Sie sind auf Seite 1von 7

Toothpaste Purchasing in China: Patterns of Buyer Behaviour

Mark Uncles and Simon Kwok


University of New South Wales

Track 17 Continuation of the work of Ehrenberg and Bass

Background

Buyer Behaviour. Patterns of buyer behavior have been described, modelled and
used by Western researchers for at least fifty years. An understanding of these
patterns has developed hand-in-hand with the collection of consumer panel data
(Sudman and Wansink 2002, Bucklin and Gupta 1999) and with the refinement of
stochastic and econometric models (Massy, Montgomery and Morrison 1970,
Ehrenberg 1988). A vast body of knowledge has emerged, giving rise to numerous
practical uses (e.g., auditing the performance of established brands). In addition,
many implications for our understanding of consumers, brands and the marketing mix
logically follow from these frameworks (e.g., the typical consumer exhibits
polygamous buying behavior) (Ehrenberg, Uncles and Goodhardt 2004). Virtually all
of this knowledge has been derived from studies of Western consumers and Western
markets.

Chinese Buyer Behaviour. By contrast, there have been very few published studies of
Chinese buyer behaviour, or of Chinese consumer behaviour in general. An
exhaustive search of the literature for the years 1979-1997 (Sin and Ho 2001) shows
that out of 75 journal articles in consumer research on “Greater China” only two were
on choice and purchase in the People’s Republic of China (Chan and Lin 1992,
Samuel, Li and McDonald 1996). In no instance was use made use of consumer panel
data, and no longitudinal data analyses are reported. Similar conclusions are drawn
from another, wider-ranging, review (Ouyang, Zhou and Zhou 2000). It is not an
exaggeration to say that the formal and systematic study of buyer behaviour remains
in its infancy in China.

The Research Problem. There is a need to expand our understanding of Chinese


buyer behaviour, and to do so in a formal and systematic way. In particular, to see if
established Western patterns, models and findings generalise. For instance, in repeat-
purchase markets, is there a lack of segmentation between brands? Do Chinese
consumers exhibit polygamous buying behaviour or is there a tendency to be
exclusively loyal (over some stated time period)? The fact that our understanding of
buyer behaviour has generalised so well in the West does not necessarily mean it will
do so in China. Indeed, there are many profound differences. The importance of
cultural differences between China and the West has been stated time and again
(Hofstede 1991, Clark 1990). The need for multinationals to localise has been debated
at length (Levitt 1983, Usunier 2000). At this stage it is not possible to address all
these questions and issues. However, we make a start by reporting on preliminary
analyses of Chinese buyer behaviour, with a focus on the toothpaste market in Beijing
over the years 1999-2002.

Toothpaste Purchasing in China

2580
The Data. We use data provided by CVSC/TNS. This agency operates nationwide,
with household panels focusing on packaged-goods in 16 cities across China. All
purchase records for household panels are analysed separately for each of the years
1999 to 2002. The number of reporters ranges from 734 in 2002 to 851 in 2001.
Roughly 50% of the panel live in 3-person households and over time there is a clear
trend towards smaller families and away from household sizes of more than 3 people.
Using such data, basic brand performance measures can be extracted using the
BUYER software (Uncles 1989) and MATLAB-based macros (Rungie, Goodhardt
and Driesener 2003). The latter make use of maximum-likelihood estimation, in
contrast to earlier buyer behaviour software.

Descriptive Patterns. Initial inspection of the data shows that in the Beijing toothpaste
market Colgate and Zhonghua are the market share leaders for all four years, and
increasingly so, with combined shares of almost 50% in 2002. This is at the expense
of other leading brands in the market, which have mostly recorded declining market
shares. The only exceptions are LSL and some smaller brands with less than 2% of
purchase occasions. Colgate has significantly high penetration of between 76%-80%;
however, it only captures about 40% of its customers’ total purchases (with share of
category requirements of about 40%).

A number of patterns are evident across all years:


• The measures of brand size have much greater variance than loyalty-related
measures (consistent with prior knowledge).
• A brand’s penetration (i.e., the % of households buying the brand) varies in line with
a brand’s market share.
• There is a double jeopardy effect, in that brands with lower market shares not only
have fewer buyers (i.e., lower penetration) but their buyers also buy less frequently on
average (i.e., average purchase frequencies).
• The average share of category requirements (SCR) of a brand is only 24%. This
means, on average, a brand’s customers will buy other brands 76% of the time. This is
evidence of divided loyalty.
• On average, only 1% of a brand’s customers are 100% loyal (in a full year). This is
again evidence of divided loyalty.
• In fact, as the average portfolio size results indicate, households typically buy from a
repertoire of 3-4 brands over the course of a year.

Significantly, as we show next, these patterns appear to conform to predictions from


buyer behaviour theory.

Modelling Framework. Conventionally, stochastic and econometric models are used


to describe buyer behaviour. There are many models from which to choose. One
empirically useful model is the NBD-Dirichlet (or Dirichlet) (Ehrenberg, Uncles and
Goodhardt 2004). In this model consumers are seen to have, for the time being, steady
personal purchase propensities – or stochastic probabilities – for when they buy the
product and what brands they then choose. A feature of the model is that few inputs
are required, whereas a large number of brand performance measures are predicted
(e.g., % buying in a year, the average number of purchases per buyer, category
purchasing per buyer, share of category requirements, 100%-loyals, repeat-buyers,
lapsed-buyers). All these output measures are available for the Beijing toothpaste
data.

2581
Observed and predicted Dirichlet values of brand size and loyalty-related measures
are compared. Leading brands with market shares greater than 2.5% for each year are
analysed with all other brands grouped as “others”. This results in the same 9 leading
brands being analysed throughout the study, although their rankings varied between
the different years. In general, the observed values are closely predicted by the
Dirichlet model. This is evident from “eye-balling” the data and is shown by the
generally low scores for the mean average deviation across all measures (2002 data
are shown in Table 1). However, there are some deviations.

Table 1: Nine Leading Brands in the Beijing Toothpaste Market: 2002

Brand Market Penet-ration Ave SCR 100%


Share Purch Loyal
Freq
%O %P %O %P O P %O %P %O %P
Any 100 100 100 100 11.6 11.6 100 100 100 100
Colgate 32 31 79 80 4.7 4.4 41 38 5 3
Others 16 16 58 58 3.2 3.3 28 28 2 1
Zhonghua 16 15 52 54 3.6 3.1 29 27 2 1
Crest 8 8 36 36 2.7 2.7 23 23 1 1
LMZ 8 9 41 39 2.2 2.8 21 24 2 1
Bluesky 7 7 32 32 2.4 2.6 22 23 2 1
Signal 4 4 21 20 2.2 2.5 19 21 1 1
LSL 3 3 16 16 2.4 2.4 21 20 1 1
Jieyin 3 3 16 16 2.3 2.4 22 20 2 1
Heimei 3 3 13 13 2.3 2.3 19 20 1 1
Average 10 10 36 36 2.8 2.9 24 25 2 1
MAD 0.01 0.01 0.21 0.01 0.01

Patterns of toothpaste purchasing have been reported in earlier work (Ehrenberg,


Uncles and Carrie 1994; Kau, Uncles, Ehrenberg and Barnard 1998). Therefore, we
are able to compare the Beijing figures with those from other markets and other time
periods – see Table 2. Importantly, the patterns do not differ from one another in any
noticeable ways.
Table 2(a): Six Leading Brands in Four Toothpaste Markets: Penetration

Brand China (2002) UK (1991) US (1985) Japan (1983)


%O %P %O %P %O %P %O %P
Any 100 100 84 84 71 71 95 95
1st 79 80 34 37 41 41 70 74
2nd 52 54 22 21 27 27 31 27
3rd 36 36 22 21 23 23 26 23
4th 41 39 20 18 9 9 15 16
5th 32 32 11 12 6 6 13 10
6th 21 20 11 11 4 4 3 2
Average 44 44 20 20 18 18 26 25

Table 2(b): Six Leading Brands in Four Toothpaste Markets: Average Purchase Frequency

Brand China (2002) UK (1991) US (1985) Japan (1983)


O P O P O P O P
Any 11.6 11.6 7.6 7.6 4.3 4.3 9.8 9.8
1st 4.7 4.4 4.1 3.8 2.6 2.6 5.4 5.1

2582
2nd 3.6 3.1 3.2 3.3 2.2 2.2 2.7 3.1
3rd 2.7 2.7 3.1 3.3 2.1 2.1 2.7 3.1
4th 2.2 2.8 3.0 3.3 1.8 1.9 3.1 2.9
5th 2.4 2.6 3.3 3.1 2.0 1.9 2.1 2.8
6th 2.2 2.5 3.2 3.1 1.9 1.8 1.7 2.6
Average 3.0 3.0 3.3 3.3 2.1 2.1 3.0 3.3

Table 2(c): Six Leading Brands in Four Toothpaste Markets: Share of Category Requirements

Brand China (2002) UK (1991) US (1985) Japan (1983)


%O %P %O %P %O %P %O %P
Any 100 100 100 100 100 100 100 100
1st 41 38 45 40 49 49 50 47
2nd 29 27 34 34 36 39 27 26
3rd 23 23 35 34 31 37 27 25
4th 21 24 33 33 28 32 27 24
5th 22 23 36 31 27 31 18 22
6th 19 21 34 31 28 30 14 21
Average 26 26 36 34 33 36 27 27

Extensions and Further Work. Only some of the results are reported here. The
Dirichlet model also can be used to compute other important brand performance
measures, such as frequency distributions (i.e., the percent of buyers of a brand who
buy once, twice, three-times, etc.) and the duplication of purchases between brands
(i.e, the percent of buyers of one brand who also buy named other brands). And do so
for different length time-periods (e.g., for each quarter). Repeat-buying computations
can be derived from the NBD model (i.e., quarter-to-quarter repeat-buying and rates
of purchasing by new, lapsed and repeat-buyers). There is also some scope for
regression-type analyses – with possible independent variables being household
characteristics (income and size) and purchase attributes (purchase quantity, price
paid, package size, on/off promotion, store type).

In many respects this is a highly developed area of research. The modeling framework
is tried and tested. Yet there are many unresolved issues that deserve further
investigation. Further analysis of the Chinese data will greatly assist in this process.
Three areas are identified here: (a) extensions to new conditions, (b) deviations from
the Dirichlet model, and (c) elaborating the Dirichlet model.

(a) Extensions to new conditions. The Dirichlet provides benchmarks when analyzing
data for another year, country, or product category. Instead of unfocused data-mining,
it is simpler to check whether the Dirichlet patterns recur. This study illustrates the
value in comparing Chinese data with patterns that have been found to hold generally
in the West. Further extensions are envisaged in terms of looking at different product
categories (traditional Chinese goods, as well as packaged-goods), choice of stock-
keeping units, and store / store-type choice. There is also the question of how
repertoire and subscription markets might differ in Chinese markets (Sharp, Wright
and Goodhardt 2002).

(b) Deviations from the Dirichlet model. Discrepancies in the model predictions occur
and they need to be explained (Bhattacharya 1997). Some of these turn out to be
idiosyncratic effects, which are unlikely to recur. Nevertheless, some systematic

2583
discrepancies have been reported (e.g., for some market leaders, annual purchase
frequencies are a unit or so higher than predicted). A question is whether these
systematic deviations appear in the Chinese data. And are there any further deviations
to report? If so, can they be explained?

(c ) Elaborating the Dirichlet model. The Dirichlet, and related stochastic approaches
to modelling buyer behaviour, have been elaborated in two main ways. First, case-
specific “mass points” have been used to model consumer heterogeneity, rather than
the Dirichlet’s smooth Gamma and Beta distributions (e.g., Chintagunta, Jain and
Vilcassim 1991). There is scope to develop these studies, however this must be
achieved without forsaking parsimony and generalisability. Second, consumer
heterogeneity has been related to various possible causal sources such as socio-
demographic and attitudinal factors (e.g., Fader 1993). Further work on this, using the
Chinese data, is proposed. Apart from the empirical motivation to do this, there are
theoretical reasons in that with these models linkages can be found between the
Dirichlet and logit/logistic models (Fader 1993, Rungie 1999). This sees the
stochastic modeling and econometric modeling traditions as theoretically related,
rather than opposed to one another.

The preliminary results shown here are limited, nevertheless they are indicative of the
value of a more formal and systematic approach to the study of buyer behaviour in
China.

Acknowledgements

We are grateful to CVSC-TNS Consumer Panel Research for the Chinese data,
particularly the assistance of David Fell at TNS and Jiang Tao at CVSC. Comparative
data for the UK, US and Japan are from Ehrenberg, Uncles and Carrie (1994). Note:
category penetration is a 100% in the Chinese data, which may be a feature of the way
data have been provided, rather than a true statement of toothpaste penetration levels.
If so, this would impact absolute penetration values but not the patterns.

References

Bhattacharya, C.B., 1997. Is your brand’s loyalty too much, too little, or just right?
Explaining deviations in loyalty from the Dirichlet norm. International Journal of
Research in Marketing, 14 (5), 421-435.

Bucklin, R. and Gupta, S., 1999. Commercial Use of UPC Scanner Data: Industry
and Academic Perspectives. Cambridge, MA: Marketing Science Institute.

Chan, T.S. and Lin, G.S., 1992. An empirical analysis of consumer decision processes
in the Republic of China. Journal of International Consumer Marketing, 4 (4), 33-49.

Chintagunta, P.K., Jain, D.C. and Vilcassim, N.J., 1991. Investigating heterogeneity
in brand preferences in logit models for panel data. Journal of Marketing Research,
28 (November), 417-428.

Clark, T., 1990. International marketing and national character: A review and
proposal for an integrative theory. Journal of Marketing, 54 (October), 66-79.

2584
Ehrenberg, A.S.C., 1988. Repeat-Buying: Facts, Theory and Applications (2nd
edition). London: Edward Arnold (Griffin).

Ehrenberg, A.S.C., Uncles, M.D. and Carrie, D., 1994. Armed to the teeth: an
exercise in brand management. Cranfield, UK: European Case Clearing House
(reference M94-005:594-039-1/594-039-4/594-040-1/594-040-4)

Ehrenberg, A.S.C., Uncles, M.D. and Goodhardt, G.J., 2004. Understanding brand
performance measures: Using Dirichlet benchmarks. Journal of Business Research
(forthcoming).

Fader, P., 1993. Integrating the Dirichlet-Multinomial and Multinomial Logit models
of brand choice. Marketing Letters, 4 (April), 99-112.

Hofstede, G., 1991. Cultures and Organisations: Software of the Mind. Berkshire,
England: McGraw-Hill.

Kau, A.K., Uncles, M.D., Ehrenberg, A.S.C. and Barnard, N., 1998. Competitive
brand-choice and store-choice among Japanese consumers. Journal of Product and
Brand Management, 7 (6), 481-494.

Levitt, T., 1983. The globalisation of markets. Harvard Business Review, 61 (May-
June), 92-102.

Massy, W., Montgomery, D. and Morrison, D., 1970. Stochastic Models of Buyer
Behavior. Boston, MA: MIT Press.

Ouyang, M., Zhou, D. and Zhou, N., 2000. Twenty years of research on marketing in
China: a review and assessment of journal publications. Journal of Global Marketing,
14 (1/2), 187.

Rungie, C., 1999. Heterogeneity in Brand Choice. University of South Australia:


Doctoral Thesis.

Rungie, C., Goodhardt, G. and Driesener, C., 2003. Research note: calculation of
brand performance measures from the parameters of the Dirichlet model. School of
Marketing, University of South Australia.

Samuel, S.N., Li, E. and McDonald, H., 1996. The purchasing behavior of Shanghai
buyers of processed food and beverage products: Implications for research on retail
management. International Journal of Retail and Distribution Management, 24 (4),
20-28.

Sharp, B., Wright, M. and Goodhardt, G., 2002. Purchase loyalty is polarized into
either repertoire or subscription patterns. Australasian Marketing Journal, 10 (3), 7-
20.

2585
Sin, L.Y.M. and Ho, S.C., 2001. An assessment of theoretical and methodological
development in consumer research on Greater China: 1979-1997. Asia Pacific
Journal of Marketing and Logistics, 13 (1), 3-42.

Sudman, S. and Wansink, B., 2002. Consumer Panels (2nd edition). Chicago, Il:
AMA.

Uncles, M.D., 1989. BUYER: Buyer behaviour software. London Business School,
London.

Usunier, J.-C., 2000. Marketing Across Cultures (3rd edition). London: Prentice Hall.

2586

Das könnte Ihnen auch gefallen