Over a month, a quarter, or a year, the sales of any brand are determined by the number of people who buy it, how often they do so and which other brands they buy. Marketing practitioners, therefore, invest substantial sums to influence consumer behaviour through communications, sales promotions and loyalty schemes, all designed to increase the penetration and repeat purchase of the brands they manage. The effectiveness of these investments is often assessed by the extent of market-share growth achieved.
One survey of nearly 700 senior managers found share to be the marketing metric most regularly reported to the board (1).
The attention is hardly surprising. In many consumer goods categories a single share point may be worth several million pounds a year. Importantly for managers, the concept is easy to understand, responsive to intervention in the short term and quick to report. It should be the ideal metric; in fact it has one serious limitation.
STATIONARY BRAND SHARES
Consumer panel evidence reveals a paradox that is often missed in day-today decision-making. Despite quarterly fluctuations, in most established fmcg categories, brand shares remain roughly stationary, at least over a year or two. Any gains or losses are only temporary.
Two explanations have been proposed, and both probably apply. The first, that the effects of competing marketing activity are usually offsetting, may instinctively feel right; the second runs counter to traditional marketing wisdom.
Far from being loyal to single brands, it appears most consumers are both experienced and habit-driven, and regard competing brands as substitutable within a repertoire. This once led Andrew Ehrenberg famously to remark, 'Your customers are the customers of other brands who occasionally buy you' (2).
DO THE LAWS OF MARKETING HOLD LONGER TERM?
Consumer repertoire buying falls into regular patterns that repeat across categories, countries and cultures, from which the few well-known laws of marketing science have now been established. These include the laws of double jeopardy (DJ) and duplication of purchase, which describe mathematically the market structure effects of repertoire buying (3).
Market structure may also be estimated more fully with the NBD-Dirichlet (4) which models many common measures – including the purchasing of light and heavy buyers, overall brand loyalties and brand switching – from just a handful of inputs.
An underlying assumption of the laws and the model is that category consumers have independent but steady buying propensities across a portfolio of brands, at least for the time being.
As long as these remain steady, relative shares remain roughly in equilibrium. Yet even in the face of some short-term market dynamics, the Dirichlet output fits surprisingly well and it is therefore often used to analyse category structure for tactical marketing mix planning, and brand performance benchmarking.
One question that has not yet been fully addressed is whether its application might be extended to a strategic level – the longer term. The models fit because buying propensities mostly remain approximately stable, at least over a quarter or two.
Given levels of marketing expenditure on long-term brand equity benefits, in time one might expect to see at least some permanent change to buying behaviour, a shifting in established short-term patterns that would violate the Dirichlet assumptions. However, to date little work has been conducted on strategic effects, largely because of an absence of suitable data.
Now for the first time, with the benefit of a new six-year consumer panel, we can address three questions:
- Market-share equilibrium is normal for most brands in most established categories over a few quarters; but for how long does it really hold in the face of relentless competitive marketing-mix investment?
- If, as we might expect, a number of brands are seen to grow or decline permanently, can we identify any commonalities behind such change?
- To what extent does current knowledge inform an extended time frame? Do the existing laws still adequately describe long-term behaviour, even where markets are dynamic? Can the Dirichlet be used both to model strategic outcomes, and to assess them?
In answer to these questions we report three findings that establish an important extension to our knowledge of buying behaviour, extend boundary conditions for the Dirichlet, and bring us closer to an understanding of how brands grow. We describe the new panel data before moving on to the findings and a discussion of their implications.
A TWENTY-SIX QUARTER CONSUMER PANEL
Our data consist of a continuous and recent purchasing record of nearly 4,000 UK households reporting between 1999 and 2005 (six years) supplied by TNS UK. This panel is smaller than standard but the important difference is that it contains only continuous reporters rather than the usual sample with replacements, which makes it possible to examine repeat purchase at both brand and individual household level for the entire period. It is believed that a panel of this extent and nature has never been constructed before.
A range of quarterly marketing metrics has been extracted for 18 frequently purchased categories covering analgesics to vitamins, and including 140 well-known brands.
Our approach is one of replication and extension of empirical generalisation: to identify and describe known patterns observed in the data, benchmarking them with theoretical Dirichlet output. Categories have therefore been selected that differ widely in purchasing style, with average quarterly market penetrations ranging from 95 per cent to just 11 per cent, and average category purchasing frequencies from more than twice a week to less than once a month.
THREE KEY FINDINGS
1. Our first finding is that astonishingly, over six years, brand-share stationarity appears to be the norm. There is remarkably little change evident in our share data, even after 26 quarters of marketing investment.
Statisticians have developed many different approaches to measuring stationarity with more or less complex time-series analysis techniques. We believe simpler is better.
In this case a few points up or down are enough to tell the story, and using this approach we can show that, even after 26 consecutive quarters of trading, 75 per cent of our sample (105 brands), remained within just three share points of their average year one quarterly share.
A further 15 brands increased share by more than three points, but only seven from 140 increased share by more than six points; that is, by barely more than one point annually.
Table 1 identifies these brands, ordered by the magnitude of change, and includes the only two brands that lost more than six points in share over the period.
This is not entirely new. Many earlier studies have investigated share movements, including one which analysed the data in 400 academic journal articles to conclude that the effects of advertising and promotion on market share are not persistent (5).
What is new here is that our findings double the time scale: it is apparently normal for brand shares to remain approximately stationary for long periods, and certainly beyond five years.
2. What about growth? Our second finding is that exceptional, permanent, structural change in share does appear to have a common causality. It is strategic, achieved not through everyday manipulation of the promotional mix but by a major change to brand architecture, or a discontinuous innovation that demands a change in consumer behaviour.
During the period covered by our data, for example, the Unilever 'masterbrand' strategy led to the extension of Dove into personal care, and the withdrawal of the secondary brand Lux from the market. Felix cat food extended from cans to pouches and Kenco launched a new sub-brand, Rappor. Such strategies, although potentially significant to category structure, risk unacceptable levels of cannibalisation, and we discuss evaluation in the next section.
3. Our third finding is that the DJ relationship between penetration and purchase frequency holds in dynamic as well as in stationary categories.
This is illustrated in Table 2, with a six-year comparison of data from the instant coffee category. Individual brand shares have remained stationary from beginning to end, with the single exception of Kenco, which has gained eight points and moved up one place in the rank order through its Rappor extension. The table shows the familiar Dirichlet relationships in both first and last quarters. Brand shares and penetrations remain closely correlated; Nescafé has a 45 per cent market share and a penetration of 34 per cent, Birds has a 2 per cent share reaching 2 per cent of consumers.
On the other hand, although penetration and share values move in line, average purchase per buyer for any brand remains roughly constant at around two thirds the total category average (3.2), or twice per quarter.
Nescafé metrics are well above this average, but remain governed by the DJ law. (The constant, w (1 – b), is 1.8 in Q1 for both Nescafé and Birds, and about the same in Q26, reflecting how smaller brands predictably have fewer buyers who buy them less often).
Since large and small brands have roughly equal loyalties, then penetration must explain brand size. Kenco simply reached more buyers with the launch of Rappor. All other measures have remained stationary, so the share increase cannot be due to buyers drinking more coffee (in fact the data show a small decrease from 1.9 purchases to 1.8 in the last quarter), nor to increased loyalty, since SCR drops slightly too.
NO BRAND IS ABOVE THE LAW
Yet despite the fact that the market has so clearly evolved, the Q26 Kenco brand measures still conform to Dirichlet norms (and although not shown here, are in fact within a point or two of the theoretical output). Performance is entirely predictable, even five years ahead, along with the entire category structure.
Was Kenco successful? The target of the launch was Nescafé, but we can see that the share gain appears to have drawn from Maxwell House (a Kraft brand) too. Was this to be expected?
Using the duplication of purchase law, we note that Kenco has increased penetration, and therefore its duplication with other brands, but that the draw is disproportionate.
It should attract a third of Nescafé buyers but wins only a quarter of them, compensating with its own Maxwell House consumers. The cannibalisation is clear to evaluate.
Faced with the evidence of extended share stationarity some managers have said, 'Well yes, we thought as much all along …' (but were afraid to say?). Others, perhaps more cynical, have asked, as we do here, 'What's the point of marketing anyway?' The implications of our findings, however, are emphatically not that advertising is a waste of money or that marketing 'doesn't work'.
Rather, we think it is time to re-evaluate the use of share growth objectives; they are simplistic, short term and, in most cases, unrealistic.
We argue that the point of marketing in established categories (and that means most markets) is to maintain the share you already hold in the face of intense competition.
It is clear from our data that households remain polygamously loyal over extended periods, buying habitually from repertoires of brands they know well.
For each consumer some brands are infrequently bought, others more regularly so (think about the last time you bought a Bounty or Lilt), but the continuing role of brand marketing must be to keep improving customer value, to nudge that brand towards the top of each repertoire in the face of competitors' activities designed to do the same.
As the discussion of DJ shows, the vital factor is penetration. Purchase frequency or loyalty remains roughly similar from brand to brand.
PLAN IN TERMS OF CUSTOMER ACQUISITION, NOT RETENTION
The difference between brands of different size is the number of buyers they attract in a given period, and loyalty measures. Then simply follow in line with the DJ law.
Marketers should plan in terms of customer acquisition rather than retention, encouraging back some of the huge pool of experienced but light buyers they share with their competitors. This is working hard to stand still.
Although share may increase in one period it will invariably fall the next, remaining in approximate equilibrium over long periods.
What of permanent brand growth? It is rare because it involves a restructuring of the market through a permanent change in deeply ingrained habitual behaviour.
The risks of cannibalisation, failed innovation and loss of share are serious, and the evidence points to the fact that marketing is perhaps not as good at changing behaviour as it sometimes takes credit for.
In all this, the one certain factor is that market structure continues to be predictable from the Dirichlet even in evolving categories, and now over the long term.
That is very good news indeed. We suggest that a better knowledge of category structure can only lead to more effectively defined strategic objectives and benchmarks, and so to better marketing.
1. Results from a Havas survey, discussed in Ambler, Tim (2000) Marketing and the Bottom Line, 2nd edition. Harlow: Prentice Hall p.121.
2. Ehrenberg, ASC (1988) Repeat Buying 2nd edition. London: Charles Griffin and Company. These two laws usefully describe constant and therefore predictable relationships between common measures. The double jeopardy relationship between penetration (b) and purchase frequency (w) in a established category can be mathematically expressed as w(1-b) = a constant. Duplication of purchase captures the normal proportion of your customers' total buying going to each of your competitors. The percentage of buyers of brand A who also buy brand B (bB|A) can be compared with the percentage of buyers in the population who buy brand B (bB), and hence the duplication coefficient, D, calculated such that: bB|A=D x bB.
3. Ehrenberg A, Uncles M and Goodhardt G (2004) Understanding Brand Performance Measures: Using Dirichlet Benchmarks, Journal of Business Research, 37, 1307–1325
4. Goodhardt, Gerald J., Ehrenberg, Andrew S.C. & Chatfield, Christopher (1984) The Dirichlet: A Comprehensive Model of Buying Behaviour. Journal of the Statistical Society, A, 147, 621–655
5. Dekimpe, Marnik and Hanssens, Dominique (1995a) Empirical Generalizations about Market Evolution and Stationarity, Marketing Science 14. 3. (Pt. 2). 109–121.
ABOUT THE AUTHOR
Charles Graham is an associate of the Ehrenberg Centre for Research in Marketing and a senior lecturer at London South Bank University.
This is an edited version of the Academy of Marketing's Best Paper Award for 2009. From 2010 this will be sponsored by the Marketing Society as the Tim Ambler Award – for academic paper most helpful to practitioners.