Increased sales for a brand tend to show up more as additional customers than as additional purchases by existing customers. But until now, stressing customer acquisition for growth has only been an extrapolation from the familiar cross-sectional “Double Jeopardy” phenomenon (where at one given point in time, bigger brands usually have much higher penetration than smaller brands, but only slightly higher purchase rates – Ehrenberg-Bass Institute Report 26 for Corporate Members.

Recently, however, there has been some direct empirical confirmation of acquisition for growth. For doctors’ prescriptions, Prozac’s vast 20-fold growth over 8 years showed up as a 10- fold increase in the number of doctors prescribing it (from 6% to over 60%) with only a doubling of their annual prescription frequency (Ehrenberg-Bass Institute Report 30 for Corporate Members). A similar but less dramatic pattern has been quoted for a major US dairy brand over 4 years (Anschuetz 2002). And extensive Canadian fmcg data has shown that customer acquisition (or loss) predominated over loyalty shifts for sales changes over a 5-year gap (Baldinger et al 2002).

But these long-term results still leave uncertain the nature of shorter-term and smaller sales trends, where even a growing brand usually remains much the same size (e.g. a 6% penetration growing to a steady and very welcome 8% in a following year, not Prozac’s stunning 60% after several years). For such smaller share differences, the well-established cross-section DJ relationship has too much local scatter to be extrapolated confidently to growth (or decline) over time. We therefore need to examine shorter-term (say annual) gains or losses directly.

New empirical evidence has now confirmed the overall emphasis on customer acquisition in 157 cases of annual brand-share changes (79 gains, 78 losses) in 20 fmcg product categories from the TNS Impulse and Superpanel data in the UK. This is summarised in Table 1, for large, medium and small brands with increasing or decreasing shares.

The emphasis on acquisition shows up even for most of these year-by-year share increased penetration by 31% on average, three times as much as their increase in purchase frequency (10%). Brands that lost share similarly dropped twice as much in penetration (-20%) as in purchase rate (-10% again). These ratios (3 to 1 for gains, 2 to 1 for losses) vary between cases, partly, as illustrated in the table, for arithmetical reasons because of the size of the initial customer base: smaller brands show larger percentage changes. (The table omits seven very small brands with exceptionally large gains from a near zero base.)

In absolute terms, these annual changes were typically not very big. On average almost 3 percentage points for penetration gains in Table 1, about 2 for losses. And less than 1 purchase for the frequencies (on average only 0.5 for gains with none for the large brands, 0.9 for losses).

Table 1: Year by year changes in Share, Penetration and Purchase Rates (Absolute and Percent changes)

There are deviations from the overall pattern. In a fifth of the cases, purchase frequencies varied proportionately more than the penetrations. Nearly two-thirds of these occurred in a few exceptionally heavily-bought categories with much multi-brand buying (e.g. pet foods, carbonated drinks and confectionery); in such cases, purchase frequencies can vary over time without penetration changing much. Some of these deviations could also be due to specific marketing actions such as extended 2- for-1 promotional offers, or the occasional larger sampling error. But these remain exceptions. The longer term evidence (e.g. Prozac) implies that over successive years, the predominance of penetration growth would reassert itself.

Consumers’ buying-behaviour patterns help to explain why over time penetration would tend to grow more than loyalty (e.g. Ehrenberg-Bass Institute Report 1 for Corporate Members). In particular, the possible growth in customers’ rates of buying a brand is more limited than the possible growth in penetration, because a brand’s share of category usage (its SCR) is already higher than its share of category users (its penetration), for almost all except very large brands.

The Basic Role of Purchase Propensities

Despite changes in penetration being more common, marketing action should not solely be aimed at penetration growth. The underlying mechanism of buyer behaviour is telling here. Thus it is well-established that individual consumers have their own personal propensities-to-buy each brand (probabilities in the Dirichlet model – e.g. Ehrenberg-Bass Institute Report 1 for Corporate Members). These propensities will be at or near zero when a brand is never or rarely bought by that consumer, higher for her more popular brands, all adding up to 1 if the consumer is in the market at all. In a chosen analysis period such as a year, these ongoing propensities then show up as Penetration or Reach (how many people bought the brand at all) and also as Frequency (how often they then bought it).

Marketing inputs will affect some of these propensities to buy. Any such change would then necessarily be reflected in changes in both penetration and average purchase frequency from one period to the next. The results in Table 1 merely showed that the number of customers generally changes proportionately more than the purchase frequencies, although both do usually change.

Normal Buying Patterns Each Year

This interpretation is confirmed by the finding that in the new data, the starting- and end-position for the short-term sales changes tend to follow the normal steady-state (cross-sectional) Dirichlet patterns of buying behaviour.

Table 2 illustrates this: in each separate year, penetrations and purchase frequencies were overall just about as predicted by the steady-state model. The average fits are very close, as also for most of the 157 individual cases. This supports the rather simple hypo-thesis that the annual measures of penetration and loyalty move together, both for gains and losses, rather than that loyalty change either precedes or follows penetration change. It is consistent with the changing purchase propensities being the driver.

Table 2: Observed and Theoretical Measures in each Year (Overall, and typical examples of market-share changes)

There are, as always in such data, occasional deviations. One is illustrated here for PG Tips Pyramid Bags, a highly successful line extension at the time. In this case, the observed Year 1 penetration (O = 17%) was higher than the theoretical prediction (T), and the observed purchase frequency (3.2) was comparatively low. This could have been due to patchy retail availability in the first Year 1 (i.e. soon after launch). In Year 2, the brand had come into line.

Brand Switching Between the Two Years

A start is also being made on establishing where the individual gained or lost buyers might come from, or go to. The results so far are simple: competition still follows the normal steady-state patterns. Purchase duplication between any two brands for two such “dynamic” years varies just with each brand’s size. But together with the traditional clustering for variants of the same brand (implying some cannibalisation). This is in line with the long-established “Duplication of Purchase Law” in steady markets (e.g. see Ehrenberg-Bass Institute Report 1 for Corporate Members).

The fact that market shares have changed therefore does not seem to affect brand choice behavior disproportionately.

Table 3 illustrates these patterns for UK tea bags. Lyons Tetley was bought in Year 2 by 40% of category buyers and by about 40-odd% of the customers of each of the other three brands in the previous year. Similarly, the “Other” variants of PG Tips were bought by only some 10% or 12% or so. However, the previous customers of these earlier “other” versions of PG Tips switched more to its new Pyramid variant (50%).

Table 3: Purchase Duplication Between the Years

This is the kind of within-brand clustering that also occurs in steady markets.

Customer retention (repeat-buying from year to year) is also consistently higher than duplication or switching between the brands, as usual in more or less steady markets. Thus, competition still followed the normal patterns even when market shares changed between years.

Targeting Goals

The empirical evidence and the underlying theory imply that one can set realistic goals for how an increase in sales should normally show up in consumer terms: there should be an expected number of extra customers and also a lesser expected increase in their loyalty. And similarly for sales losses, a predictable number of customers would normally be lost and those who remain would be buying the brand some-what less often. Table 2 shows that the metrics from Year 1 to Year 2 can be quantified: how many more customers and how much extra loyalty. For Comfort Fabric Conditioner, for example, a 2-point increase in market-share would be expected to show up as a 5-point increase in penetration (12% to 17%) and a 0.4 increase in average purchase frequency (2.8 to 3.2).

However, we distinguish such specifiable consumer goals from how the marketing input itself might be competitively formulated or targeted (the advertising, pricing, merchandising, product development, etc.).

In an extreme case, the campaign objectives might be entirely to attract new customers, or (conversely) entirely to lock in current customers with loyalty incentives. But the outcomes of such apparently specific targeting are not usually so selective. A successful loyalty scheme for example is likely also to attract newcomers (e.g. by simply providing publicity, or the promise of rewards, or acting as a testimonial).

As marketers, we seldom know how far a marketing input will in practice convert, or nudge, or merely maintain. But the implication of the findings here is that whatever the form of the marketing effort, it should seek to increase some or all consumers’ underlying propensities of buying the brand. This will usually have more acquisition than loyalty consequences.