Beruflich Dokumente
Kultur Dokumente
Chapter 9
Consumer panels
Denis Delmas
Dominique Levy
To structure the sample, we make use of the descriptive criteria for the
population mentioned in the previous paragraph. First, the principal
segmentation criteria are selected on which it is planned to stratify the
universe (for example region x type of residence x number of people in the
household). In this way cells are defined with known content. This determines
the segmentation of the sample. Each panel member recruited will correspond
to a designated cell. We then decide on the secondary criteria, i.e. those
characteristics of the population in question that we want the sample to
represent (children, age of the head of household, socio-economic group).
The last step in designing the sample is the sampling plan. Once the total
sample is divided into cells, the ideal number of panel members in each cell is
determined, generally with the use of a quota method. The average sampling
fraction is the ratio between the total population and the number of households
in the sample.
Consumer panels 277
permit very clear distinction between two items from a single range which
differ, for example, only by the colour or the perfume. Since the completion of
the diary is disconnected from the purchase itself, it is sometimes done 'by
memory', and can reflect consumers' normal habits rather than their real
buying trends.
It is equally clear that the use of the post service is only possible in countries
or regions where the courier system is functioning correctly. Even then, this
mode of communication involves a delay which can not be reduced ('posting'
by the panel member and carriage). Finally, once the questionnaires are
received by the institute, they still need to be sorted, which permits a human
control of the responses but involves extra cost.
Home scanning
Principle. The institute provides each household with an electronic scanner, an
apparatus which permits the optical reading of bar-codes and the recording by
the panel member of a number of details (quantity bought, price, place of
purchase). Bar-codes are directly read from the product packaging, when
available, and on a 'code book' (that is a collection of bar-codes with
descriptions of corresponding products) provided and updated by the institute
for use when the product does not have a bar-code. The collected information
is transmitted by telephone (the households are provided with a modem)
without any interaction from the panel member.
Strengths. The identification of products is precise (as is the codification by
bar-codes) and objective. Home scanning does not imply any pre-
determination of the categories to be audited, and permits the establishment of
a priori databases for non-subscribed categories. The co-operation of the
panel member is simplified and, for comparable products, its duration is
significantly reduced (compared with the diary book). The direct collection of
the data by telecommunication link between the institute computers and the
sample household means that there is no fear of any risks of forgetting to send
the information, or of delays in dispatch. This system has the advantage that
there is no need to type the information into the computer. Also, any changes
in the data to be collected can be effected simply by remote-controlled
modification 'of programs stored in the home scanner.
Weaknesses. Exploitation of the information given by home scanners depends
on the existence of a coding system. This requires the creation and
maintenance of a dictionary where all the bar-codes of the products being
purchased are listed, so the panel member can select the appropriate one. Each
bar-code is associated with a detailed description of the characteristics of the
product (category, segment, brand, name, size, special features, etc.). The
eventual quality of the panel will be influenced by the quality of this
dictionary which must be compiled before starting the panel and be updated
regularly. This represents one of the heavy investments necessary for setting
up a 'home scanning' panel, which also requires the acquisition, installation
and maintenance of each of the home scanners and the training for each of the
panel members. It is therefore obvious that the installation of such a panel is
materially complicated and very expensive. It also needs the establishment of
expert systems of validation of the returns from the panel members of a
particularly sensitive kind. These reasons, in addition to the requirement for
282 Denis Delmas, Dominique Levy
over the last three years (following the United States). New panels are likely
to be created in the years to come, especially in Asia, South America and
Eastern Europe.
At the data processing stage, questions may arise that are particularly sensitive
and critical, which (almost) every institute will resolve in its own way. There
is no question here of entering a technical debate on that subject, so we will
limit ourselves to mentioning the principal problems: the panel members'
holidays and the specific treatment (or editing out) of purchases which are
abnormally high.
The reading and analysis of panels thus implies using a number of indicators,
285
Principal indicators
a) The number of buyers, NA (NA = Nombre d'Acheteurs) corresponds to
the number of households that purchased the product at least once during
the given period. Penetration is the most frequently used indicator. We
talk of absolute penetration (NA brand/total number of households) or of
relative penetration (NA brand/NA market, i.e. percentage of households
purchasing the market that purchased the brand). The penetration
measures the size of the customer base, that is the proportion effectively
reached by the brand. Penetration gives a first diagnostic of the uptake of a
brand or a market.
The meaning of this notion varies markedly according to the length of the
period studied. Calculated over a short period, penetration gives an almost
photographic indication of the sales traffic generated by a brand. If this
information (monthly penetration for example) is compared with the same
figure obtained from a longer period, we introduce notions of frequency or
repetition. The graph of accumulating penetration over time gives us
information on how long it takes for a brand to reach its total customer
base. Comparison of this accumulated penetration with the penetration
cal~ulated for separate periods is a first indicator of repetition.
b) The average purchases per household. This may be expressed in terms of
volume (QA/NA, = Quantites AcheteeslNombre d'Acheteurs, i.e. the
average quantity bought per buyer), in consumption units (VC/NA, =
unites de consommation/NA, i.e. the average number of units purchased
per buyer, or in value (SD/NA, the average expenditure on the brand per
buyer). This indicator, whether it is called average frequency of purchase,
or level of consumption, measures the quantities purchased (or average
amount of money spent) per household for the period studied. The
transition from quantities to amounts spent is effected by means of the
average price, an indicator which is provided by all panels.
c) The quantities purchased (QA = Quantites Achetees) of a category or a
product are determined by the multiplication of the number of buyers
(NA) by the average purchases per buyer (QA/NA or SD/NA).
High levels of quantities purchased (QA) can be the result of a high
number of buyers (high NA) or a high average buying rate (QA/NA).
286 Denis Delmas, Dominique Levy
High levels of QA/NA can result from frequent buying, or buying which
is infrequent but of a high volume or amount. Therefore, they are not
predictive of frequency of purchase or of loyalty to a brand. To begin to
interpret them in that sense, they must be decomposed into two indicators:
QA per purchase occasion and number of purchase occasions.
c) The number of purchase occasions corresponds to the average number of
purchases in a household in a period, i.e. its purchasing frequency.
Depending on the collection method, there will be a number of purchase
occasions or a number of purchasing weeks (in those methodologies using
the diary method or the home audit). From this we will deduce the repeat
buying rate (percentage of buyers who purchase on at least two
occasions).
Clearly linked to the number of purchase occasions is the notion of the
purchasing cycle or purchase interval. This corresponds to the average
length of time (in days or weeks) between two purchase occasions in a
household. It is therefore a way to measure the familiarity of consumers
with a category or a brand (the more often they buy it, the more likely it is
that they will remember characteristics such as the price).
These indicators (frequency, purchasing cycles) are very useful for the
planning of certain promotional activities (duration of a game operation,
number of points to be collected, etc.).
e) The other variable which can explain quantities bought overall by the
household is the QA/purchase occasion (average quantity bought per
occasion) or the SD/purchase occasion (average amount spent per
occasion). This indicator is very useful for establishing a formulation
policy for the brand, e.g. what to include in a range, or the design of
promotional formats. It also provides a first approach to understanding the
level of involvement consumers have with a category or a brand (the
higher the amount spent on a purchase, the more we can suppose that the
consumer will have made a considered choice).
f) The last indicator commonly used is the share of requirements. It
measures the fidelity of consumers to a brand, and corresponds to its
relative weight within their purchases. It is calculated as QA/NA
brand/QA/NA category (i.e. the ratio of brand purchases to category
purchases) among those consumers who buy the brand.
Depending on the size of a brand's customer base, the size of its
competitors' universe and the level of consumption of its buyers, the share
of requirements can be read in very different ways. A high share of
requirements for a brand whose customer base is small and essentially
composed of light buyers of the category will suggest an atypical
positioning. At the other extreme, the same share of requirements for a
brand which enjoys a large customer base, and which recruits heavy
Consumer panels 287
Panel results are also sub-divided in accordance with the places where the
purchases were made. We then have a reading of the performance of a brand
(always using the same indicators) according to the type of retailer (store close
by vs. hypermarkets, specialised shops vs. general shops, mail order etc.) or
the store name.
Once the periods for study have been selected, together with judicious
segmentation criteria, the list of brands to be monitored and the most suitable
indicators, it is then possible to set up a standard reporting system, to be
reproduced for each period. This reporting system will permit the comparison
of brands, periods (for dynamic reading), sub-populations, types of trade or
store names. The objective of this sorting of the results is to identify the
important facts. These may involve elements which discriminate between two
brands (the weakness of my brand compared with a specific competitor is due
to our respective levels of penetration), relevant developments between two
periods (in such and such a period there was a break in the trend line for my
brand: what event can explain it?) or differences in performance according to
one criterion or another in relation to a type of population or trade (the
increase in my market share is accounted for by households with children. The
QA/purchase occasions of my competitor are especially strong in this or that
named store).
Another way to read the indicators of the panel is to use decomposition
schemes to highlight the characteristics of a brand. These permit us to quantify
the contribution of each indicator to the overall performance of a brand and to
direct the analysis accordingly.
The first and most commonly used scheme is to decompose the total quantities
(QA). It is extremely easy and can be written as follows:
QA = NA (number of buyers) x QA/NA (average quantity bought per buyer)
-
(Figure 1).
OAINA Figure
0.0- 1ft prilXi ( 1
~ =
0)
3
I
/
BRANDV(llJJ\£
j
\
1(1r&nl8q
( N?Pcm )
Consumer panels 289
--
It permits a first obvious interpretation. If we consider two brands with the
same buying volume (l00 litres for example), this volume can be reached in
very different ways.
Brand A might have a customer base of fifty households each buying two
litres, whilst brand B achieves the same sales with 200 households each
buying 0.5 litre (Figure 2).
Figure 2
QA = NA x QA I NA
NA
200
50
A
J
0.51 11 21 QA/NA
One can pursue this analysis by segmenting the QAlNAs using the following
equation:
QA/NA = Number of purchase occasions x QA/occasions
Returning to the above example, it will be seen that the sales of brand A are
achieved by fifty households, each with an average of four purchases of 0.5
litre each, whilst the 200 households of brand B each made an average of two
purchases with a volume of 0.25 litre per purchase (Figure 3).
Figure 3
QA/NA = QA I P. Occ x No. of Ps
No.of if
Purch.5
4 DA
2 3
Another approach to the decomposition of the indicators takes its source from
the market share, which is the most common measure of overall performance.
Market share can be read as the product of the following elements:
Market Share = relative NA X share of requirements X index of consumption
where:
Market share = QA brand/QA category;
Relative NA = NA brand/NA category;
Share of requirements = QA/NA brand/QA/NA category among buyers of
the brand;
Index of consumption = QA/NA category among buyers of the
brand/QA/NA category among all buyers.
Thus:
INDEX
SHAREOF
OF
(brand X X
(brand buyer)
buyer) CONSUMPTION
(all
OAiNA
OAiNA
OAiNAbuyers)
category
category
Figure 4
category
X MARKET
X 11
1 I~
The use of this decomposition diagram will logically lead the analysis to the
following problems:
o Weakness of the relative NB Recruitment problem;
Distribution (type of distribution);
Brand appeal (reputation, image,
level of promotional activity,
importance of the price, etc.).
292 Denis Delmas, Dominique Levy
One can see that reading the results of a consumer panel rapidly leads to
specific marketing problems. Through this reading, one is led to ask questions
about the brand, either its customer base or its mix, its competitors, the market
in which it is positioned. Whether these questions are motivated either by the
reading of the panel itself or as a result of events arising within the markets or
products, we can, to answer them, dispose of a wide range of tools and
methods in which standard results and special analyses will be mixed.
There follow some examples of questions, or marketing issues, and how a
consumer panel can answer them.
MARKETING ISSUES
The answers obtained from a consumer panel to marketing questions will be
the indicators that we have just described and also more specific approaches
such as special studies.
Intended to complete and to deepen the analysis of the standard results, a
special study is a 'segmentation' of these results. Technically, a special study
uses disaggregated data (that is data separated panel member by panel
member, purchase by purchase) and rearranges them according to criteria
which define a population and/or a period or a sequence of purchases.
There exist a certain number of studies of this type, which are used, with
variations, by all the institutes. One can nevertheless imagine what sort of
questions are asked. To bring about a special study one might, for example,
wonder who are the new buyers of a brand, or try to understand what the
consumers who are giving up a product are doing, or to identify the heavy
buyers of a category.
Consumer panels 293
It is thus the question that must precede the choice of a methodology. We have
tried to identify here the principal types of marketing issues, and name
examples of special studies relevant to them.
We will discuss in turn certain broad issues, which are:
o understanding the market (the product environment)
o the different aspects of the product's life (launch, customers, competition,
variables of the mix)
o the trade
trying, with actual examples, to explain the most common methodologies and
their practical use.
Who buys?
Once we know how many people buy and how frequently, the first question to
ask is: who are the buyers?
294 Denis Delmas, Dominique Levy
We will answer this question first by defining their profile, that is by detailing
the level of penetration and consumption of the category according to different
criteria for describing the population.
This analysis permits us to:
o represent the structure of the customer base and of the category and, by
comparison with the structure of the population, to identify the groups that
are under- and over-consuming
o measure the contribution of the same groups to the total volumes of the
category.
By doing this, we then know which are the types of households which are
under- or over-consuming, and if this level of under- or over-consuming is the
result of a level of penetration or of QA/NA.
In the following example (Figure 5) it can be seen that an over-consumption
of the category by people aged 65 years and older is the result of a strong
penetration and a level of QA/NA much higher than normal. Logically, this
profile could equally be the result of a high level of consumption in
households with two people, a portion of the population in which the elderly
are represented.
Figure 5
CATEGORY A: PENETRATION AND CONSUMPTION LEVEL
BY SOCIO-DEMOGRAPHIC GROUP
1pe"l 127.3
2 pe"l 130.3
3 pe"l -: 129.3
4pers + 1 127.4
+child'n 1 130.5
- ehBd'n I I 26.9
rIn the following example, we will consider that the 'heavy' buyers are the
20% of households for which the QA/NA are highest. Thirty percent of the
households following these form the medium buyers group, and the 50%
remaining will be the light buyers. We can equally fix definitions according to
the consumption level (heavy buyers = households that purchased X litres and
more during the period of study), or the number of purchase occasions. The
advantage of specifying these segments by percentages of buyers is that it
permits a homogeneous segmentation of one category (or a brand) compared
with another and makes it easier to set up the comparison and to establish
norms.
The first information produced by this type of segmentation is the relative
weight of each group of households in the total sales of the category. In our
example, the 23% who are the heaviest buyers account for 58% of the total
sales of the market (Figure 6).
Figure 6
CHEESE
Year 1996
NA: 23.2 % = 100 % QAlNA in gr
755 gr
225 gr
41
o LIGHT 625 gr
.MEDIUM
IililHEAVY
1,930 gr
Segmentation Segmentation
of customers of volumes
Besides the degree of concentration of the market, the study of light, medium
and heavy buyers allows us to describe the different groups in terms of:
o profiles
o buying behaviour (QA/NA overall, number of purchase occasions,
QA/purchase occasion, number of brands purchased)
o nature of the purchases (what other categories do they buy, what brands
within the category do they consume?).
This study allows us to identify precisely the heart of a target market and to
identify the elements of differentiation between an average and a heavy buyer.
296 Denis Delmas, Dominique Levy
Figure 7
MIX BETWEEN SEGMENTS
QA/NA QA/NA
A B
NAp100 66.8% 1.9 kg 4.8 kg
5.5%
1.9 kg
40.6 % I I I I 4.2 kg
I
16%
14%
r-
~
r-1
8%
11%
5%
14%
Ii r-1
,--,
10% -,
8%
12%
10%
8%
6%
4%
2%
0%
B c D E F G H
Established products
Questions regarding established products are many and varied. However, we
may consider that the principal ones are related to the following themes.
o How many buyers are there?
o Who are the buyers? Are they loyal? What else are they buying?
o What is the competitive universe of the product? Who are the most
dangerous competitors?
o How can we refine and make an evaluation of the different components of
the mix?
To build a customer base requires, in the first place, recruiting buyers and then
establishing customer loyalty. During the lifetime of a brand, its customer base
is enlivened by a continuous motion of renewal: at every moment the brand is
losing and gaining customers. The illustration in Figure 9 illustrates this
continuous flow.
Figure 9
THE BRAND IS IN PERPETUAL MOTION
As can be imagined, the aim of a brand is, at the same time, to maintain or
increase the level of water in the bathtub, and to limit leaks by introducing
enough 'fresh' water; in other words, to have a balance of gained buyers
against lost buyers and to keep its customers as long as possible. The
challenge is to keep recruiting new consumers (which is usually expensive)
and/or to increase the regularity or loyalty of the ones already attracted.
Depending on the maturity of the product, marketing efforts will tend towards
one or other of those objectives.
From this point of view, the consumer panel will, first, allow us to identify the
nature of the major problem of the brand (recruitment or loyalty). It will also
help us to understand the consumers' behaviour and thus identify the reasons
for any lack of regularity or loyalty, and the means to cure the problem.
We will start by measuring consumer flows. The simple reading of the
evolution of the cumulated penetration rate and of the repeat buying rate
permits an initial diagnosis. To improve on it, we very often use the analysis
of the loyal, lost, gained customers of a brand.
This study consists of two distinctive periods (two semesters, or two years for
example) and segments the customers into three groups:
o The loyals or renewers: they are the consumers that purchased in period 1
and in period 2.
o The lost: the consumers that purchased in period 1 but not in period 2.
o The gained: the consumers who did not purchase in period 1 but did in
period 2.
300 Denis Delmas, Dominique Levy
Figure 10
Period 1 Period 2
Brand X buyer
Competitor
buyer Competitor
buyer
Figure 11
YOGHURT
BRAND A: RENEWERS, LOST, GAINED
NA= 6.5%
NA= 5.8%
OAINA: 4.0 kg OAlNA: 3.1 kg
f-- - --t--
T1 T2
T1 = Year dating from first week; T2 = Year dating from first week
high renewal rate of its customer base and, above all, by a worsening of
QA/NA among the renewal buyers. In total, the brand is losing approximately
15% of its volume between the two periods.
We could complete the analysis by studying more closely the purchases of the
lost consumers (is their loss due to a total or partial withdrawal from the
market or a result of switching to other brands?), and of the retained
consumers. Regarding this last population, the question is to find out whether
the fall in their QA/NA is due to a reduction in frequency (perhaps
foreshadowing their defection), or in QA/purchase occasion, and if it is
expressed in a reduction in the brand's share of requirements, that is a drifting
away to other brands.
These two concepts: frequency (and regularity) and loyalty (or relative
importance of the brand in question amongst all purchases) are two essential
indicators of the solidity of the brand. To analyse them, we often use studies
of distribution.
The distribution of the frequency of purchasing consists of classifying the
buyers of the brand in accordance with the number of purchases made during
a period. It permits the identification of the occasional buyers, regular buyers
and assiduous buyers, and to measure their contribution to the sales volumes
of the brand. Depending on the stage of maturity of the product, a high share
of occasional buyers can be a sign of a period of recruitment or, in the case of
a product in decline, a sign of a weakness in the brand.
In either case, the analysis of purchase frequencies and their distribution is
important. A deficit of assiduous or regular buyers can lead to questioning the
policy respecting the formulations (packaging, etc.) or varieties of the brand,
or to organising promotional activities to encourage repeat purchasing
(collecting tickets, discount coupons) or the circulation of the buyers between
the varieties on offer, rather than instantaneous promotional mechanisms
which are better adapted to a problem of recruiting new buyers.
If the number of purchases of a brand by a consumer is a good indicator of his
familiarity with the brand, it is not directly linked to his loyalty to the same
brand. Loyalty is a relative concept, which will be measured by the part
played by the brand in the purchases of its customers. If a consumer buys
brand A very regularly but another brand even more regularly, it may be
considered that he is assiduous with regard to both brands, but loyal to neither
of them. On the other hand, a very occasional buyer of a brand who does not
buy anything else in the category will be an occasional but also a loyal buyer.
Loyalty is measured by the share of requirements, the distribution of which
will allow us to classify the consumers in terms of their degree of loyalty to
the brand.
The example in Figure 12 refers to a milk brand. We can see that the 20%
most loyal buyer group accounts for 52% of the volume of the brand, which