1. BACKGROUND

An analysis of twenty-three successful new brands or line-extensions has shown an unexpected but clear finding: their purchase frequency is almost instantly normal. This means that loyalty to a new brand is from the start at the same level as for established brands.

Traditionally, a new brand’s repeat-buying loyalty is thought of as growing slowly over time (e.g. Franzen 1999 and many earlier references). But some practitioners with new-brand experience seem to have views (unpublished) which are much closer to the “instant-loyalty” one that is put forward here. (We are planning to survey opinions on this more formally.)

The growth of loyalty has often been modeled in terms of consumers’ so called “depth of repeat”, i.e. increasing with whether a consumer had already repeat purchased the item once, twice, three times, etc. (e.g. Fourt and Woodlock 1960, Baum and Dennis 1961, Eskin 1973, and Kalwani and Silk 1980):

One thing is certain – there is no rule about the level of repeat purchasing to be expected at different levels of penetration (Parfitt and Collins, 1968).

We expect there will often be a period of instability during which consumers’ preferences for the new product are evolving (Fader and Hardie, 1999).

But despite 25,000 new products a year being launched in the US (Economist 1998), there is little systematic knowledge about how loyalty to new brands does actually develop (e.g. see Hardie et al 1988, Wright and Sharp 1997).

Our Study

We have therefore sought to examine how newly-launched brands or brand extensions of frequently-bought consumer goods and the like perform in terms of a variety of repeat-buying, brand-switching and attitudinal brand performance measures.

The approach was explicitly exploratory, since for new brands we basically did not know what we were looking for. But for established brands’ in-market performance, extensive generalisable empirical results and theory have long been available (e.g. Ehrenberg 1959, 1972/88, Goodhardt et al 1984, Stern 1995, Stern and Ehrenberg 1995, Sharp 1999). This provided potential norms and a possible conceptual basis to evaluate the performance of new brands.

The work was for newly-launched brands which were broadly “successful”, i.e. available for retail sale for at least a year or two. New brand failures have so far been outside the scope of the analyses here, since detailed in-market data for them is more sparse. The box below however shows how varied the causes of new-brand failure can be.

In the event, our results began to suggest a much simpler outcome: near normal loyalty for new brands almost instantly. This led to a radical re-focusing of our further goals. Instead of having to research how differing marketing-mix and extrinsic factors would influence new-product performance, we could reverse the process to ask how our simple new-product finding might help to explain various marketing-mix issues.

Methods 

We needed to obtain data on brand performance measures for new brands and competitive established brand (e.g. their market shares, penetrations, average purchasing frequencies, share of category requirements, the incidence of 100%-loyal buyers, etc.).

At this early stage we sought data over quarterly time-periods as a compromise between periods too short to observe much repeat-buying and too long to track the new brand’s changing development over time. Sometimes data for 3 or 4 such quarters from launch were available, sometimes more. We partially discounted the first quarter in each case, since its results would tend to be confused by the precise date (or dates) of the launch.

Table 1 lists the nine product categories in which we have a total of 23 new-brand cases or line-extensions, and data on almost 100 established brands as possible benchmarks. All but one of these cases gave uniformly the same instant-loyalty result. (The exception being on much the smallest sample base and not statistically significant.) The high consistency of the main finding suggests that selection bias of the category is not an issue. Nonetheless, extensive follow-up work is now being pursued.

2. Results

Consumers’ purchasing of given brand X in an analysis-period like a quarter of a year can be broken into two main factors:

1. The Penetration: The percentage of consumers who bought X at least once in the period.
2. The Purchase Rate or average purchase frequency, i.e. how often on average buyers of X bought it in the analysis period.

The “Sales Equation” is that Total Purchases (or Sales) = Penetration × Average Purchase Frequency.

The main finding is that a successful new brand’s average purchase rate was at or near its subsequent “normal” level and also equal to the rates for established brands, virtually from the start.

In contrast, the new brand’s quarterly penetration typically increased initially, (sometimes very much so for a “real winner”). The number of buyers in a quarter then usually leveled out and/or dropped again to a lower steady level (see Davis 1964).

2.1 An Example: Prozac

The new anti-depressant Prozac has been very successful since its launch in the early 1990s. (The data are from an ongoing panel of some 200 UK GPs who recorded all their new prescriptions week by week). Table 2 shows that the percentage of doctors who prescribed Prozac at least once in a given quarter – i.e. the new brand’s quarterly penetration – rose some 20-fold over the first two years, from 1% in QI (the launch quarter) to 21% two years later.

In contrast, the average number of new Prozac prescriptions written in a quarter per prescribing doctor was at a mostly rather steady level of round about 2.3, after the typically mixed-up “launch” quarter. The somewhat higher prescription rates at 2.9 in Quarters VI and VII may merit further in-depth study but are not typical for our other data. (They may have been due to promotional activities, or something like a flu epidemic.) Overall, Table 2 shows that repeat-buying loyalty was almost instantly at its longer-term level.

2.2 The Various New-Brand Cases

The similar outcome for 22 of our new brands is summarized in Table 3 for Quarterly Purchase Rates and in Table 4 for Penetrations (one detergent case is excluded, as mentioned). For conciseness, the two tables show the first three quarters and the last one available to us, this result being typical also of the intermediate quarter.

The observed buying rates in Table 3 were essentially steady from the first full post-launch Quarter II onwards, about 1.9 on average. They are also close to the established brands’ benchmark figures averaging at 1.9 in the last column. This was so far really new brands like Prozac as well as for line-extensions such as the new liquid detergents in our data-bases.

The leading established brands used for comparison here had however much larger market penetrations than most of the new brands, as was to be expected. The established brands therefore also had relatively high purchase rates – the well established Double Jeopardy (DJ) phenomenon (e.g. Ehrenberg, Goodhardt, and Barwise 1990, Ehrenberg and Uncles 2000).

The last column of figures in Table 3 has therefore been adjusted for DJ, using the standard Dirichlet model to re-estimate the established brands’ purchase rates corresponding to the new brands’ average market shares (Ehrenberg 1988). The adjustments are however small and hardly affect the comparisons (e.g. the average purchase rate for the Detergents cases for instance reduces from an observed rate of 2.2 down to an adjusted one of 2.0).

Quarterly market penetrations of the new brands however increased during the launch periods, from effectively zero before the launch to an average of 4% in Quarter II, as shown in Table 4. The penetrations then either more or less stabilized (as is quite common for successful new brands, e.g. Davis 1969), or in the case of the two medical blockbusters went on increasing greatly to 22% of the population (here “Prescribing GPs”). This highlights the relative stability of the purchase rates in Table 3.

2.3 Other Loyalty Measures

The average frequency with which a brand is bought by its buyers in a period such as a quarter is one measure of its customers’ “loyalty”. It has been widely shown for established brands that this correlates closely with other measures relating to loyalty, such as the incidence of 100%-loyal buyers, the brand’s shares of category requirements, and the levels of period-to-period repeat-buying (e.g. Ehrenberg 1988, Ehrenberg and Uncles 2000).

For our new brand cases here, the corresponding results also largely hold, except for quarter-by-quarter repeat-buying. At 24% this is much less than the norm for established brands (about 50%). Period-to-period repeat-buying would however be especially sensitive to the occurrence of once-only triallists who are especially likely with new brands. This can be readily checked once suitably detailed data are available, by using so-called “Conditional Trend Analysis” (Goodhardt and Ehrenberg 1967, Morrison 1969, Ehrenberg and Uncles 2000). “CTA” analyses light and heavier buyers separately, with interpretative theoretical norms.

2.4 Previous New-Brand Cases

Some isolated new-brand cases giving instantly normal average purchase frequencies had already been reported previously (Ehrenberg and Goodhardt 1967; Wellan and Ehrenberg 1988; Wright and Sharp 1997). But this was always discounted, as an unexpected aberration: “The fact that the new brand looked like an existing brand so quickly is a little curious” (Wright and Sharp 1999). The received wisdom was, as we have noted, that new brand loyalty had to develop slowly: Only now, with a range of 20 cases all giving the instant-loyalty outcome, can one see that the earlier isolated cases were prophetic, rather than aberrations.

Supportive evidence also comes from the Unilever / RBL / Research International “MINIVAN” test panel operation. This was a mobile grocery shop which signed up a panel of housewives who could conveniently shop there once a fortnight as one of their retail outlets but under experimentally controlled conditions (e.g. Charlton et al 1972).

The Minivan attracted a good deal of research interest from manufacturers at the time, but was ultimately abandoned. It rather accurately predicted a new brand’s real-life repeat-buying but failed to predict its crucial market penetrations. That is in line with the new result here, that normal repeat-levels are already achieved instantly by new brands, and hence should have been easy to predict.

More generally, we understand that successful commercial volume-forecasting services nowadays tend simply to use the NBD/Dirichlet model or related results to predict the test-brand’s repeat-buying. The more difficult-to-predict key to success is the new brand’s ultimate market penetration i.e. the number of new customers.

3. WIDER IMPLICATIONS

The new-brand finding has implication for some half-a-dozen wider marketing issues which we have already been able to explore.

Brand Loyalty

We have previously argued that any loyalty measure for a successful new brand must ultimately settle down to a typical norm for the product category (e.g. Ehrenberg 1991). This has had powerful planning implications (e.g. see Ehrenberg-Bass Institute Report 1 for Corporate Members). The new “instant loyalty” finding for new brands now greatly strengthens this conclusion.

Brand Penetration and Salience 

More generally, we have been arguing that what distinguishes one brand from another is whether the brand is “salient” to the consumer, i.e. is in his/her “consideration set” (e.g. Ehrenberg et al 2000). “Instant loyalty” supports this because all that can matter to a brand’s success is to how many consumers the brand is salient.

Brand Differentiation

The traditional view of competitive advantage is that a brand must be differentiated: consumers, it is thought, need a ‘reason’ (functional or emotional/image-related) for choosing brand A rather than brand B.

Our contrary view has long been that competitive advantage is soon copied and hence not sustainable: in practice, brand competition consists of matching your rival on what matters rather than of being different (e.g. see Dawes and Page 2000, Ehrenberg et al 2000, and earlier references there). This view is now strengthened by finding that even loyalty to new brands is normal, implying that they are also not seen as different.

Brand Equity 

A common view in marketing is that brands possess different levels of “equity”, i.e. that brands are either “strong” or “weak”?

We have long argued the contrary view: that brands are primarily just either large or small, and that this depends only on their number of buyers and not on differences in their buyers’ loyalty or commitment (e.g. Ehrenberg 1983).

The reason is that there are essentially no such differences in loyalty after allowing for the market-share-related Double Jeopardy effect (Ehrenberg, Goodhardt and Barwise, 1996). This is again supported by the new finding.

Segmentation 

New brands’ having the same degree of loyalty as established brands also fits in with the finding that similar brands appeal to similar kinds of consumers, i.e. that competitive brands do not “segment” the market (Kennedy and Ehrenberg 2000).

Attitudes 

Attitudinally, users of brand A look at A much like users of the competitive brand B look at B (e.g. Barnard and Ehrenberg 2000 and previous references there). We now need to establish whether or not users of a new brand also view it in the same way.

Price Promotions 

Temporary price-cuts for established brands (i.e. “deals” or “price promotions”) appeal virtually only to past customers (Hammond, Ehrenberg, and Goodhardt 1996, Hammond, Ehrenberg and Long 2000). Price-promotions therefore do not recruit new customers to established brands. However, for new brands, the mechanism must differ since there are no past customers. This needs to be explored further.

Advertising

Received wisdom is that advertising is mostly persuasive, e.g. that advertising motivates the consumer to become loyal to the brand. We have however been arguing that this is not so (e.g. Ehrenberg et al 2000). We believe that the new-brand finding of instant loyalty further undermines the traditional “persuasive” views and supports the publicity perspective.

4. DISCUSSION

The research outcome that successful new brands have normal loyalty virtually from the start seems sufficiently simple for companies to incorporate it fairly quickly into their new product planning and evaluation (or to reject it) and for academics to discuss it in their relevant teaching and research.

Although largely unexpected, the finding now makes much sense, with hindsight. Thus we believe that experienced consumers know that competitive or substitutable brands are similar and hence substitutable, which is why the brands are in fact competitive. No extensive “learning” about the new brand’s attributes is therefore needed: the consumer either chooses it as a normal “repertoire brand,” or not.

Possibilities for follow-up research include more systematic and in-depth studies of successful new brands in other product categories and countries, including in emerging markets, and also studies of unsuccessful new brands. The task of collecting suitable data is eased now that one knows what to look for.