom Worldwide – the brands were either clients or their main competitors. The average brand
analysed had a volume share of 12 per cent, so the sample tended to consist of major grocery brands. Its price to the shopper averaged 13 per cent higher than the average for sales in its category. Individual brands varied considerably around these averages.
Advertising in these categories was mainly on television, in most cases overwhelmingly so. In some categories other media were significant; for a few brands, print or outdoor were the major media (radio was in some cases also studied, but had always been used less than TV or print). Where other media were clearly significant, they were treated exactly like television. TV language is used below. Some 3,600 target ratings a year, or 70 ratings per week, supported the average brand. This was 16 per cent of the total ratings in its category (share of voice).
Data organisation, from the usual sales and media data sources, was by local Starcom offices, in a common format suitable for the Battlefield programmes used later. This is both a management information system, designed on the basis of an earlier method (Broadbent, 1992), and is used to reorganise the variables, creating new ones like relative price and adstock (see Broadbent and Fry (1995) and earlier papers for details). The main use of Battlefield is to improve local understanding of the environment in which the advertised brand competes.
The key facts about each brand were recorded – whether it was new or established, its average sales share and relative price, and so on. A unique feature was that the local team answered a questionnaire about the marketing and advertising objectives, the category and media environment, how they expected communications to work and how effective they thought they were.
Advertising variables (some for several media) were turned into adstocks with a range of half lives. Volume shares were then modelled (in a few cases of brands with different seasonalities from their category, actual volumes were modelled). Most of these models included relative price and availability as explainers; in some cases, other specific factors entered, as did, on many occasions, competitors’ advertising.
Half lives were chosen for the adstocks giving the best fits and in the most influential media. This meant that either none of our adstocks was significant, or only one, or in some cases two (e.g. a short and a long half life, or a TV campaign and a print campaign). For short-term effects, diminishing returns were investigated.
The coefficient for the adstock(s) chosen describes the size of the ad effect. Elasticity, the percentage by which sales increased for 1 per cent more overall ratings, is:
Coefficient x average adstock/average sales
A single coefficient (hence a single elasticity) was estimated for each brand. This was because the object was to learn the average ad effect on the brand. Of course ad effects vary by campaign, and when we investigate it we need separate adstocks for each campaign, but this was not necessary here.
Before generalising, two examples are given. One reason is to demonstrate how much the sales shares of typical brands actually move. It is well known that, over weeks, price promotions cause big swings. Some of the literature gives the impression that over longer periods the surface is calm. Sales shares for the two brands in Figure 1 are therefore plotted over months. The second reason is to show that modelling was able to explain a significant share of the movement of the brands, as Figure 2 shows.
Brand A was the brand leader in a crowded category, with a high relative price, 36 per cent above the average. This meant it could afford to dominate advertising in the category, with 24 per cent share of voice. Although the largest brand, it has clearly been able to expand. The obvious questions are what parts price and advertising played in this growth, and in particular in the two sales peaks around months 16 and 22.
Brand B was a small brand in the same category, closer to average price and in another country. Though small, it was a serious advertiser, with a share of voice nearly double its market share. This time, the question is to find the reasons for its decline, which has not been steady, but in fits and starts.
The proportions of variation in sales explained were 92 per cent for Brand A, 83 per cent for Brand B. In both cases, price and advertising were strongly associated with sales, with display being an additional reason for
Three points have now been made: the odds of our brand being chosen can vary, both short and longer term; adstock modelling can explain most of this variation; we obtain estimates of the relevant parameters. More about these case histories is given in the Appendix.
General conclusions from ‘113 Cases’
There is an equivalent of Newton’s First Law: brand share continues in a steady state unless affected by marketing activities, either ours or competitors. We look for reasons for changes in share.
Our share of category sales generally increases when we have advertised recently. ‘Generally’ because there are exceptions where, if there was an effect, it was not measurable. This does not mean the campaign was necessarily ineffective, though this certainly happens. Sometimes, the data and situation prevent us from a significant measurement. One example is the typical launch, when distribution and introductory pricing may explain so much of the brand’s sales that the advertising, however effective, fails to appear in an econometric analysis. Other methods are needed to demonstrate its influence. A second example is the large brand with a steady sales share and continuous advertising. Regression depends on there being enough change in the variables to estimate effects – no change, no measurement.
What is ‘recently? How big was the effect? The average half life for short-term effects averaged three-and-a-half weeks. This is consistent with Roberts (1998), who uses a completely different technique and quotes a 16-day half life. ‘Interval bias’ (he uses days, and this data is for weeks) is enough to explain the difference. The average is also very close to the four weeks quoted by Hall (1998).
Since the famous paper by Clarke (1976), it has generally been known that such estimates depend on the length of the data interval. A correction for this has been published by Fry et al. (in press). This was used to standardise the findings in 113 Cases, where data intervals of four weeks and one month were common in the analysis, but unbiased estimates in weeks were needed.
If we say a half life of half a week or a week is very short, so that ‘recent’ means ‘largely over within a week’, then less than a quarter of all measured short-term effects have a very short half life (compare Jones, 1995a). In fact short-term half lives seem to be normally distributed round the mean, and are typically two, three or four weeks.
The size of the effect is traditionally expressed as the advertising elasticity. The average here was about 0.1. Since the break-even elasticity of the average brand is about 0.2, short-term effects are unlikely to pay off directly. The advertising elasticities have a J-shaped distribution, with a long tail of low effectiveness, and a few much more effective – the largest third average 0.2. For the middle third, the average is only 0.06.
Up to now, linear fits have been described (the difference from multiplicative or log fitting is not serious). An estimate of the steepness of diminishing returns is needed in scheduling decisions. For the shape of the convex curve used, see Broadbent and Segnit (1967). The parameter which describes its steepness is F, which here is allowed to vary between 20 and 80. F is the percentage of maximum response reached when adstock is at the level of 100 ratings (equivalent on average to one OTS) per week. The very steep response suggested by Jones (1995a) corresponds to F = 80.
This shape was found in about a quarter of the cases analysed. Another peak was at F = 20, which corresponds to very gentle convexity, not so far from linear, as reported by Jones (1995b). In between there were fewer cases, so the distribution of F was U-shaped.
Longer term effects
The half lives for cases with longer term adstocks averaged 23 weeks. This is not a precise estimate as it depends on the half lives tested. Compare Hall (1998) who quotes a half life of about a year, but for brands
which already had a particularly positive response. Their advertising elasticities averaged twice those of the short life effects, at 0.2. They again have a tail of lower effectiveness, and a few are much more effective – the largest third average 0.4. For the middle third, the average is only 0.13. Hall (1998) quotes a baseline effect of 4–8 per cent for every 4 per cent short-term contribution. Millward Brown have suggested higher figures in some cases.
It is common ground that longer term effects are the main reason for advertising. Indeed, the above figures suggest they may on average be responsible for two-thirds of the commercial return in the medium term.
The industry has considerable difficulty in defining, let alone measuring, long term effects. In this part of the paper, the definition is simple: the effects are proportional to an adstock with a half life of ten weeks or over. The shape of an adstock variable is always the same, only the rate of decay varies. That is, the effect is highest immediately after the ad has been seen, and then declines exponentially.
Some say that long-term effects do not exist and mock the ‘time-bomb’ theory, others that there can be no long-term effects without short-term ones. From the definition in the previous paragraph it can be seen that there is no question of an adstock effect being zero in the week just after an OTS, but having a high effect per week several weeks later. It is however true that with a long half life, the proportion of the total effect in the first week or two is relatively small. It may be so small as to be statistically undetectable, while the larger, longer term results may be measurable.
There is therefore no time-bomb in adstock modelling, no longer term effects without immediate results. But there can be a significant long half life adstock term in the fit, without there being necessarily a short-term one, and vice versa.
Adstock half lives form a continuum. The differences are due to persistence of copy effects in the memory, plus the ideas and habits about the brand which they have stimulated. In many cases, only one adstock is chosen in the fit, which may be an average of a short and medium term effect; in others, two adstocks fit better, so the effects (or people affected) are of two sorts. For some, the ad may be just a reminder; for others, more persists. In Section 4, ‘more’ is re-examined.
Learnings from the questionnaire
This feature was included to help Starcom use the details of later cases to decide appropriate schedules. It also indicates when we most expect various outcomes in the marketplace.
Cases with no apparent ad effect are often those in ‘stalemate’ categories where advertising is the price of entry and competitors cancel out, but no one dares withdraw. They also have maintenance as the objective and recent support from other activities has been normal, as opposed to low or no support. They tend to be brands with higher ratings per week but against high advertising competition.
We particularly expect to see short-term ad effects when share of voice is high and communication is easy, and when the purpose of advertising is to reassure and reinforce. We expect it to be over particularly quickly if it is repeating a known message, with few other marketing activities to help it, against heavy opposition. Less repetition of an ad is required per week when the brand has a large share or relatively low price, or the copy calls for immediate sales. More is needed for a launch.
We see long-term ad effects when we want to launch or establish the brand against competitors to whom shoppers feel low loyalty, and in a high involvement category. The purpose is to change people (which is likely to be a slow job). If a longer term effect is expected, an ad which is to reinforce attitudes, with a simple message, in a high involvement category and with help from other marketing activities, has the longest lasting effects.
How campaigns affected shoppers’ choices in 113 Cases has now been described. Next, data from other sources are also used to generalise on the ‘what’ question: In what way do advertisements affect overall
There follow three sections on the ‘how’ question, about the likely mechanisms through which advertisements cause these effects.
Effects on behaviour
Campaigns differ in what they do for brands. The reasons for variety are the categories in which brands compete (history, weight of advertising, substitutability of products, habits of shoppers), the situations individual brands are in, their competitive environments and other activities used, the intentions of the advertisers and the effectiveness of the copy. For the majority of campaigns we can estimate the advertising effects. In these individual cases, with qualitative information, we can reach a view about how the advertisements worked.
What generalisations are possible? Certainly not ‘Advertising always works this way’. But we can define a structure or shape of a model that can often be fitted. A set of parameters determines how the model looks in any particular case, and therefore what the effects on brand choice were in that case.
The distribution of the parameters can be ascertained from any collection of case histories. These distributions depend on the case histories chosen – the countries, categories and dates, for example: Russia in an economic crisis, the US in steady growth, a routine purchase in a category without excitement, a deliberated and costly decision – who expects uniformity?
Generalising about groceries in a sound Western economy, the evidence suggests that many campaigns have both a short and a long-term effect on brand choice; some have one but not the other. A small minority have no measurable effect. Short-term effects average a three- or four-week half life, with a normal distribution about this mean. The size of the effect is J-shaped, with some large effects and many small ones. Diminishing returns are normal, with a U-shaped distribution – some very steeply convex, some near-linear. On average, the investment does not pay back in this way.
Medium term effects (adstock with a half life of ten or more weeks) average a half life of about six months, so there are still effects more than a year later. Again, the distribution of size is J-shaped.
All these effects are obscured in raw data. There are stronger forces at work: product experience, product improvements, price skirmishes, promotions, launches and new variants, retailer decisions, erosion. On the latter, Ehrenberg et al. (1997) correctly comment, ‘Without some share of competitive marketing inputs, the brand’s long run share of market will be expected to fall.’
There is also a possible long-term effect, part of what Jones (1995a) calls the brand’s ‘internal momentum’. Brand size does not explain everything. The simplest counter-example is price, and the easiest demonstration is the scatter plot for brand averages of volume share and relative price. We often see on this plot two brands of similar size (volume share), but one is at a higher price than the other. Or, at similar prices, but one with higher share.
Figure 3 is an example of data from 113 Cases; it shows the most re cent year for the main brands; Brand A was in the examples given above.
The brand leader (A) is neither the cheapest nor the most expensive; its relative price is average for the brands included (other cheap brands and Own Label brought the overall average down to 100). In fact price seems to have little to do with share across brands (within brands, normal price effects are clear). Though the very expensive brands M and N have very low sales, for the others the relationship seems counter-intuitive – the more expensive, the higher the sales. How can this be, unless brands like A, B, C and D have higher perceived values? Where do these values come from? We return to this in the final section.
What sort of effects when?
The conditions under which all these possibilities are likely to occur were indicated in the section on the
questionnaire above. These were not given in full, nor are they repeated here. Other analyses are starting to look at this question (Hall, 1998).
Jones (1997) is right to point out that short-term effects certainly exist. They are ‘blips’, not adequately discussed in either the ‘strong’ or ‘weak’ theories. ‘Nudge’ does not seem sufficient to describe them, though ‘Here I am’ is part of the mechanism. There can be more information content than this: ‘Here’s a promotion’, or, ‘Remember, I can help with this’. After most blips, things return to normal, but occasionally the brand has been pushed up to a new level.
Very short half lives are more likely when the purpose is to reassure, not to change, and there is no new news. Diminishing returns are steeper with long ads calling for immediate sales. Adstocks for medium half lives decay especially slowly when the purpose is reinforcement of a simple message.
New, small and weak brands are more likely, if the advertising is suitable, to instruct and change shoppers. New brands and line extensions are more responsive. ‘News’ is an ingredient which usually makes effects more likely; hence the announcement of other marketing activities in advertising is recommended, and a discernible and relevant change in brand strategy can help.
Communications for large brands are more likely to remind and reassure shoppers. Campaigns with longer half lives are likely to have higher ad elasticities and, although ‘weak’ in current parlance, may be giving powerful support to their brands.
Campaigns with a short half life and high ad elasticity (sales effectiveness) for these brands will appear to be ‘strong’ advertising. Ehrenberg et al. (1997) ridicule the ‘strong’ theory and ‘brand-building’ when they are taken to mean that advertising on its own can increase the shares of all the brands in a category. They are wrong to suggest that a brand cannot grow when supported by advertising (Brand A was a counter-example), but this nearly always happens when other activities share in this task – product improvement and new variants, new uses, new sales outlets, reductions in price, and so on. The real culprits are those who suggest that advertising alone can always raise sales. Brand-building is best interpreted as the relationship between shopper and brand being kept in good repair – it is building in the sense of maintenance, not ‘new-build’. A major method is ‘framing’, when the ads call attention to brand features that are confirmed in product use.
So far, the discussion is like supply-side economics and has concentrated on production: what advertisers do and what they expect. This is unbalanced. Demand matters at least as much and a people-centred view is in the end the only explanation of ‘how’.
There may be some, even in 2000, who hold that a human being, in all her complexity and with all her other concerns, is impelled by advertisements down tracks like a tram. Ignorance of a brand, attention, interest, learning, feeling, desire, conviction, and so on are actually not stations in a one-way progress. They may be states we can recognise, but they recur, more or less strongly, as the shopper leads her life according to her own priorities. Each time she is in one of these stages, she brings new experiences and perhaps different motivations. Punctuated by the ‘empty box’ and brand choices, the sequence is rarely the same. Hence Huey (1999) recognises we should ‘examine advertising effects in a new way, from a non-linear, multidimensional perspective … Persuading consumers (and keeping them persuaded) is a continuous process… and advertising plays a continuous role in that process.’
What limit can possibly be set on how people vary in these regards, and what measures should we use to learn about the course of thought and action? The states are clear enough (ignorance, etc. as above); speculation about the process is easy, but conflicting overall accounts of it are hard to choose between. For a single category, aunique individual, one brand choice – a description is feasible. But togeneralise? – we are back to variety in an even more diverse population.
A key criterion about people and brands, though ultimately circular (again, see the final section), is the shopper’s loyalty. What proportion of her purchases in the category does our brand satisfy? How does it stand in her repertoire? Those who have never bought our brand are much less likely to be affected by advertising. Brands with many occasional buyers already (low loyalty) are more likely to be responsive.
Next in importance is the shopper’s weight of exposure to communications in our category. When light viewers are exposed, they respond to advertising more than do heavy viewers (they view differently, they see fewer competitive ads).
Thus, the shopper’s chances of being affected by our advertisements can depend on demographics, weight of purchase in the category, previous product experience and attitudes and beliefs (circularity again). There is probably also a scale for ‘openness to persuasion’, about which we know little.
We can make sense of data like those in Figure 3 (even allowing for differing distribution) only by assuming that the perceived benefits of brands like A, B, C and D balance or outweigh their costs. Such relationships can be quantified as Consumer Brand Equity (Broadbent, 1997). The causes will include real differences in product quality, history, habit, and so on – but past campaigns may also play a part. Sometimes, when the other factors have been steady, we can see this campaign effect in simultaneous changes in equity and in long-term advertising weight.
Individuals strike their own balance between cost and benefits. Each shopper has her particular way of reducing the dissonance between her limited budget, family requirements, product experience and associations, recent advertising, and so on. These preferences are not as rational as some seem to believe. Shoppers create brand personalities to simplify and populate their worlds. This is a proper study for anthropologists, as well as for market researchers, as White (1999) points out. Traditional attribute scales show the differentiation, as do more subtle and specific questions and focus groups. The reasons are well explained by Lannon (1999).
We meet this construct of experience and beliefs under various names: brand personality is only one. ‘Genuine’ or rational brand values come into it; so do history and habits, loyalties, attitudes, emotional and ‘perceived’ values. Internal momentum, Consumer Brand Equity and ‘branding’ attempt to describe it.
Is it possible that there is a simple split of this Gestalt into rational and emotional causes? Can we dissect it – so much due to product, so much to advertisements, so much to other marketing activities, etc? It seems unlikely. It is like asking which does the thinking – the brain or the mind. Branding is still a conundrum and it is only honest to admit this.
Over the long term advertising can increase ‘brand equity’ i.e. the brand’s ability to support a high market share, a premium price and profitable brand extensions… However it has not been possible (and probably never will be) to pin down the precise contribution to brand equity and long term profitability of advertising as a separate factor.
Yet, for many manufacturers, brand equity is their most valuable and potentially longest lasting property. Many advertisers and agencies see its creation, nurture and defence as their main task.
What this paper shows is that we are increasingly well informed about advertising effects over weeks and longer. In many individual cases, we can explain behaviour and also agree how shoppers used the advertisements. We are beginning to generalise about what advertise ments do for a brand over a year or two. We know advertisements contribute to branding and brand equity. We have a measure for this concept; we can list some of its components. It is not simply the sum of these parts, it is the joint result of all of them. We do not know exactly how to dissect it. The really long-term effects of advertisements still elude us.
ACKNOWLEDGEMENT 113 Cases was designed with and commissioned by Starcom Worldwide, Chicago. I gratefully acknowledge their permission to use some of the findings, and also comments by Jayne Spittler and Kate Lynch.
Assmus, G., Farley, J.U. & Lehmann, D.R. (1984) ‘How advertising affects sales: the meta-analysis of econometric results’, Journal of Marketing Research, February, 65–74.
Barwise, P. (ed.) (1999) Advertising in a Recession. London: London Business School.
Broadbent, S. & Segnit, S. (1967) ‘Response functions in media planning’, in Ten Years of Advertising Media Research, Thomson Medal Papers, 187–238. The convex shape has often been confirmed and it has not been necessary to change the shape of the fitted function since. See e.g., J.L. Simon & J. Arndt (1980) ‘The shape of the advertising response function’, Journal of Advertising Research, August, 11–28.
Broadbent, S. (1992) ‘Using data better – a new approach to sales analysis’, Admap (January), 48–54.
Broadbent, S. & Fry, T. (1995) ‘Adstock modelling for the longer term’, Journal of the Market Research Society (October), 385–403, and earlier papers.
Broadbent, S., Lynch, K. & Spittler, J. (1997) ‘Building better schedules – new light from the single source’, Journal of Advertising Research (July/August), 27–31.
Broadbent, S. (1997a) ‘Single source – new analyses’, Journal of the Market Research Society (April), 363–379.
Broadbent, S. (1997b) Accountable Advertising. Henley-on-Thames: Admap Publications.
Clarke, D. (1976) ‘Econometric measurement of the duration of advertising effect on sales’, Journal of Marketing Research (September), 345–57.
Ehrenberg, A., Barnard, N. & Scriven, J. (1997) ‘Justifying our advertising budgets’, Admap Conference, December.
Fry, T., Broadbent, S. & Dixon, J. (in press) ‘Estimating advertising half life and the data interval bias’, Journal of Targeting, Measurement and Analysis.
Hall, J. (1998) ‘How advertising works 2’, European Advertising Effectiveness Symposium, Hamburg.
Huey, W. (1999) ‘Advertising’s double helix: a proposed new process model’, Journal of Advertising Research (May–June), 43–51.
Jones, J.P. (1995a) When Ads Work. New York: Lexington Books. Very short half life and steep diminishing returns are suggested here, and are quoted by Erwin Ephron as support for the Recency theory of Scheduling, but are controversial – see S. Broadbent & E. Ephron (1999) ‘Two views of TV scheduling – how far apart?’, Admap (January), 22–25.
Jones, J.P. (1995b) So wirkt Werbung in Deutschland. GWA-Publikation.
Jones, J.P. (1997) ‘The essential role of communications’, Admap Conference, December.
Jones, J.P. (ed.) (1999) How to Use Advertising to Build Strong Brands. London: Sage Publications.
Lannon, J. (1999) ‘Brands and their symbols’, in Jones (1999).
Roberts, A. (1999) ‘Recency, frequency and the duration of the sales effects of TV advertising’, TSMS & Taylor Nelson Sofres. See also ‘Measuring the short term sales effects of TV advertising’, Admap (April), 50–52.
Tellis, R.J. (1988) ‘The price elasticity of selective demand: a meta-analysis of econometric models of sales’, Journal of Marketing Research (November), 331–341. This summarises a collection of 367 price elasticities, with many comments on methods, etc.
White, R. (1999) ‘What can advertising do for brands?’ International Journal of Advertising, 18(1), 3–17.
APPENDIX: MORE ABOUT THE CASE HISTORIES
A steady reduction in relative price is the main reason for overall growth for Brand A. Early in these two-and-a-half years it cost 50 per cent above the average; towards the end this had fallen to about 20 per cent. We estimate this added 50 to the sales index, probably not a profitable move since the price elasticity was only –1.4. The strategic benefits are obvious however.
Price was not the only major factor. A long-term (13-week half life) TV advertising effect was measured, with one of the highest advertising elasticities in this study. If this had not been present, the fall in the sales index is estimated at 35.
There was also a similar, but larger, effect from the total of competitors’ advertising. This however fell during months 12 to 20 to about two-thirds its level during earlier and later months. This is one of the reasons for a rise in Brand A’s sales at this time.
There remain some isolated peaks to explain, in months 4, 15 and 16, 22 to 24 and 29. It is no surprise that these corresponded with particularly low prices (month 29 was the minimum). But some of these were accompanied by print advertising, and this is estimated to have a short-term effect, with a half life of two weeks. This has added up to 10 to the share index.
Here is a typical story. Management has decided to gain share by reducing the relative price level, with some isolated price promotions. Some of these promotions have been supported with print advertising, increasing their effect. Meanwhile the brand is defended by rather continuous, brand-building, television advertising. This is not planned for immediate effect, but it has made a major contribution to the sales level. Competitors’ advertising also had long term effects, continuously threatening the brand with a decline which would have wiped out all their gains.
The decline of Brand B can be traced to two causes, both of about equal size. Its advertising weight nearly halved during the two-and-a-half years. The adstock fit with a 26-week half life tells us that the support which this provided in the early months was worth 30 to the share index. The fall in advertising pressure at the end cost the brand about 15 on this index.
The brand lost almost all its display over the same period, and this meant another fall of 15. The overall drop was not as bad as the total of 30, because management tried to maintain share by price reductions.
The price elasticity was very small (–1.1 was estimated) so the effect was minimal (it must also have affected profit badly). There were isolated price promotions (particularly in months 21 and 25, the latter well supported with display) whose effect can be seen in the plot.
In this case management may not have been aware of the long-term effects of advertising: no short-term effects could be seen, though the ad elasticity for the long-term effects is in fact good. They may also not have assessed the poor return to price cuts. They would otherwise hardly have led the brand downhill in this way. The loss of display must have been obvious. The attempts to stabilise the brand over the past year have had some success.
NOTES & EXHIBITS Simon Broadbent
Simon Broadbent is director of BrandCon Limited, a brand consultancy. His books include ‘The Advertising Budget’ and ‘Accountable Advertising’. His work has
influenced ad thinking over three decades. FIGURE 1: TWO BRANDS' VOLUME SHARES FIGURE 2: FITS TO VOLUME SHARE FIGURE 3: EXAMPLES OF BRAND POSITIONS: PRICE AND SALES
Steinzeitmensch beim Honigsammeln, Cueva de la Arana bei Bicorp, Valencia, 4000 v.Chr. Aus: Hans Baumann: Die Höhlen der grossen Jäger, K.Thienemanns Verlag, Stuttgart 1972. Honig war die einzige dem Steinzeit-Menschen zugängliche wirkliche Süssigkeit. "Ich nehme als die 6te Ursach an die Schwelgerei, welche zwar keine epidemischen Krankheiten verursacht, aber zu allen Zeiten und an alle
Tappa 4 - Da Granges di Pragelato a Balziglia (15,8 km) Si parte dall’Hotel Passet di Pragelato in direzione dei Trampolini del Salto, lungo la strada asfaltata, fino al ponte. Da qui si imbocca la sterrata che corre lungo la sinistra orografica del Chisonetto e la si percorre fino ad arrivare al ponte successivo, costruito per le Olimpiadi Invernali del 2006. Hanno inizio i campi da golf