<%@ Language=VBScript %> <% CheckState() CheckSub() %> Advertising: Strongly Persuasive or Nudging?
Journal of Advertising Research

Journal of Advertising Research

Advertising Research Foundation
641 Lexington Ave., New York, NY 10022; (212) 751-5656

January/February 1997

Advertising: Strongly Persuasive or Nudging?

Neil Barnard and Andrew S. C. Ehrenberg

Advertising practitioners take a view on what consumers are like and how they can be influenced. Here we focus on a key consideration: individual customers' loyalty to established brands and the advertising implications this has.

On the one hand, do consumers fall into distinct segments of 'loyals' (each committed to just one brand) and 'switchers' (who are not committed to just one brand)? A prime objective would be to increase the number of your brand's exclusive loyals. Advertising would have to be strongly persuasive, with its brand-building role emphasized.

On the other hand, do consumers mostly have split-loyalties to several brands, with habitual brand repertoires? Consumers would all differ from each other, ie, be infinitely segmented (rather than clustering into switchers and a few distinct groupings of loyal, one for each brand).

On the latter view, there would then generally be very little dynamic switching from one brand now to another in future, but steady, split-loyal repertoire-buying. Advertising's role would be to reinforce consumers to continue buying your brand and, at times, to nudge them but with brand-maintenance even then remaining the main task.

In the next section we draw out this distinction more fully. We then turn to the empirical evidence. This supports the unsegmented split-loyalty perspective and hence the reinforcing and nudging view of advertising. Finally, we briefly consider the targeting of advertising and its creative content in this context. More detailed 'Notes' are given at the end of the paper in the appendix.


Very different tasks for advertising are implied by the two contrasting views of brand loyalty. We therefore explore the distinction in more detail, starting with the strongly persuasive view.

Advertising to committed loyals and switchers: Strongly persuasive

Any consumers who are committed or hardcore loyals would be matched with a brand that uniquely satisfies their needs, both tangible and intangible. This would help protect the brand from competitive attack and support premium pricing. Loyals are seen as valuable in sales terms.

In contrast, switchers would not have a brand that is just right for them. They would move from brand to brand as variety-seekers (looking for change), or rotators (moving within a fixed repertoire of brands), deal-sensitives (buying brands on promotion when possible), private-label prones (with choice depending on which store they are in), price-sensitives, and so on. A switcher's contribution to a brand's sales is traditionally thought to be much less than a loyal's.

Acquiring loyals, therefore,would mean converting customers to one's brand from other allegiances. It would require some markedly different persuasive tasks:

The brand maintenance goals in this scenario would differ yet again:

All in all, the 'committed loyals and switchers' perspective would invoke various, very different but all strongly persuasive advertising tasks (though in practice one might decide to restrict oneself to just two or three and ignore the rest). But creating any such persuasive advertising would be difficult and can hardly be produced to order. This is borne out by there being relatively few markedly successful brand-building case histories for established brands.

The alternative: Advertising to split-loyals

Our split-loyalty point of view sees a given brand's customers as generally having idiosyncratic repertoires of one or more other habitual brands. This leads to a much more homogeneous and less ambitious task for advertising than in the loyals/switchers scenario.

Each consumer would choose from her or his repertoire with mostly steady purchase propensities (ie, with only occasional shifts and departures from habit). The brand's advertising to such multibrand buyers does not have to be strongly persuasive, not trying to convert them whole-heartedly to that single brand. In this scenario a brand has few near-100% loyals anyway (as is the case in real life).

There is instead scope for advertising to (1) reinforce your brand's customers' existing propensities to buy it as one of several, (2) 'nudge' them to perhaps buy it somewhat more often, and (3) get other consumers perhaps to add your brand as an extra or substitute brand to their existing brand repertoire (first usually on a 'trial' basis - 'I might try that' - rather than already strongly convinced or converted - see also JOAB Report 6).

The specific contexts for this will be varied (for similar or differentiated brands, different need-states, or different end uses or end users). But the common task is the same: mainly to influence the purchasing propensities for your brand within consumers' existing multibrand repertoires.

The immediate effects would usually be small, advertising's reinforcing and nudging mechanism mainly being long-term. It would involve the advertising leaving implicit or explicit memory traces and associations (as we discuss further in our final 'bring-it-together' JOAB Report 8). The context would be any reinforcing effects of the rest of the marketing mix (eg, availability) and of the actual acts of buying and using the brand (almost certainly the dominant factors), together with the related dissonance-reduction ('I'm buying my brand') and self-justification which advertising can help ('That is my brand'). This is broadly consistent with what many commentators seem to mean when they talk about long-term advertising effects, without implying that advertising would have to act like a delayed 'time-bomb.'

Positive nudging can then, however, also occur for your many competitors and that needs, therefore, to be countered. Hence there is an on-going need for active brand maintenance. But the task of thus defending your customers' propensities to buy your brand and that of positively nudging them can be quite similar. It can be accomplished by effectively reminding consumers about your brand and influencing whether it is salient to them. As we have said, this appears both more uniform and less ambitious than the highly varied advertising tasks in the loyals/switchers scenario. It seems close to traditional issues of influencing long-term brand awareness, familiarity, salience, and brand associations ('leaving a trace').

The question now is the extent to which either of these very distinct views of the consumer correspond to the observed realities - segmented loyals and switchers versus steady split-loyalty propensities to buy.


We now outline the evidence. Do consumers either fall into distinct segments of loyals and various kinds of switchers? Or simply have many varying degrees of split - but habitual loyalty?

A recent Leo Burnett study in the United States gave evidence on the traditional loyals/switchers classification (eg, McQueen et al, 1993). But contrary to general expectations (including, it seems, the Leo Burnett authors'), near-100% loyal customers were not an important segment for any brand. This is in line with our own finding over the years (see Notes). It is also not clear that there were in fact distinct if small subsegments of consistently dynamic switchers. In contrast, the unsegmented split-or shared-loyalty view of consumers has long been supported by extensive empirical evidence and also quantitative theoretical ('Dirichlet') predictions across more than 50 product categories.

Are there distinct segments of loyals?

It is increasingly being recognized that for frequently bought goods, many or most consumers are 'multibrand' buyers. (See Notes. For durables, they similarly have 'consideration sets' of several brands which they might buy and which they often do buy taking a long enough view.)

Yet the notion that there are also many committed 100% or near-100% 'loyals' is still found intuitively appealing. This is partly because most people can say 'But I always buy Brand X,' picking on a product category where they do buy only just one brand.

In addition, much observed data simply looks like it: In a week nearly everyone appears 100% loyal because they mostly buy the product category or subcategory only once (if at all). And similarly there are many 100%-loyals in a month. But not in longer analysis periods. Typically in a year (or over a succession of 10 purchases or so) only around one in ten of the buyers of the brand are 100% loyal or nearly so. Such numbers are moreover predictable without having to assume any 'hardcore loyals.' If most consumers are in fact polygamous, some will happen to still appear monogamous (in closely predictable numbers) in any chosen analysis period simply by dint of being light buyers of the product category who, therefore, have little opportunity to be disloyal.

Indeed, longer term loyals are not heavy buyers even of their own particular brand. They therefore do not account for any specially attractive amount of its sales. This has been found to be so for any brand or category (see, for example, Table 2 in the end of paper Notes). Even a brand's heavy buyers still buy other brands almost half the time. Heavy buyers are not loyals to that brand who just occasionally fall out of line. Instead they still have serious steady split-loyalty relationships.

Loyals as usually defined (near-100% loyal), therefore, do not make up either a sizeable or a specially attractive subset of customers for any brand. They should not be ignored. But they do not warrant any heavy advertising or other heavy marketing activities geared specifically to them (eg, to persuade or convert other consumers to become 'loyal' to your brand).

Are there 'switcher' segments?

The idea that the labeling of consumers as 'deal-prone' or 'price-sensitives,' etc, actually reflects distinct subsegments of switchers is also lacking in effective supporting evidence, as we now note:

All in all, there is no systematic evidence in the literature that rather homogeneous, distinct, and substantial segments of 'loyals' or various 'switchers' can clearly be identified. Instead it seems that consumers mostly follow a great variety of split-loyalty patterns. This leads, as we will now see, to closely predictable numbers of just how many behave in the various different ways. There would therefore be little point in devoting advertising or other resources to try to persuade these elusive switcher segments to drop their other brands or, more generally, to make your brand's franchise abnormal, compared with all other brands.

The unsegmented split-loyalty (or Dirichlet) view

The contrary split-loyalty view of consumers is embodied in the Dirichlet theory. This has a wide range of highly successful predictions (as illustrated for Folgers instant coffee in Table 1 in the Notes here, and also for many other products - over 50 in all - and different countries, etc, as noted in JOAB Report 1). Such a split-loyalty perspective then leads to the very different reinforcement-and-nudging view of advertising.

The basic supposition in the theory is that each consumer tends to choose from a personal repertoire of brands with steady habitual propensities: Jane Doe might have a repertoire of three brands, bought in proportions of .6, .3, and .1. Over time she would buy the three brands in an irregular (probabilistic) manner, reflecting her varying reasons, including sheer habit. But in a near-steady-state market, the brands would still account for some 60%, 30%, and 10% of her purchases over time. The make-up of such brand repertoires, however, varies greatly between different consumers.

Jane Doe's low-probability 0.1 brand would still be bought 'loyally' even if infrequently, ie, habitually for 1 in 10 of her category purchases. (In a typical case study for the 10 leading US detergents, the Dirichlet model predicted that of those customers who bought a brand only once in a quarter, 41% would buy it again during the next quarter, compared with 39% observed.) Consumers in this empirically based Dirichlet-type split-loyalty theory are, therefore, like the 'rotators' in the segmented view of the market but with three differences:

1. Virtually all the customers of a brand show up as such rotators (other than for very light buyers who hardly have the chance to buy more than one brand).

2. The numbers are predicted (how many multibrand buyers there should be, how much they should buy, their shares of category requirements, and so on, instead of all being left up in the air).

3. The predictions tend to come out right, both in short and in long time periods (as has been documented over and over again in the last 30 years - see, for example, the Table 2 illustration in the Notes and details and references there).

The evidence, therefore, widely supports the view that consumers' mostly split-loyalties cover the full range of habitual propensities-to-buy from high to low and do not fall into highly distinct loyalty subclasses.

Flexible habits

Strictly speaking, the Dirichlet model is a model of the steady-state market, with stable market-shares. But the model does not say that such stability will or must happen. Instead, it only successfully predicts what markets are like when they are stable or close to stable. In particular, consumers' individual brand-choice propensities should then also appear to be steady and habitual.

Nonetheless, such habits can change at times. As we all know, habits are difficult to change when it involves the nature of the habits, eg, drinking coffee without milk, or giving it up altogether. But habits can change fairly easily when it only involves changing the object of the habit in minor ways. For example, consistently buying brand A in place of the near lookalike brand B. But even such 'minor' changes in brand choice are relatively rare insofar as there is not much reason for making a switch, eg, if the brands are near-lookalikes. And not changing one's brand allegiance at all takes even less decision-making effort. Brand loyalty in our view is primarily a matter of habits (nothing else is assumed in the Dirichlet model, for example). And habits are, we think, adopted because they are convenient. (One does not have to think much about which brand to buy, at least not this time.)

When a significant and lasting upset in market share does occur, the evidence implies that customers still split their loyalties in much the same way as before but with some changes in the make-up of their repertoires. This is consistent with the view that advertising for established brands can work by occasionally nudging and mostly reinforcing such split brand choice propensities, rather than by strong persuasion or total conversion.


The evidence, therefore, shows that split or shared loyalty is an ever-present part of the context in which advertising has to operate. (For durables, etc, consideration sets with more than one option are analogous to split loyalty repertoires and with a good deal of multibrand buying over a long enough time perspective.) We now briefly note some of the implications for targeting and for the creative advertising content. To what extent can campaigns be effectively targeted or convey very brand-specific messages (as might, for example, seem required in any 'persuasive' model)?

Targeting in general

Targeting a particular consumer group presupposes that that group is important. But the advertising will also reach many other consumers. Any effects will probably depend more on those actually reached, ie, including the spill-over, than just on those successfully 'targeted.' Targeting considerations will in any case seldom differ much between competitive brands, since more or less similar brands (and often even somewhat dissimilar ones) tend to appeal to much the same types of consumer (see Notes).

In talking to young people, for instance, one therefore cannot simply ignore that older people are overhearing what is being said. This can lead to conflicts in the creative positioning, which may matter with advertising that aims to be strongly persuasive. But such conflicts are far less likely (though not impossible if inconsistency creeps in) with ads which merely aim to reinforce and nudge, by seeking to influence that the brand is salient to those reached, rather than trying to change what they think or feel about it ('strong persuasion').

Reassuring heavy buyers and reminding light ones?

Heavy buyers in a brand's franchise are always few, yet account for much of its sales, ie, the old '80:20' rule. This being predictably much the same for any brand (ie, always about 80:20), neither advertising nor any other marketing activity can markedly change it.

But even heavy buyers of a brand are at risk. The findings are that in a year they still buy and use other brands in total almost as much as they buy their 'heavy brand' - they are therefore routinely exposed to other brands anyway. And they could also switch to be heavy buyers of another (similar) brand. They are therefore not to be taken for granted. Advertising which reaches the heavy buyers of a brand can give reassurance that buying this brand (heavily) is ok.

Most buyers of a brand are, however, light buyers of it. For example, about 50% of the annual buyers of a US instant coffee brand buy the brand only once in the year. Such light buyers of a brand are nonetheless generally 'loyal' to that brand in that they buy the brand again, and predictably so, but just not very frequently. And they also buy competitive brands even more. They could, therefore, overlook a brand they buy so infrequently and are worth reminding.

Even your 50% lighter buyers still account for some of your sales (about 20%) and appear worth holding on to. They are your brand's most numerous ongoing customers; they are familiar with it and are endorsing it by continuing to buy it (occasionally) despite all their ongoing experience of other brands. Such customers tend to be the main source of a brand's awareness, familiarity, and reputation.

When a brand's sales do increase, most of the extra buyers gained will again be light buyers of the brand. Even with transient sales promotions, the findings show that consumers have already to be familiar with the brand - even if only as light buyers - to make use of the deal at all (eg, Ehrenberg et al., 1994). So light buyers still matter.

A somewhat 'leaky bucket': Recruiting new buyers

We have emphasized that individual consumers' purchase propensities are mostly stable and predictably so. But over a year or so, it is found that there is nonetheless some erosion in any brand's repeat-buying level (eg, 48% repeat-buy from one quarter to the next, but only 40% to a quarter a year or more later [see Notes]).

Established brands, therefore, do not maintain their market shares simply by retaining all their existing customers. There is something of a 'leaky bucket' that needs to be more or less continually topped up by recruiting some new trialists (some effective nudging). Advertising would be expected to play a part in such proactive brand maintenance.

Creative content

Creatively, can advertisements which are explicitly designed to recruit new trialists also serve to (1) remind existing light buyers and (2) reassure heavy ones? This may be asking too much of one campaign. By focusing just on attracting new buyers, for instance, it may fail to take the existing buyers properly into account - in their just needing to be reminded or reassured in some way, a heavy selling message may jar or at least be screened out.

Resolving such trade-offs between different target groups would be difficult since advertising campaigns are not, as we have noted, neatly separable in whom they reach. Hence so much advertising does not in fact try to make a specific claim targeted only to a specific consumer group. Instead, judging from the general run of advertisements, the creative approach usually aims to make an impactful presentation that will be associated with the brand more broadly, by simply promoting the existence or salience of the brand effectively. When the product is being bought, the advertised brand may then come to mind or even appear reassuringly familiar.


In this paper we have argued that it is not the task of advertising to persuade consumers to be highly committed or loyal to one particular brand. The empirical evidence does not support that this 'committed' view can work.

If consumers had been strongly persuaded in the past, there would be many near-100% loyal customers for each successful brand, and they would be important to its sales. Instead, over a year or so, few customers of a brand remain exclusively loyal to that brand (or even near-100% loyal), nor are they then particularly heavy users of it.

Such low numbers of observed near-100% loyals are not just due to marketing mix factors like out of stocks or competitive price cuts, since they occur also in steady-state markets. And the Dirichlet model predicts the low numbers without assuming any such marketing-mix factors. Instead, most consumers have multibrand repertoires (as is also predicted). They can therefore merely choose another of their already habitual brands if it is on deal, or if their intended one is out of stock, without this having to affect the overall selection of brands they are loyal to overtime.

Another possible counterargument is that few households appear 100% loyal in the usual scanner-panel data because there are different end users in the household, each of whom may separately be 100% loyal to his or her own preferred brand. But both the predicted and observed incidence of 100% loyals are low, even in one person households, and also for products where there is virtually only one end user in a household anyway (as for heavy duty laundry detergents). The low incidence of single brand loyalty is, therefore, a genuine feature of the marketplace.


Given that a brand's customers mostly have split loyalty repertoires of habitual brands, our conclusion is that advertising for an established brand can reinforce and occasionally positively nudge their propensities to buy the brand - but only as one of a number of familiar and habitual brands for that consumer.

We believe that this adapts and extends the modern consumer-centered view of advertising. Instead of advertising having to do something to the otherwise indifferent or unwilling consumer, the consumer would be doing something with the advertising. This would work with the grain rather than against it when the consumer willingly endorses, or at least acquiesces in, what the advertiser sets out to achieve.

It seems that consumers' ongoing tendencies to be loyal to several brands (or to have limited consideration sets for durables) result from the need to simplify decision making, making frequent brand choices manageable, or being content for a time with one's previous choice of durables.

Advertising can then play a part in influencing the make up of such brand portfolios or wider consideration sets. It helps consumers to do what they seem inclined to do anyway: choose from a limited range of habitual brands. Advertising can achieve this by what we have here called reinforcing and nudging: it can affect how many such split-loyal customers each brand has rather than how loyal they are to it.

Brands that are not in the consideration set or portfolio are not necessarily always rejected but more often merely ignored. They may never have been tried. Such competitive brands could therefore be brought into consumers' consideration sets as possible choices. This provides opportunities for nudging consumers toward your brand. But it also provides threats from your competitors which your brand has to guard against.


Aaker, D. A. Managing Brand Equity. New York: Free Press, 1991.

R. Batra, and J. G. Myers. Advertising Management (4th ed.). Englewood Cliffs, NJ: Prentice Hall, 1992.

Barnard, N. R., A. S. C. Ehrenberg, K. Hammond, and M. D. Uncles. 'Model Discrepancies: Two Approaches.' SBU Working Paper, 1997.

Bartholomew, D. J. Latent Variable Models and Factor Analysis. Oxford: Griffin, 1987.

Bass, F. M. 'The Theory of Stochastic Preference and Brand Switching. 'Journal of Marketing Research 11, 1974: 1-20.

Blattberg, R. C., and S. K. Sen. 'Market Segments and Stochastic Brand Choice Models. 'Journal of Marketing Research 13, 1976: 34 45.

Brown, G. H. 'Brand Loyalty-Fact or Fiction?' Advertising Age, 1952-53.

Colley, R. H. 'Defining Advertising Goals for Measured Advertising Results.' New York: Association of National Advertisers, 1961.

Colombo, R. A., and D. G. Morrison. 'A Brand Switching Model with Implications for Marketing Strategies.' Marketing Science 8, 1989: 89-106.

Copeland, M. T. 'Relation of Consumers' Buying Habits to Marketing Methods.' Harvard Business Review 1, 2, 1923: 282-89.

Cunningham, R. M. 'Brand Loyalty - What, Where, How Much?' Harvard Business Review 34, 1956: 116-28.

De Chernatony, L., and M. McDonald. Creating Powerful Brands. Oxford: Butterworth-Heinemann, 1992.

Dekimpe, M. G., and D. M. Hanssens. 'Empirical Generalizations about Market Evolution and Stationarity.' Marketing Science 14, 1995: G 109-121.

Diggle, P. J., K.-Y. Liang, and S. L. Zeger. Analysis of Longitudinal Data. Oxford: Clarendon Press, 1994.

East, R., and K. Hammond.'The Erosion of Repeat-Purchase Loyalty.' Marketing Letters 7, 1996: 163-71.

Ehrenberg, A. S. C. Repeat-Buying (2nd ed.). London: Edward Arnold, 1972; New York: OUP, 1988.

-. 'Repetitive Advertising and the Consumer.' Journal of Advertising Research 14, 2, 1974: 25-34.

-. 'Car Switching.' SBU Working Paper (1998, in preparation).

- , K. Hammond, and G. J. Goodhardt. 'The After-Effects of Price-related Consumer Promotions.' Journal of Advertising Research 34, 4, 1994: 11-21.

-, and M. D. Uncles. 'Dirichlet-type Markets: A Position Paper.' SBU Working Paper, 1997 (submitted to the Journal of Marketing).

Fader, P. S., and D. C. Schmittlein. 'Excess Behavioral Loyalty for High-Share Brands: Deviations from the Dirichlet Model for Repeat-Buying.' Journal of Marketing Research 30, 1993: 478-93.

Frank, R. E. 'Correlates of Buying Behavior for Grocery Products.' Journal of Marketing 31, 4, 1967: 48-53.

Goodhardt, G. J., A. S. C. Ehrenberg, and M. A. Collins. The Television Audience: Patterns of Viewing (2nd ed.). Aldershot: Gower, 1975, 1987.

Greene, J. D. Consumer Behavior Models for Non-Statisticians. New York: Praeger, 1982.

Grover, P., and V. Srinivasan. 'A Simultaneous Approach to Market Segmentation and Market Structuring.' Journal of Marketing Research 24, 1987, 139-53.

Hammond, K., and A. S. C. Ehrenberg. 'How Important Is the favorite Brand: An Empirical Investigation.' MSI Conference, Tucson, 1994.

N. R. Barnard, and A. S. C. Ehrenberg. 'The Near-100% Loyal Buyer.' SBU Working Notes, 1995.

-, A. S. C. Ehrenberg, and G. J. Goodhardt. 'Market Segmentation for Competitive Brands'. European Journal of Marketing 30, 12, 1996: 39-49.

-, S. Long, and A. S. C.Ehrenberg. 'Follow-up Studies on Consumer Promotions.' SBU Working Notes, 1997.

Hedges, A. Testing to Destruction. London: The Institute of Practitioners in Advertising, 1974.

Jacoby, J., and R. W. Chestnut. Brand Loyalty: Measurement and Management. New York: Wiley, 1978.

Joyce, T. 'What Do We Know About How Advertising Works?' The Market Researcher Looks at the Way that Advertising Works. Amsterdam: ESOMAR; also London: J. Walter Thompson, 1967.

Kamakura, W. A., and G. J. Russell. 'A Probabilistic Choice Model for Market Segmentation.' Journal of Marketing Research 26, 1989: 379-90.

King, S. 'Can Research Evaluate the Creative Content of Advertising?' Admap, June 1967.

Lal, R., and V. Padmanabhan. 'Competitive Response and Equilibria.' Marketing Science 14, 1995: G101-G108.

Lannon, J. 'Advertising Research: New Ways of Seeing.' Admap, October 1985.

Lavidge, R. J., and G. A. Steiner. 'A Model for Predictive Measurements of Advertising Effectiveness.' Journal of Marketing 25, 4, 1961: 59-62.

Massy, W. F., R. E. Frank, and T. Lodahl. Purchasing Behavior and Personal Attributes. Philadelphia: University of Pennsylvania Press, 1968.

-, D. B. Montgomery, and D. G. Morrison. Stochastic Models of Buying Behavior. Cambridge, MA: MIT Press, 1970.

McQueen, J., C. Foley, and J. Deighton. 'Decomposing a Brand's Consumer Franchise into Buyer Types.' In Brand Equity and Advertising, D. A. Aaker and A. L. Biel, eds. Hillsdale, NJ: L. Erlbaum Associates, 1993.

Mellens, M., M. G. Dekimpe, and J-B. E. M. Steenkamp. 'A Review of Brand-Loyalty Measures in Marketing.' Tijdschrift voor Economie und Management, 1996.

Mercer, D., and A. S. C. Ehrenberg. 'Brand Segmentation Across a Range of Markets.' Open University Working Paper, 1997.

Rossiter, J. R., and L. Percy. Advertising and Promotion Management. New York: McGraw Hill, 1987, 2nd ed., 1996.

Rust, R. Advertising Media Models. Lexington, MA: Lexington Books, 1986.


JOAB Reports 1 to 8:

1. Ehrenberg, A. S. C., and J. A. Scriven. 'Added Values or Propensities to Buy?' Admap, in press, 1997.

2. Barnard, N. R., and A. S. C. Ehrenberg. 'Advertising: Strongly Persuasive or Nudging?' (This paper).

3. Ehrenberg, A. S. C., J. A.Scriven, and N. R. Barnard. 'Advertising and Price.' Journal of Advertising Research (forthcoming).

4.- , and N. R. Barnard. 'Advertising and Product Demand.' Admap, 1997, in press.

5. -, -, and J. A. Scriven. 'Differentiation or Salience.' Journal of Advertising Research, 1997, in press.

6. -. 'New Brands and ATR&N.' Admap, 1997, in press.

7. Barnard, N. R., and A. S. C. Ehrenberg. 'Consumer Attitudes.' (1997, In preparation.)

8. Ehrenberg, A. S. C., N. R. Barnard, and J. A. Scriven. 'Justifying Our Advertising Budgets.' (To be published.)


Notes and references

This paper is based on one of eight reports in the two-year South Bank project on 'Justifying Our Advertising Budgets' (JOAB). It aims to help companies better establish how their large scale advertising can be justified. Technicalities and references are given in these Notes under the same headings as in the main text.

JOAB participants in the United States and Great Britain have included: BMP DDB Needham, British Gas, BP, British Telecom, CBS, Coca-Cola, Colgate, General Motors, IDV, IPA, Philip Morris/Kraft, Procter & Gamble, Prudential, Publicis, Scottish Courage, Seagrams, Shell, South Bank, Unilever, and United Biscuits. We are also indebted to Helen Bloom, Pam Mills, Colin McDonald, and other colleagues for helpful comments.

Advertising: Strongly persuasive or nudging?

The differing views of customer loyalty have different advertising implications. The notion of brand loyalty has a long history. Copeland (1923) wrote about 'brand insistence' in the first volume of HBR. Jacoby and Chestnut (1978) gave a classic review of the many opinions about brand loyalty and possible measures of it emphasizing commitment, but with little systematic evidence.

Mellens et al. (1996) recently reemphasized the popular view that brand loyalty should represent some deeper commitment to the brand, but again cite no evidence. They also still fail to note the extent to which the different measures of loyalty correlate across different brands in a split-loyalty context, and that the measures do so without theoretically requiring any deeper commitment to each brand.

The question here is then how advertising can influence the consumer. For simplicity we have focused on two polar opposites: segmented loyalty (eg, groups of customers who are highly loyal to particular brands) versus unsegmented or divided loyalty, varying more or less continuously across different consumers.

Brand loyalty and advertising

Advertising to committed loyals and switchers: Strongly persuasive

The idea that persuasive advertising brings about conversion and commitment to a brand is an old one (eg, DAGMAR and AIDA, see Colley [1961] and Lavidge and Steiner [1961]). Such Hierarchy-of-Effects models have, however, been criticized (eg, Joyce, 1967; King, 1967; Ehrenberg, 1974; Lannon, 1985).

Even with advertising nowadays being seen in the broader context of 'branding,' the idea of committed loyalty to a brand would still entail conversion of the consumer to be 'loyal' (but by the brand as a whole rather than just by its advertising). Hence the modern rhetoric of brand-building, of the power of brands (eg, Aaker, 1991, de Chernatony and McDonald, 1992), and the older manipulative view of 'persuasion,' eg, by adding values.

The alternative: Advertising to split-loyals

Seeing consumers as splitting their loyalties between several brands implies nonexclusive propensities to buy each of several brands (the polygamous consumer). It would then be possible to vary these propensities without requiring any radical change in consumers' habitual brand repertoires or behavior. There is thus scope for nudging through advertising and, consequently, also the need for defensive reinforcement. This nonpersuasive view of advertising has seldom, it seems, been very explicitly or formally developed (but see Hedges, 1974), or linked with the supportive evidence that is now available.

Segmented or unsegmented loyalty: The evidence

Segmented and unsegmented views of loyalty have each been widely discussed but they have seldom been directly contrasted.

George Brown (1952/53) initiated the inspection and classification of individual purchase sequences from panel data, typically of 'loyals' and various types of 'switchers.' See also Cunningham (1956), Frank (1967), and Massy et al (1968), and for overviews again Jacoby and Chestnut (1978) and Mellens et al (1996). Advertising texts also still refer to such theoretical classifications (eg, Rossiter and Percy, 1987; Aaker et al, 1992), and practitioners also use them (eg, the Leo Burnett study - McQueen et al, 1993- actually tested it, with, we think, unexpected results).

Assumptions of distinct segments of hardcore loyals and switchers have also been incorporated in theoretical models of buyer behavior (eg, Massy et al, 1970; Bass, 1974; Blattberg and Sen, 1976; Grover and Srinivasan, 1987; Colombo and Morrison, 1989; Kamakura and Russell, 1989; and others). The models require specific parameter estimates for each segment (they are not parsimonious) and no generalizable results seem to have been reported.

The alternative view that consumer heterogeneity is distributed in a smooth unsegmented way is also part of a wider modeling tradition (eg, Bartholomew, 1987; Diggle et al, 1994). The Dirichlet model - originated and empirically calibrated by Gerald Goodhardt and Chris Chatfield and theoretically also by Frank Bass and his colleagues is one particularly simple case (Ehrenberg, 1988, and earlier references there; Ehrenberg and Uncles, 1997).

Are There Distinct Segments of 'Loyals'?

Looking at purchase sequences usually reveals some runs like AAAAAA ('100%-loyal' in that data) and AAABAA ('near-loyal'). So some consumers certainly look like loyals. Even in pairs of consecutive purchases, for instance, repeat-levels (AA) of about 50% are both observed and predicted (for grocery products and for motor cars say, eg, Grover and Srinivasan, 1987; Fader and Schmittlein, 1993; Barnard et al, 1997; Ehrenberg, 1996).

However, over longer purchase sequences or in say a year, typically only around 10% to 15% of a grocery brand's buyers appear 100% loyal to it (including near-loyals). This is one of the routine predictions from the Dirichlet (see references above). The brand performance audit in Table 1 taken from JOAB Report 1 (see also Ehrenberg and Uncles, 1997) and reproduced below, illustrates this in the seventh column of figures as one of many predictions for Folgers (13% - 100% loyal observed, 14% predicted). Table 3 in the same reference showed it for a wider range of products.


% Buying in


% Buying


100% Loyal





per buyer


5 +

Q by Q




























% Who also bought




















Note: O = observed T = theoretical. (Annual unless stated), *Low (see text), ±Share of category requirements

In the model, consumers are very heterogeneous in their choice of brand. Some will buy only one brand just 'by chance' (the model is 'stochastic'), without assuming a segment of hardcore loyals. (In the whole product category the number of one-brand-only buyers builds up, because the 100% loyal buyers of each brand by definition do not overlap). A stochastic analogy is with tossing pennies. When throwing a penny n times, heads may up n times; this does not mean that it was a double-headed penny ('hardcore loyals'). Instead it is predictable on a chance basis.

Occasional lapses from 100% loyalty do not upset these conclusions. (Near loyals, with 10% to 20% other brand purchases, are still only a few % in number - Hammond et al, 1995; and Hammond and Ehrenberg, 1994.)

Are there switcher segments?

With few of a brand's customers buying only one brand in a year, the great bulk are multibrand buyers. In the empirically based Leo Burnett study (McQueen et al., 1993) most buyers were classified as 'rotators,' ie, each consumer moving fairly regularly within a limited set of brands. This is, by another name, the habitual split loyalty consumer of the Dirichlet model. (But this model assumes that such rotators are extremely varied so that they can hardly be termed a 'segment.' The Dirichlet also predicts all the numbers involved - how many rotators there should be, how much they should buy, etc, thus providing stringent tests of its assumptions.

In a deal-intensive environment some purchase sequences will involve a fair number of purchases on deal.

Some consumers will therefore look deal-sensitive. But are they really 'sensitive to deals'? Apparently not: Consumers buy their own familiar brands when they are on deal but do not switch to other 'unfamiliar' brands when they are promoted (eg, McQueen et al, 1993; Ehrenberg et al, 1994; Hammond et al, 1997). The Leo Burnett study in the United States reported this both for its 'loyals' and its 'rotators.'

Similarly, price-segments appear to be predictable (see JOAB Report 3).

The unsegmented split loyalty view (the Dirichlet)

This split loyalty view is nowadays embodied in the Dirichlet model (see Ehrenberg and Uncles, 1997, for an overview and earlier references, eg, Ehrenberg, 1972, 1988).

The Dirichlet assumes that each consumer has ongoing purchase probabilities for the brands in their own repertoires (eg, the .6, .3, and .1 example in the main text). These probabilities are thought to capture consumers' apparently irregular and diverse reasons for choosing a particular brand on each given occasion (eg, varying end uses or end users or need states, varying retail availability, etc). These 'reasons' vary greatly across different consumers and over time, as represented in the model by smooth probability distributions.

The same form of theoretical probability distribution is used for different brands and product categories. It is of the 'beta' format (which is also successfully used for exposure distributions in media research, eg, Goodhardt et al, 1975; Greene, 1982; Rust, 1986). The numerical input needed to calibrate the Dirichlet model is very parsimonious (three basic product category statistics, together with the steady state market share of each brand as the only brand specific input - see earlier references).

These theoretical assumptions are so limited and simple that they could easily prove wrong in practice. But the fit of the predicted brand performance measures (as in Table 2 above, and across many other different brands and products, eg, Ehrenberg and Uncles, 1997) shows that in practice the Dirichlet provides very good benchmarks (correlations generally of .9 + ). Effects of market mix factors over and above their influence on market shares will show up as deviations and subpatterns from the predicted benchmarks.

In using a brand's market share as an input, the Dirichlet does not seek to forecast either short or longer term variations in market shares (unlike more 'dynamic' models: econometric ones, logits, etc). Instead, the Dirichlet provides steady-state norms for diagnosing and evaluating the nature of such market-share variations in consumer terms.

The general view that in the medium term most markets are mostly near stationary most of the time is increasingly widely supported in the literature (eg, Dekimpe and Hanssens, 1995; Lal and Padmanabhan, 1995; and earlier references there).

Neither the Dirichlet nor most alternative models, however, explain the $64 million question why a brand is the size it is. This we discuss in broad terms elsewhere (see JOAB Reports 6 and 8).

Advertising with Split Loyalties

Background references to the literature for this section of the main text are:

Targeting in general

In a segmentation study of various product categories in the United States, United Kingdom, Germany, and Japan, Hammond et al (1996) showed that similar brands generally appeal to similar types of consumers. The study is currently being extended (Mercer and Ehrenberg, 1997). The result is in line with the lack of systematic and positive evidence of strong brand segmentation in the more general literature.

Reassuring heavy buyers and reminding light ones

The classic (but rough and ready) 80:20 rule says that the 50% of your customers who are the heavier buyers account for 80% or so of your sales, and the 50% lighter buyers for only 20%. For the 12 products documented in Table 2A (JOAB 1 Notes), some 49% of the average brand's customers bought it only once (ie, were light buyers) and accounted for 14% of the brand's sales (ie, close to about 20%).

A somewhat 'leaky bucket': Recruiting new buyers

Our colleagues' current work (East and Hammond, 1996) has brought out that even in apparently steady-state markets, some leaky-bucket type of erosion occurs. It does so across the board, ie, for light and heavy buyers, for different brands, and different product categories. The extent of such erosion (eg, 48% down to 40% as noted in the main text) is marked but still a far cry from the larger-scale instability or relentlessly dynamic 'switching' between brands that is often assumed to be normal. It is probably as limited as it is because of most brands' active brand maintenance efforts (eg, ensuring stable retail distribution, competition, advertising, etc).

© Advertising Research Foundation