The Ultimate Secrets of Advertising


John Philip Jones

  • Citations
  • Add to My List
  • Text Size

  • Chapters
  • Front Matter
  • Back Matter
  • Subject Index
  • Copyright

    View Copyright Page


    For W.M.H.J.


    I learnt that what emerged from a dispassionate scientific analysis of the confused events of war counted for no more in the debates in which I became engaged than the in-built convictions of various members of what were called the intelligence and planning communities.

    SollyZuckerman, 1978
    (from his book From Apes to Warlords)

    Zuckerman—empiricist and iconoclast—refers to a conflict that characterizes all fields of activity: the conflict between unsupported convictions and disinterested facts that contradict those convictions. Our knowledge of how advertising works has until now been opaque, incomplete, and variable in quality. And although some influential analysts remain bemused by the difficulties, generations of practitioners and academics have developed unshakable (and often contradictory) views on the subject. This is a problem that is addressed by this book.

    List of Tables and Figures


    Foreword: Red Threads

    This ambitiously titled book tries to answer four questions:

    • Does advertising work?
    • How does advertising work?
    • How much advertising works?
    • How can advertising be measured and made accountable?

    The last question effectively sums up the rest, and it can only be answered by our finding a way of calculating the financial return from advertising during a finite period (factoring out all other influences on the sales of a brand) and comparing this return to the dollar cost of the investment. Readers will be baptized in Chapter 7 with the method I recommend to do this difficult job, and Chapter 10 gives details of advertising accountability as it relates to 17 real brands.

    The responsibility for the effectiveness of advertising is divided more-or-less equally between client and agency. Advertising cannot be successful unless the client markets a brand that will satisfy the consumer (normally well enough to encourage repeat purchase), and the brand must be sold at an acceptable price and be distributed widely throughout the retail trade. Besides these responsibilities, the client is an active partner in all matters concerning the advertising itself and takes the lead in determining the budget. The agency is totally committed to the advertising task. It writes the campaigns, assesses and endorses the budget, and plans the media (although this particular function is increasingly being done by outside organizations).

    This book does not focus on how to write effective campaigns, nor how to arrive at the optimum advertising budget, nor how to phase this budget over time in the best media. These three matters—creative, budget, and media—are the most important operational concerns of advertising. I am, however, convinced that it is totally futile to plan such crucial contributions to the enterprise unless we are informed by knowledge of advertising effects. How can we talk about creative ideas without some degree of professional knowledge of the wide range of creative possibilities, and in particular, how specific creative ideas have performed in the real world? How can we propose an advertising budget without knowing about the effectiveness of similar (and also different) budgets? And how can we plan the details of media without knowing what specific media and media deployments have accomplished in the past?

    My relatively brief discussions of the creative process, budgets, and media are therefore no more than extensions of this book's agenda, which is devoted essentially to the measurement of effects.

    The book contains a good deal of information that has never before been published. This is important in itself. However, even more important is the fact that a good deal of time and effort has gone into analyzing and synthesizing—and generally thinking about—the data. Inhabitants of the world of information believe that everyone has an insatiable demand for facts. We are, in reality, drowning in them.1 It is already exceptionally—and increasingly—difficult to select the relatively few facts that are important and to work out how to use them to our advantage. This is what I attempt to do in this book.

    This volume is empirical, and the theory that it postulates is based on facts. The framework of how advertising works is described in Chapter 1 as the Gate-keeper model. Parts of this are repeated in subsequent chapters, where the elements of the model described in these extracts are examined with the use of marketplace data (and, in my judgment, confirmed).

    The book is organized in a simple way. Chapters 1 through 5 are concerned with short-term and medium-term effects. Chapters 6 through 10 are devoted to advertising's long-term effects in all their complexity. The book concludes with two appendixes. Appendix A describes a wide range of tracking techniques by which the progress of advertising campaigns can be studied over time. The main part of this book does not study continuous advertising effects (for reasons given in Chapter 10). But tracking studies can represent an important way of looking at how the contribution of advertising develops, and readers should be made aware of their potential value. Appendix B is devoted to two supplementary methods of measuring the long-term effects of advertising. These are more theoretical than practical. However, given their potential importance, I believe that they should be used experimentally, and it is possible thereby that we will eventually be able to put flesh on their theoretical bones.

    In Northern Europe, the local languages use a valuable metaphor, which can be translated into English as a red thread. It brings to mind a fabric, in which a red thread is a visible but not overly visible element of the weave. A red thread is used here to describe an underlying theme in a piece of writing—something that is constantly but not too aggressively evident. It is different from the structure or organization of a text, its basic blueprint. The structure of this book has just been described, but the writing itself contains a biblical total of seven red threads. These more than anything else reveal the underlying concepts to which the book is devoted. All parts of the book bear on them.

    1. Behavioral Effects

    Advertising works by influencing consumer behavior, and it does this in many ways. It can boost sales of a brand, thus increasing share of market in all circumstances except when the category is growing faster than the brand itself. It can defend a brand's sales. It can decelerate sales decline. It can increase penetration or purchase frequency or both. It can maintain sales in the event of a price increase.

    Cognitive effects, which relate to consumers' awareness of brands, brand at-tributes and advertising campaigns, and also attitudes toward brands, are not much discussed here. Nor is the moot point of whether cognitive effects influence consumer behavior, or whether they are in turn influenced by it.

    It is only by studying behavioral (and especially sales-related) effects that we can get close to the major topic addressed by this book, which is advertising accountability, the comparison of dollar return against dollar investment.

    2. Advertising's Patchy but Multifaceted Effectiveness

    There is not a scintilla of doubt that advertising can work and can influence consumer behavior. This does not mean that all advertising works. In fact, effective advertising represents no more than a substantial minority of all the advertising that is exposed.

    How advertising works varies widely. Among the more important ways in which it can operate are (a) to build penetration (to bring in new users) or increase purchase frequency (to boost business from existing users); (b) to operate directly to stimulate brand trial by new users or work indirectly to reinforce existing users' inclination to buy the advertised brand repeatedly; and (c) to emphasize a brand's functional features or its added values—values that exist in the psyche of the consumer.

    3. Three Orders of Advertising Effect

    Advertising can produce a short-term effect, felt generally within 7 days of its appearance; a medium-term effect, measured over the course of a year; and long-term effects (in the plural), also measured over a year but representing effects that have built up over a number of years before this. The three effects are sequential, and each acts as gatekeeper to the one that follows. Most important, a short-term effect is a precondition for all other effects.

    The shorter the period measured, the more the influence of advertising is felt on its own as an isolated stimulus. The longer the period, the more that advertising works in cooperation with other stimuli influencing sales.

    Although this book is empirical, it aims to formulate an inductive (i.e., fact-based) theory. This theory aims to be general rather than specific, which means that all important aspects of advertising are covered by it. In a full sense, it tries to solve the very difficult problem of how advertising works.

    4. Scale Economies

    Large brands are generally strong brands (with few exceptions). Small brands are generally new or weak brands (with not quite such few exceptions). Large brands benefit from scale economies in all aspects of their manufacture and marketing.

    Advertising-related scale economies are of cardinal importance. The best way to measure the long-term effects of advertising is via an advertising-related measure of these scale economies. The reasoning behind this is that large brands are indeed those that have benefited most from the long-term effects of previous advertising. These effects operate in conjunction with consumers' satisfaction with the functional delivery of these brands in a process of continuous mutual reinforcement.

    5. Diminishing Returns

    Surprisingly, incremental advertising pressure almost always yields diminishing returns when we measure carefully the sales that result progressively from increments of advertising. This is a short-term outcome, and it sets limits on how much we should concentrate advertising pressure, a matter of great importance in determining a brand's media strategy.

    It seems paradoxical that although advertising works according to diminishing returns in the short term it can yield significant scale economies in the long term. This contradiction is best solved by focusing on the overall strength of the brand itself. Consumers will prefer a large brand over a small one on the basis of its performance. A large brand is likely to have a richer battery of added values in the minds of users. A large brand is probably in fuller retail distribution and carries greater influence with the retail trade. For those and other reasons, the overall response to advertising is likely to be greater for a large brand than for a small one. But even so, the pattern of incremental response for a large brand will still follow diminishing returns, as it does for a small brand. But for a large brand, the pattern of diminishing returns operates at a high absolute level; in contrast, for a small brand, it operates at a low absolute level. (This point is demonstrated diagrammatically in Appendix B, Figure B.5.)

    6. A Snapshot of Accountability

    To establish the measure of accountability for any advertising campaign, we must pick a finite period during which the dollar input and dollar output are measured. This process resembles taking a snapshot—a picture of advertising effect frozen in time. This is the reason why tracking studies—which are continuous and open-ended (like movies)—are not featured much in this book. The measurement of effectiveness in the short term, medium term, and long term is in every case discrete.

    7. The Analytical Tools

    This book is based on research that is state of the art. This includes pure single-source research; consumer panel research; a predictive technique to pretest television commercials; econometric evaluation; and studies of market and brand aggregates.

    Although the data in the book are invariably based on sound samples and proven measurement techniques, I have tried very hard to explain them with the use of the simplest mathematics and (I hope) the clearest prose. I am not a supporter of the view that complicated research requires complicated exposition—in fact, just the opposite: Difficult concepts absolutely demand that they should be explained carefully and sequentially and in simple language. I have always been struck by the powerful image conjured up by Anatole France, who claimed that lucid prose is like white light; it is self-evidently simple, but its simplicity conceals great complexity.

    What I find most frustrating about advertising evaluation is not the difficulty of the task, great though this is. It is that the plentiful discussion and controversy surrounding it are carried out exclusively among clever and well-informed technicians rather than by advertising decision makers. As a simple technician myself, I am happy in the company of my peers. But technicians are not the people who are in a position to implement the things I write about. I believe that people who are senior enough to make the decisions should be as passionately interested as I am in advertising evaluation in all its aspects. After all, it is their money that funds effective—or, more commonly, ineffective—advertising. But the issues have regrettably not penetrated either because they have not been presented with enough clarity or force or because senior decision makers have too many other priorities to spend much time engaged in discussions about advertising.

    There is nothing new about this, and I have on many occasions addressed the upper echelons of management in large marketing companies on this topic.2

    I have not achieved much success, but readers of this book can rest assured that I will not give up my efforts. Advertising evaluation really matters. And it matters most to advertisers who are interested in the profits earned by their brands. More accurate evaluation is the starting point to generating more effective advertising, and this is a guaranteed system for reducing the plentiful amount of waste. The result will, of course, be an inevitable increase in the manufacturer's profit. And profit is surely the most direct measure of the overall success of any business.


    1. See Anonymous, “Quantifying Information: Byte Counters,” The Economist, October 21, 2000, 96.

    2. See John Philip Jones, “Advertising: The Cinderella of Business,” Market Leader: The Journal of the Marketing Society, Summer 2000, 20–25. Also see John Philip Jones, “The Mismanagement of Advertising,” Harvard Business Review, January/February 2000, 2–3.


    My first thanks go, as always, to my wife, Wendy, to whom this book is dedicated. She has supervised the enterprise from first to last and carried out all the complex administration. She is my best (and toughest) critic. Not least, she has prepared an immaculate typescript for the publisher. She is one of the few people who can translate my handwriting into meaningful prose, and my multiple manuscript amendments are generally even more difficult to construe than my first drafts.

    I am also extremely grateful to Media Marketing Assessment (MMA), which has made available to me a large battery of its most valuable data on the econometric analysis of advertising effects. This has formed the foundation for Chapters 6 through 10. Gerry Pollak has been the “point man” for MMA, and he not only presented me with large amounts of information in a form that I could understand completely, but he also took great pains to check all my calculations and test my interpretations of the figures. He was closely supported within MMA by two other experienced and talented analysts, its president Sunny Garga and its senior vice president Bob Wyman.

    Another research organization, one with which I have worked extensively in the past, rsc THE QUALITY MEASUREMENT COMPANY, provided me with information of great depth and importance from which I constructed Chapter 3. They also deserve my gratitude.

    I am equally grateful to the Institute of Practitioners in Advertising, the organization representing the advertising agencies in Great Britain, for giving me access to its incomparable collection of more than 600 rigorously evaluated case studies. I have selected from these the examples of tracking studies—all important cases of best professional practice—that appear in Appendix A.

    The manuscript has been seen in whole or in part by a number of friends and professional associates, all experienced advertising practitioners and/or academics. In North America, these include George Black, Meg Blair, Erwin Ephron, Gary Gray, Allan Kuse, and Pam Shoemaker; and in Europe, Paul Feldwick, Robert Heath, and Nick Phillips. Despite the perceptive and constructive observations of these people, I must take the responsibility for all errors of commission and omission in this book.

    Scott Bunting, of Industrial Color Labs in Syracuse, New York, produced all the superb computer-generated statistical diagrams not only on time but with his customary accuracy.

  • Appendix A: Tracking Studies

    The traditional way to examine the effects of advertising is to follow continuously the progress of a brand's sales alongside different variables that are directly or indirectly connected with them (e.g., advertising, consumer price, retail distribution, and seasonality). This tracking is carried out using techniques that vary widely in their sophistication and purity (i.e., their ability to isolate the influence of advertising from the many other stimuli that affect the sales of a brand).

    The marketing literature is full of tracking studies, and this appendix is devoted to examining their range. It features exclusively examples taken from a single collection, one that represents the most solid corpus of case-study material on advertising available anywhere. This body of cases has been built up over the past two decades by the Institute of Practitioners in Advertising (IPA) from studies submitted by the British marketing and advertising communities for the annual IPA Advertising Effectiveness Awards.

    More than 600 cases are now available in electronic form,1 and the best of these have been brought together and published in 11 substantial hardback volumes (a new one comes out every second year). In view of the range and analytical rigor of the IPA collection, it is difficult to understand why these British cases are not more widely studied in the United States.

    The 21 examples featured in this appendix were chosen from the IPA cases published since 1990. All 21 were chosen as being typical, although the number of different tracking studies that can be carried out is very large indeed. My selections are illustrations of the techniques, and the facts of the individual cases are touched on only briefly.

    There are two basic types of tracking study. First, what I call simple studies show the progress over time of a single variable or a number of separate (generally unconnected) variables affecting a brand. Second, complex or multivariate studies examine a number of separate influences on a brand's marketplace performance, but they aim to compare the effects of these influences and to evaluate their relative importance. This is a difficult procedure, but it can enable the role of advertising, in particular, to be isolated. The technique requires high-order mathematics, and the most common method, multivariate regression, is what gave this type of tracking its name.

    Virtually all the IPA studies feature simple tracking. A substantial minority also use multivariate tracking, and these are of particular interest because they represent analyses at the state of the art.

    Simple Studies

    When advertising is successful and succeeds by working directly, it is generally possible to detect a close relationship between the advertising and the consumer action that it prompts. Figure A.1 shows the progress of a campaign to persuade the Scottish public to stop smoking. The television campaign is indicated by the vertical bars, and the numbers of telephone calls requesting help to quit smoking are shown by the continuous line. The connection between the two variables is not invariably precise, but the cause-and-effect relationship is quite clear.2

    In this diagram (as in most of the others in this appendix), the two variables are totally different in nature from one another, but the scales used for comparing them are chosen to emphasize the tightness of the fit between the two. This is an arbitrary but generally legitimate procedure, although on occasion it can exaggerate the closeness of the relationship and therefore mislead by implying visually a stronger connection between cause and effect than actually exists.

    Figure A.1. Health Education Board for Scotland Antismoking Television Advertising and Television Inquiries

    Figure A.2 introduces two variables in addition to advertising: sales tracked for heavy and light television viewers separately. The brand in question is Nescafé Gold Blend, whose campaign used a series of commercials featuring a man and a woman who became romantically involved when one of them borrowed a jar of Gold Blend from the other (a campaign that was also used by Nestlé in the United States). The objective of the advertising was to stimulate interest in the stories in the commercials with the aim of boosting interest in the brand's starring role. The heavier television viewers, who would have seen more of the advertisements, would therefore have been likely to buy more Gold Blend than light viewers would. Figure A.2 shows this hypothesis to be valid.3

    Figures A.1 and A.2 have described ongoing campaigns. A close relationship between advertising and sales can often also be established in successful new brand introductions, but it should also be remembered that these are rather rare. Figure A.3 shows what happened when Häagen-Dazs ice cream was introduced into the United Kingdom. The response of sales to the periods of introductory advertising was quite direct.4

    The examples so far have illustrated the immediate, short-term effect of advertising. However, in Figure A.2, the sequential short-term effects are super-imposed on an upward trend in the consumption of Gold Blend by heavy television viewers. Tracking studies are extensively used for the explicit purpose of demonstrating long-term trends. Figure A.4 shows a 4-year sales trend connected with a new advertising campaign for Stella Artois, an expensive brand of imported beer (represented by the heavy line at the top of the chart). Note that the sales of the other five brands are quite flat over a period of 6 years, thus emphasizing the significance of Stella Artois's performance.5

    Figure A.2. Nescafé Gold Blend Television Advertising and Share of Market by Heavy and Light Television Viewers

    Figure A.5 demonstrates a slightly more subtle point. It tracks the worldwide sales of diamond jewelry, which are ultimately controlled by the international cartel De Beers, Over a 5-year period, sales of diamond jewelry, measured by individual rings, brooches, and other pieces, declined slightly. However, the dollar value of sales significantly increased.6 This was partly a result of the advertising campaign run by De Beers, and the cognitive effects of this are shown in Table A.1, discussed later in this appendix.

    One of the best ways of tracking long-term movements is to use moving totals or moving averages. A moving total for a year (although it can be done for any period) starts with the total for the first 12 months—for example, January through December 1999. The next figure is arrived at by adding January 2000

    Figure A.3. Introduction of Häagen-Dazs Ice Cream in Television Advertising and Value of Retail Sales

    Figure A.4. Sales Trends of 6 Brands of Beer

    Figure A.5. Worldwide Sales of Diamond Jewelry

    and subtracting January 1999. We next add February 2000 and subtract February 1999, and so the progression continues. A moving average for a month (again, any period can be chosen) is arrived at by taking the first year's total and dividing it by 12, then adding January 2000 and subtracting January 1999 and dividing the new total by 12, and so on.

    The object of this procedure is to smooth short-term movements, particularly seasonal ups and downs, and concentrate them as the resultant trend line. In Figure A.6, this process is shown in action for two major brands of chocolate confectionery, Roses and Quality Street.7

    Figures A.1 through A.6 relate advertising to sales. Tracking studies are also employed very widely to show how advertising influences nonbehavioral variables—in particular, various measures of cognitive effect:

    • Advertising awareness, that is, whether the campaign can be recalled by the public either spontaneously or by being prompted (sometimes both)
    • Brand awareness, measured either spontaneously or by being prompted (also sometimes both)
    • Brand associations, that is, whether the brand is associated in the consumer's mind with certain specific functional features and nonfunctional attributes

    Figure A.6. Moving Annual Totals for Brand Shares of Roses and Quality Street

    Figure A.7 illustrates the launch of an extremely successful new brand of mobile telephone with the unusual name “Orange.” Spontaneous brand awareness of Orange (shown in the heavy line) is seen to be increasing strongly and largely without interruption, a rise obviously associated with the periods of advertising (indicated by the horizontal lines enclosed within arrowheads). Orange's strong upward trend contrasts with the unchanging awareness levels of each of the 5 competitive brands.8

    Figure A.8, also devoted to Orange and its competitors, examines a series of brand associations embodied in three separate phases: “sets the standard for the future,” “leads the way in new technology,” and “is dynamic.” How much consumers agree with these phrases is given an averaged score for each brand. The rapid growth in the average score for Orange is again compared with the flat or declining scores for leading competitive brands. The periods of advertising for Orange are indicated by the horizontal lines, as in Figure A.7, and again the relationship is clear.9

    The Orange research is concerned with the launch of a new brand, and with such successful launches the measures can show strong increases. With existing brands, cognitive measures are usually extremely stable. Figure A.9 tracks the average of a number of consumer attitude measures relating to 6 leading marques of car. Note that the left vertical scale only covers the top range of variations, but even with this rather exaggerated presentation of the data, consumer attitudes change very little over a period of 5 years.10 (This stability echoes the trends in Figures A.7 and A.8 for the existing brands that were on the market before Orange.)10

    Figure A.7. Spontaneous Awareness of Brand-Name Orange Related to Periods of Advertising

    Long-term cognitive tracking can provide confirmation that an advertising campaign is having a progressive effect as planned. In other words, it can provide valuable post hoc diagnostic information. Figure A.10 describes the continuous tracking of consumer awareness of advertising for the Meat & Livestock Commission, whose advertising campaign was aimed at raising the public profile of beef. Figure A.10 shows that the advertising campaign that ran between 1990 and 1993 was remembered by a steady proportion of people, although there was a good deal of short-term volatility in the awareness figures.

    With a new campaign introduced in 1994, the average level of awareness jumped up to a higher plateau, which was maintained for the next 4 years, albeit with a similar pattern of up-and-down movements to the earlier period. The advertising campaigns ran over the whole 8 years relatively continuously, and there does not appear to have been a higher level of expenditure during the last 4 years than during the earlier period. The new campaign was therefore clearly more memorable.11

    Figure A.8. Image Associations of Brand-Name Orange Related to Periods of Advertising

    Image attributes can change gradually over time. Table A.1 shows such changes in consumer perceptions of diamond jewelry.12 Note the overall similarity between the trends in both the United States and Great Britain. In both countries, diamonds maintained a high rating for “worth the expense” while showing perceptible improvements in the other measures. The increasing strength of consumers' personal valuation of diamond jewelry therefore helps explain the increasing dollar value of sales, seen in Figure A.5.

    One of the most useful types of tracking study is one that relates a brand's advertising not only to its market share but also to its penetration. Figure A.11 tracks these data for Unilever's remarkable brand of margarine “I Can't Believe It's Not Butter,” a brand that had been marketed first in the United States with similar success. Figure A.11 shows the development of the brand's SOM from

    Figure A.9. Image Associations of 6 Leading Automobile Marques

    Figure A.10. Meat & Livestock Commission Advertising Campaign: Claimed Advertising Awareness

    TABLE A.1 Image Perceptions of Diamond Jewelry: United States and Great Britain (in percentages)

    Figure A.11. Share of Market, Penetration, and Advertising for Launch of “I Can't Believe It's Not Butter”

    Figure A.12. Growth in Per Capita Consumption of Marmite

    the time of its launch: a generally upward progression despite interruptions. The growth of penetration is steady. This pinpoints clearly that the main task for the brand would now be to boost purchase frequency in order to enable sales growth to keep up with penetration growth.13 This case underlines the lessons in Chapter 8.

    A similar analysis can be used to explain the success of Marmite, an old-established strong-tasting savory yeast extract that is spread on bread and toast, particularly at breakfast. (The Australians eat enormous quantities of a similar brand called Vegemite.) Despite Marmite's long history, advertising successfully boosted the brand's sales. And as can be seen in Figure A.12, this success was due to a relatively steady increase in the brand's purchase frequency. During the years 1985–1997, the brand's household penetration rose by 12%, but the per capita consumption (roughly equivalent to purchase frequency) went up by 35%.14 This is another confirmation of the conclusions in Chapter 8.

    The last example of simple tracking shows a rare method but one that provides an insight into how advertising productivity can grow. The procedure is based on a brand's share of voice/share of market relationship, the topic of Chapter 7 of this book. Figure A.13 plots the changes year by year in the SOV and SOM for Stella Artois, the brand of beer examined in Figure A.4. The sharp drop in the brand's SOV (described in Great Britain as Adspend Share) during the period when its SOM was holding steady and its volume sales were increasing significantly, is the clearest possible illustration of the increasing productivity of Stella Artois's advertising.15

    Figure A.13. Share of Voice and Share of Market for Stella Artois

    Complex or Multivariate Studies

    The IPA case studies include a substantial minority that employ multivariate regression to deconstruct the sales of a brand, in particular to isolate the influence of advertising on those sales. As discussed in Chapter 6, this procedure is often called model building, the model in question describing and quantifying the size of the various inputs that determine a brand's sales. I selected six cases to illustrate the methods used, and these are all exceptionally important. (Readers will also remember that I discussed in Chapter 10 the estimates of advertising profitability made in two further IPA cases: those for the Ford Galaxy and the Alliance & Leicester Building Society, a mortgage lender. These examples are based on the sensitive interpretation of considerable statistical data.)

    Figure A.14. Rate of Sale With and Without Advertising for Boddington's Beer in Cans

    In all these studies, the measurement is made by sophisticated statistical analysis, and the simple diagrams explaining the findings should not mislead the reader into thinking that the analyses are in any way facile.

    Figure A.14 presents in a straightforward way the incremental effects of advertising. The brand in question is Boddington's, a beer with great market strength in the north of England and one that is advertised with considerable finesse. In Figure A.14, the volume sales of Boddington's in cans are indexed to maintain confidentiality and are calculated to show sales for each percentage point of weighted distribution (thus factoring out of the calculation any changes in the distribution itself). Figure A.14 shows the modeled rate of sale, overtime, with advertising and without advertising. The difference between the two trend lines is clear, and the gap between them is seen to widen after a year and a half of the advertising campaign.16 This is visual evidence of a long-term effect.

    Figure A.15 illustrates the progress of Cadbury's Boost, a filled chocolate bar directly competitive with the Mars bar so familiar to American chocolate consumers. Figure A.15 tracks the sales of Boost with and without advertising (the top and bottom lines on the chart). It shows in addition the contributions to sales made separately by television and radio advertising. Both have a significant effect, but the contribution of television was, as would be expected, greater than radio.17

    Figure A.15. Sales and Advertising, by Medium, for Boost Chocolate Bar

    Figure A.16 presents the data in a similar way as done by Figures A.14 and A.15, but this time, the brand's sales are deconstructed to measure the separate contributions of (a) advertising, (b) the brand's distribution, and (c) general economic growth in the economy. The brand is Johnson & Johnson's Clean & Clear medicated skin lotion, and the period covered was when the brand was effectively launched. (Clean & Clear had been available in stores during the preceding 2 years but had not been successfully promoted.) Figure A.16 demonstrates the relative sales performances of two commercials, “Girls Talking” and “Real Girls,” which were run sequentially. They both had a measurable effect in the marketplace, but the second generated considerably more sales than the first, although the diagram does not answer the question whether the superior performance of “Real Girls” may have been at least partly due to synergy with the improved brand distribution and better economic conditions that were happening when this commercial was run.18

    Figure A.17 tracks the sales with and without advertising for Knorr stock cubes. “Moira” and “Hen Night” are the names of the two commercials used during the period covered. The deconstructed effect of the advertising on sales appears to have been modest, but it was nevertheless significant.19 Most important, the case study makes it clear that the effect of the advertising did not stop, but it also had a delayed effect:

    Figure A.16. Sales Deconstruction for Clean & Clear

    • —Each burst produces a sales effect with a very long “tail” (to the extent that there is no point during any year when advertising is having no effect on sales).
    • —Sales are always above the level preceding the last burst, when the subsequent burst of advertising takes place.20

    Statistical modeling is often applied to brands with the specific intention of factoring in the delayed effect of previous advertising. In such models, advertising is generally assumed to decay at a fixed rate, and the (declining) residual effect, often known as Adstock, provides a base onto which additional shorter-term effects should be added.21 However, the procedure is not free from controversy. It raises two problems. The first is that lagged effects may be relatively common but are more usually modeled than observed. (With Knorr they areobserved, which is one of the reasons why the Knorr case is so interesting.) Also, in many observed examples they appear to be absent; the effect of advertising appears to stop completely. The second problem is that the procedure does not make totally clear whether the delayed effect of advertising is the result of the advertising itself in the form of memory traces or whether it is the result of positive behavioral changes, especially improvements in penetration and purchase frequency triggered by the earlier advertising.

    Figure A.17. Sales of Knorr Stock Cubes Related to Television Advertising

    Figure A.18 uses a totally different tracking technique. The brand is again Marmite, described in Figure A.12. In Figure A.18, the advertising expenditures for each year for 30 years are shown on the one diagram, plotted against Marmite's sales during the same year.22 There is a reasonably good correlation between the two variables, although the analysis does not completely establish the direction of causality. Does higher advertising always cause higher sales? Or do higher sales encourage the manufacturer to put more money into advertising toward the end of the year, in order to reinforce success? This is a perpetual dilemma for students of advertising effects, since the majority of advertisers still base their budgets directly or indirectly on their achieved or anticipated sales.

    Figure A.18. Scatterplot of Advertising Expenditure and Sales, 1969–1998, for Marmite

    The final example of complex tracking employs an unusual and very imaginative technique. It is based on the demand curve, an analytical device derived from microeconomics. A demand curve plots at any particular time the sales associated with different prices for a brand. Almost invariably, high price is associated with low-volume sales, and low price is associated with high-volume sales. Figure A.19 describes Audi, a high-end European automobile.23 The analysis is based on the average price of used cars. Any improvement (i.e., increase) in price is seen to result from an enhancement of buyers' personal attitudes to Audi. This means, in effect, an endemic strengthening of the brand, something to which the advertising undoubtedly contributed. In Figure A.19, the supply curve, measuring volume of supply at each price level, is the dotted line rising from southwest to northeast. There are two demand curves, descending from northwest to southeast. The equilibrium level of price and sales on the first demand curve is plotted at D1.

    During the 6-month period March—September 1995, the resale value of Audi increased by about 10%. By comparison, the average for competitive marques did not change at all. Audi's price increase, which represents buyers' higher personal valuation of the marque—an outcome influenced by the advertising—can reasonably be applied as a notional price increase for new Audi cars. (The enhanced value of the brand applies equally to used and new vehicles.) This higher valuation can be interpreted either as a higher price for an existing volume of sales or a higher volume of sales for an existing price. The latter hypothesis is represented in Figure A.19 by a shift to the right in Audi's demand curve. The difference between the points plotted at D1 and D2 represents the probable extra number of cars sold as a result of the boost in demand that took place between March and September 1995.

    Figure A.19. Shifting Demand Curve for Audi

    The influence of advertising is seen therefore as taking place in the following sequence: (a) an enhanced valuation of the brand by consumers—the result of successful advertising; (b) boosted demand; and finally, (c) a national increase in the volume of sales.

    Demonstrating how advertising increases consumer demand by means of a shift in the curve to the right is not unknown in the literature of advertising. A similar example was published in 1986, based on market experience during the 1970s.24 This relates to a brand on which I myself worked at J. Walter Thompson and which is briefly described in Appendix B as the Corlett Shift. This is also the device employed by the Simon-Arndt Hypothesis, also described in Appendix B.

    Figure A.20. Stella Artois Market Share, 1980–1988

    A Coda—and a Warning

    The most common—and the most dangerous—use of trend lines is to project them into the future without too much serious thought about what this means. Such projection is often carried out in the most naïve fashion, by drawing a straight line connecting the past to the future. Even when the projection is made in a more subtle way, the procedure can result in disastrous outcomes.

    Consider Figure A.20, which describes the year-by-year market share of Stella Artois, the brand of beer examined in Figures A.4 and A.13. Figure A.20 shows the progress of the brand during the years 1980–1988.

    Figure A.21. Stella Artois Market Share, 1980–1995

    Most observers, amateur and professional alike, would understand the point that Stella Artois's market share was increasing during this period, although at a declining rate. They would therefore immediately be tempted to extrapolate the trend into future years by suggesting a very gradual rise and eventually a flattening of the vertical bars.

    The actual progress of the brand showed a very different picture, as is evident in Figure A.21.25

    The important point made by this simple example is that statistics do not have a life of their own. They are nothing more than numbers that economically and precisely describe underlying patterns. If the statistics change—in particular, if they move consistently in one direction represented by a trend line—they are measuring underlying forces, and it does not mean that these will also continue to change in the same direction during the next 5 years or even next year. If we wish to forecast the future—something we always try to do when we are planning marketing activities—we must not look at any trend line, but we should examine the underlying forces that are expressed by the statistics embodied in a trend line.

    The fallacy of extrapolating trends is apparent to a most distressing degree in all fields of human endeavor. During the closely fought presidential election of 2000, it was obvious to statisticians that the main issue dividing the two candidates was their alternative plans for spending a budget surplus of Utopian proportions that was arrived at by straight-line projections of the long-term fiscal outcomes of an especially strong American economy during the years immediately leading up to the election. Even by the time this book is published, it will be interesting to measure the outturn of these projections one year down the road and, as a result, to judge the financial prudence of the administration's election promises. As I write these words I am overcome by skepticism.


    1. Archives held by the World Advertising Research Centre (e-mail address: Website:

    2. Charlie Robertson, “Health Education Board for Scotland: Smoking. Sticks and Carrots,” in Advertising Works: Volume 8. Papers From the IPA Advertising Effectiveness Awards, ed. Chris Baker (Henley-on-Thames, UK: NTC Publications, 1995), 401–418.

    3. Colin Flint, “Love Over Gold: The Untold Story of TV's Greatest Romance,” in Advertising Works: Volume 9. Papers From the IPA Advertising Effectiveness Awards, ed. Gary Duckworth (Henley-on-Thames, UK: NTC Publications, 1997), 387–403.

    4. Nick Kendall, “Häagen-Dazs: Dedicated to Pleasure, Dedicated to Advertising,” in Advertising Works: Volume 7. Papers From the IPA Advertising Effectiveness Awards, ed. Chris Baker (Henley-on-Thames, UK: NTC Publications, 1993), 191–216.

    5. Jon Howard, Andy Palmer, and George Bryant, “Seven Years in Provence: How a Change in Strategy Helped Stella Artois Retain Market Dominance,” in Advertising Works: Volume 9, 429–457.

    6. Merry Baskin, “De Beers: Hard Times. Selling Diamonds in a Recession,” Ibid., 307–345.

    7. Liz Watts and Cindy Gallop, “Cadbury's Roses: “Thank You Very Much,'” in Advertising Works: Volume 8, 81–101.

    8. Charles Vallance, “Orange: How Two Years of Advertising Created Twelve Years of Value,” in Advertising Works: Volume 9, 5–28.

    9. Ibid.

    10. Gavin Macdonald and Antony Buck, “The Volkswagen Golf, 1984–1990,” in Advertising Works: Volume 7, 75–99.

    11. Sarah Carter and Sam Dias, “Meat & Livestock Commission: Pulling Round the Red Meat Market,” in Advertising Works: Volume 10. Papers From the IPA Advertising Effectiveness Awards, ed. Nick Kendall (Henley-on-Thames, UK: NTC Publications, 1999), 283–315.

    12. Baskin, “De Beers,” in Advertising Works: Volume 9.

    13. Jacqueline Feasey, “I Can't Believe It's Not Butter! From Extraordinary Launch to Long-Term Success,” in Advertising Works: Volume 9, 459–472.

    14. Lucy Jameson and Les Binet, “Marmite: How ‘The Growing Up Spread’ Just Carried On Growing,” in Advertising Works: Volume 10, 127–160.

    15. Howard, Palmer, and Bryant, “Seven Years in Provence,” in Advertising Works: Volume 9.

    16. Guy Murphy, “Boddington's—‘By ‘eck,’” in Advertising Works: Volume 8, 129–160.

    17. Derek Robson, “Cadbury's Boost: ‘Why Work and Rest When You Can Play?,’” in Advertising Works: Volume 8, 279–307.

    18. Polly Evelegh and Sam Dias, “Johnson's Clean & Clear: Global Advertising in a Local Market,” in Advertising Works: Volume 10, 363–399.

    19. Sarah Carter, “Knorr Stock Cubes: How Thinking ‘Local’ Helped Corn Products Corporation (CPC) Develop Advertising Which Toppled the Brand Leader,” in Advertising Works: Volume 6. Papers From the IPA Advertising Effectiveness Awards, ed. Paul Feldwick (Henley-on-Thames, UK: NTC Publications, 1991), 141–167.

    20. Ibid., 163.

    21. Lagged effects are discussed widely in the literature of advertising evaluation. They are most clearly described in Simon Broadbent, The Leo Burnett Book of Advertising (London: Business Books, 1984), 88–98. See also Chapter 6, Note 8.

    22. Jameson and Binet, “Marmite,” in Advertising Works: Volume 10.

    23. Richard Exon, “Audi: Members Only. How Advertising Helped Audi Join the Prestige Car Club,” in Advertising Works: Volume 10, 321–347.

    24. John Philip Jones, What's in a Name? Advertising and the Concept of Brands (New York: Simon & Schuster-Lexington Books, 1986), 95–96.

    25. Howard, Palmer, and Bryant, “Seven Years in Provence,” in Advertising Works: Volume 9.

    Appendix B: Alternative Systems for Measuring Long-Term Effects

    This appendix is devoted to describing two further methods of evaluating the long-term effects of advertising. These are partly in the world of theory and partly in the world of practice. If they were to be applied to the measurement of long-term effects in a wide range of specific cases, they would fulfil their potential as important inductive techniques. The first is an extension of the share of voice/share of market relationship discussed in Chapter 7; the second leans on expository devices developed in the field of microeconomic analysis, namely, shifting curves.

    Changes in Share of Voice

    The discussion of share of voice in Chapter 7 has been based on data for up to 5 years, and the averages for share of voice and share of market have been calculated in a similar way to make the comparison. The average SOV is assumed to be relatively stable and the extent to which it can continuously and safely run below SOM is an expression of the strength of the brand—a strength to which advertising has contributed in its long-term role.

    Figure B.1. Structure of Dynamic Difference/Marketing Advertising Pattern

    However, what happens if a brand's SOV is increased or decreased as part of a plan either to boost sales or to increase profit? This possibility has been widely explored, and it was examined with the use of simple mathematics at least 40 years ago by British statistician Michael Moroney, who worked for Unilever, and by American market researcher James O. Peckham, Sr., of A. C. Nielsen. These two experienced analysts were working totally independently but came to the same conclusion. The Moroney method was called Dynamic Difference,1 and the Peckham term for the same system was Marketing Advertising Pattern (MAP).2

    The system is set out diagrammatically in Figure B.1. Year-on-year changes in SOV are plotted in percentage points on the horizontal axis. (These are calculated as SOV in Year 2 minus SOM in Year 1. The assumption is, of course, that the normal stable category-average relationship is SOV = SOM.) Year-on-year changes in share of market are plotted on the vertical axis. (These are calculated as SOM in Year 2 minus SOM in Year 1.)

    Figure B.2. Dynamic Difference/Marketing Advertising Pattern Lines for Four Hypothetical Brands

    As can be inferred from Figure B.1, a positive effect from advertising will lead to a regression that ascends from southwest to northeast. If the relationship between changes in SOV and changes in SOM is found to be relatively consistent over a number of years, a line can be drawn connecting the observation points. This line is different for different brands, as can be seen in the hypothetical examples in Figure B.2. The steeper the line (e.g., Brand A), the more responsive the brand is to changes in advertising investment, up or down.

    Each observation in Dynamic Difference/MAP really measures changes in the medium-term effect of advertising. However, when we can plot a regression line derived from a number of years of data, then the slope of the line can be interpreted as a measure of the long-term effect.

    There is however a serious problem with Dynamic Difference/MAP. During the 1960s, it was possible to construct lines that fit in about 70% of cases. At that time there was less volatility in most brands' advertising than there is today, when a different campaign is introduced for most brands every 2 or 3 years. In addition, the large and increasing volume of sales promotions today tends to cloud the advertising expenditure/sales relationship. These factors mean that stable Dynamic Difference/MAP regressions are now relatively rare, although when they do exist they are capable of providing a good measure of advertising's long-term effect, a measure akin to year-on-year changes in a brand's advertising elasticity. (Advertising elasticity is discussed in Chapter 9.)

    Figure B.3. Elliott Extension: Dynamic Difference for Kellogg's Rice Krispies in Great Britain

    There is another interesting aspect to Dynamic Difference/MAP. If a regression can be established for a brand, it is possible to quantify the productivity of a successful new campaign that has produced an effect outside the regression line. Figure B.3 demonstrates this device in action with a historical example. I call this special case the Elliott Extension, named after Jeremy Elliott of J. Walter Thompson, London, who first published it.3

    Before the exposure of a new campaign in 1978, Kellogg's Rice Krispies had an established Dynamic Difference/MAP line constructed from 7 years of data (indicated by the crosses on the diagram). The 1978 campaign—based on an extremely powerful advertising idea—received advertising support of 2 percentage points SOV above the brand's SOM of the preceding year. The campaign's performance is shown by its market share improvement of 2 percentage points.

    This sales improvement was far better than the previous regression suggested. According to the earlier response of sales to changes in advertising, an improvement of 2 percentage points in SOM would have incurred an expenditure in advertising of more than 8 percentage points of SOV above the previous year's SOM. Elliott's analysis shows that the productivity of the new campaign in comparison with the old one can be calculated as a dollar value—the difference between an overinvestment of 8 percentage points and 2 percentage points. In this particular example, it meant about a million dollars at 1978 prices, a sum representing the enhanced value of sales from the more effective advertising.

    Dynamic Difference/MAP and the Elliott Extension of this are simple statistical devices. For most brands, the statistical data are readily available. Despite the fact that the first of these artifacts will only operate on infrequent occasions, and the second more rarely still, they are both capable of uncovering hidden underlying relationships.

    Two Shifting Curves
    Corlett Shift

    Microeconomic analysis is usually based on simple geometry. A demand curve, which plots volume bought against price charged, almost invariably descends from northwest to southeast, representing the commonsense idea that a high price equates to small quantities being sold but that a low price will stimulate large sales.

    Tom Corlett, of J. Walter Thompson, London, managed to demonstrate the long-term effects of advertising through a simple shift in the demand curve, but he had to collect a significant amount of data for the brand he used for his exposition.

    Corlett's starting point was to build a demand curve. He chose a major British brand in a reasonably large packaged goods market. This calculation was possible because the brand is sold at various price premiums above other brands in different regions, with resultant differences in the brand's market share, so that a range of data could be provided for a time period of defined length.

    Figure B.4. Corlett Shift

    The price was expressed as the premium over other brands, thus eliminating the effects of inflation. This also means that what is measured on one axis (the price of one brand compared with others in the market) is entirely consistent with what is measured on the other axis (the sales of one brand compared with others).

    The second stage in Corlett's work was the construction of a similar demand curve for a later period after the exposure of advertising. The campaign in question had earlier been judged highly successful on the basis of all available objective measures: ex-factory sales, market share, attitudinal and other qualitative research, and popularity polls. But the best demonstration of the campaign's success is Corlett's analysis of the brand after the campaign's exposure because his demand curve had moved quite significantly to the right, as shown in Figure B.4.

    This clearly showed that, after the advertising, (a) at a given price premium, the brand could now command a quantifiably larger brand share; and (b) at a given brand share, the brand could now command a quantifiably greater price premium.

    This made it possible quite simply to make estimates of marginal sales revenue added by the campaign. These estimates could also be compared with the cost of the campaign so as to indicate its marginal profitability to the manufacturer.4

    Simon-Arndt Hypothesis

    Two academics, American Julian Simon and Norwegian Johan Arndt, addressed the perplexing problem of the strong underlying tendency to short-term diminishing returns (as discussed in Chapter 5) and also the strong underlying tendency to long-term economies of scale for certain brands. Increasing returns and scale economies are different concepts, but in certain circumstances they can be related: The latter can contribute to the former. How can we explain this apparent contradiction between short-term diminishing returns and what looks like long-term increasing returns?

    The most plausible explanation is that the scale economies associated with successful brands apply to all aspects of those brands, stemming as they do from their functional and nonfunctional values and also from such things as distributional ubiquity and high usage.

    An elegant explanation of this phenomenon appears in Simon and Arndt's analysis of Lambin's review of 107 European brands and his findings that larger brands have advertising economies. Lambin noted, “for large brands, the inverse relationship between the advertising-sales ratio and market share.”5

    When Simon and Arndt came to consider this widespread phenomenon, they rightly concentrated on the overall scale economies associated with the larger brands:

    It is reasonable and likely that firm E has a much more extensive distribution network and a larger sales force than firm A. That would explain both why firm E has a higher response function and why it advertises more in total than does firm A. And, depending on the particular slopes of the functions, the advertising-sales ratio could well be lower for the larger firm for this reason alone.6

    Although Simon and Arndt were thinking in terms of the firm rather than the brand, they had clearly grasped the essential point, and their diagrammatic expression of this point in Figure B.5 explains the position of the larger brand with its scale economies.

    Figure B.5. Simon-Arndt Hypothesis

    The authors show the possibility of a family of advertising response curves. Here the position of the large firm with scale economies (Firm E) would, of course, be higher than that of the small one without the scale economies (Firm A). From this analysis, it is obvious why a given incremental dose of advertising has a greater effect for Firm E. The position of the two curves relative to one another is not a reflection of the advertising alone but of the relative strengths of the firms overall to which advertising has naturally made and continues to make a contribution.

    The short-term response function is driven by advertising directly. The fact that this curve has shifted to a more productive location—a long-range outcome—is the partial and indirect but unmistakable outcome of the advertising. This is a true long-term effect.


    1. John Philip Jones, What's in a Name? Advertising and the Concept of Brands (New York: Simon & Schuster-Lexington Books, 1986), 87–92.

    2. James O. Peckham, Sr., The Wheel of Marketing, 2nd edition (privately published 1981, but available from A. C. Nielsen), 108–130.

    3. Jeremy Elliott, “Kellogg's Rice Krispies: The Effect of a New Creative Execution,” in Advertising Works: Volume 1. Papers From the Institute of Practitioners in Advertising (IPA) Effectiveness Awards, ed. Simon Broadbent (London: Holt, Rinehart & Winston, 1981), 86–87.

    4. TomCorlett, “How to Make Sense of Market Analysis,” Campaign, May 26, 1978. Alsodiscussed in Jones, What's in a Name?, 95–96.

    5. J. J. Lambin, Advertising, Competition and Market Conduct in Oligopoly Over Time (New York: American Elsevier, 1976), 127–132.

    6. Julian L. Simon and Johan Arndt, “The Shape of the Advertising Response Function,” Journal of Advertising Research, August 1980, 22. Also discussed in Jones, What's in a Name?, 203–204.

    About the Author

    John Philip Jones was born in Wales in 1930 and graduated in economics from Cambridge University (B.A. with Honors; M.A.). From 1953 to 1980, he worked in the advertising agency field. This experience included 25 years with J. Walter Thompson—as a market research executive in London (1953–55), advertising account executive in London (1957–65), account supervisor and head of television in Amsterdam (1965–67), account director and head of client service in Scandinavia, based in Copenhagen (1967–72), and account director in London (1972–80).

    He worked with a wide variety of advertising clients and was most concerned with major brands of packaged goods (which in his research he titles Beta brands). His responsibilities included many brands marketed by Unilever, Chesebrough-Pond's (before its acquisition by Unilever), Beecham, Gillette, Nestlé, Pan American, Pepsi-Cola, Quaker Oats, and Scott Paper. He was international account director on Lux, the largest selling bar soap in the world, from 1972 to 1980.

    He was extensively involved with advertising education both within and outside the agency. He conducted his final agency seminar in January 1981 before becoming a full-time educator in the Newhouse School of Public Communications at Syracuse University, where he is a tenured full professor and was Chair of the Advertising Department in the Newhouse School for seven years. He has taught a wide range of advertising classes at both the graduate and undergraduate levels.

    For three years, he edited the university's interdisciplinary journal of ideas, Syracuse Scholar. He was a member of the Mellon Foundation project group, which spent two years exploring the connection between liberal and professional education and published its findings in the book Contesting the Boundaries (1988).

    He has published widely in the professional press, with articles appearing in Admap, Commercial Communications, Economist: Media & Marketing Europe, Harvard Business Review, International Journal of Advertising, Journal of Advertising Research, Journal of Marketing Communications, Marketing & Research Today, Marketing Management, Market Leader, and many other publications, including journals in Australia, Great Britain, the Czech Republic, Germany, India, the Netherlands, Scandinavia, and Switzerland. He has also contributed pieces to the New York Times and other publications.

    His books What's in a Name? Advertising and the Concept of Brands (1986), Does It Pay to Advertise? Cases Illustrating Successful Brand Advertising (1989), How Much Is Enough? Getting the Most From Your Advertising Dollar (1992), and When Ads Work: New Proof That Advertising Triggers Sales (1995) are widely used in the advertising profession in the United States and overseas and have been translated into German, Spanish, Japanese, Korean, Chinese, Portuguese, Turkish, and Arabic.

    He was editor and part-author of five major handbooks, published by Sage in 1998, 1999, and 2000. The 2000-page collection deals with all major aspects of advertising practice.

    He has developed measurement devices based on robust quantitative research. These include STAS (Short-Term Advertising Strength) and AIC (Advertising-Intensiveness Curve), which are used in professional practice and described in his books. He is a consultant to many leading consumer goods companies and advertising agencies in the United States and abroad. He also regularly addresses major professional conferences.

    In 1991, the American Advertising Federation named him Distinguished Advertising Educator of the Year. In 1996, he received a major award from Cowles Business Media and the American Association of Advertising Agencies for leadership in the media field. He received the Telmar Award in 1997 for extending the concept of STAS from television to print media. In 2001, he was awarded the Syracuse University Chancellor's Citation for Exceptional Academic Achievement.

    Every year, he presents his work to professional and academic audiences worldwide. He is Adjunct Professor at the Royal Melbourne Institute of Technology, Australia, where he teaches each May.

    • Loading...
Back to Top

Copy and paste the following HTML into your website