Post on 18-Jan-2017
Dr Adam Fenton
Brasil Brau, June 17th 2005
What do you like ? Why do
you like it ?
Why are we here ?• Measurement and analysing
preference• Brand profiling• Preference Mapping• Links to brewing • Some Examples
Intro
duct
ion
Why perform consumer research?
Why
Per
form
Co
nsum
er
Rese
arch
?
Consider a ‘boutique brewery’:-
Uses ‘Conversational’ consumer research.As a brewing company becomes bigger…..So the number of consumers increases ……Requiring more advanced techniques
Why
Per
form
Co
nsum
er
Rese
arch
?
• Product (range, features, quality, packaging…)
• Price (level, discounts…)
• Promotion (advertising, sales promotion…)
• Place (channels of distribution…)
The marketing mix is…….
Borden, N.H. (1965), “The concept of Marketing Mix”. In Schwartz G. (ed.), Science in marketing.
Measurement and analysis of consumer preference
Mea
sure
men
t and
An
alys
is of
Con
sum
er
pref
eren
ce
To ensure accurate interpretations are made, data must be collected correctly. Often, detail is overlooked, in an attempt to reduce cost.
Mea
sure
men
t and
An
alys
is of
Con
sum
er
pref
eren
ce So what can we expect?
Standard Product
Brand AVariant 1
Brand BBrand C
Brand DBrand EBrand F
Variant 3Variant 2
Variant 4Variant 5
>100 consumers rate their preference on a 1-8 scale.
From Greenhoff & MacFie 1994
In practice……….M
easu
rem
ent a
nd
Anal
ysis
of C
onsu
mer
pr
efer
ence
There are three types of measurement that we will consider today:-
• Paired Comparison Test • Nine-point Hedonic Scales• Cluster Analysis
Name………….……………Code………….…….…Date…………...
Instructions:
Evaluate Both Products starting from the left. Check the box for the product that you prefer. You must make a choice.
128 337
Paired comparison test
Mea
sure
men
t and
An
alys
is of
Con
sum
er
pref
eren
ce• 2 beers only• Which one do you prefer ? 128 337
Roberto Carlos
17 June 2005
352
X
A comparison of twoM
easu
rem
ent a
nd
Anal
ysis
of C
onsu
mer
pr
efer
ence
So, for example:-
My Brand vs Budweiser • 500 consumers assessed• 192 preferred Budweiser• 308 preferred my beer
Good news ! • Perform analysis on data to check reliability
So what’s it good for ?
Mea
sure
men
t and
An
alys
is of
Con
sum
er
pref
eren
ce
Advantages……. • Compare your brand against market leader• Low cost
Disadvantages…… • Does not tell you why, or by how muchIf there is no significant difference in this comparison test, then……
……..you can avoid wasting time
and money on further testing.
Hedonic scalesDesigned to assess more than 2 beers
Sample Code:-Dislike extremely Dislike moderately Neither like nor dislike Like moderately Like extremely
1 2 3 4 5 6 7 8 9
Sample Code:-Dislike extremely Dislike moderately Neither like nor dislike Like moderately Like extremely
1 2 3 4 5 6 7 8 9
Sample Code:-Dislike extremely Dislike moderately Neither like nor dislike Like moderately Like extremely
1 2 3 4 5 6 7 8 9
Sample Code:-Dislike extremely Dislike moderately Neither like nor dislike Like moderately Like extremely
1 2 3 4 5 6 7 8 9
224
649
176
789
Mea
sure
men
t and
An
alys
is of
Con
sum
er
pref
eren
ce
More than two beersM
easu
rem
ent a
nd
Anal
ysis
of C
onsu
mer
pr
efer
ence
A practical example-Test consumer preference for 12 beers:-
Our brandFive variants Six competitors Calculate the average preference for each beer.
The resultsM
easu
rem
ent a
nd
Anal
ysis
of C
onsu
mer
pr
efer
ence
>100 consumers rate their preference on a 1-8 scale.
From Greenhoff & MacFie 1994
Standard Product
Brand AVariant 1Variant 3Variant 2Brand BBrand C
Brand DBrand EBrand F
Variant 4Variant 5
5.995.805.695.645.635.555.555.325.285.165.06
4.75
Brand A appears most liked – but is this significant ?
The merits of scalingM
easu
rem
ent a
nd
Anal
ysis
of C
onsu
mer
pr
efer
ence
Advantages……Determine the order of preference > 2 beersDetermine the magnitude of liking or dislike
Disadvantages…….Scales are generally misused Order is critical – needs to be randomisedRelatively high cost
Cluster analysisM
easu
rem
ent a
nd
Anal
ysis
of C
onsu
mer
pr
efer
ence
Is used to classify groups of consumers OR products that are
homogeneous within themselves and
heterogeneous between each other.Organize data into meaningful information depending on their preference
How do you do it ?M
easu
rem
ent a
nd
Anal
ysis
of C
onsu
mer
pr
efer
ence
• Ask screening questions first• Measure the scores from a nine-point hedonic scale• Ask socio-economic questions after• Perform statistical analysisShows market segmentation according to preference, and therefore understand buyers’ behaviour.
Some clustersM
easu
rem
ent a
nd
Anal
ysis
of C
onsu
mer
pr
efer
ence
Cluster Freq. Brand A
Brand B
Brand C
Brand D
1 83 4.93 5.72 4.66 3.482 45 3.67 7.71 6.60 5.583 72 7.31 5.64 6.15 6.71
Total 200 5.51 6.14 5.64 5.12
Example – Mean score by cluster for 4 beer brands
*Adapted from Bogue, Sorenson, and Delahunty, 2002
Cluster 1 - preference for Brand B- least liked Brand D
Cluster 2 - preference for Brand B - least liked Brand A
Cluster 3 - preference for Brand A - least liked Brand B
…..and the consumers
Mea
sure
men
t and
An
alys
is of
Con
sum
er
pref
eren
ceExample – Mean score by cluster for 4 beer brands
*Adapted from Bogue, Sorenson, and Delahunty, 2002
Cluster 1 2 3Gender
Male 40 36 40Female 60 64 60
Age18-24 30 60 825-34 48 29 1535-44 8 7 2445-54 10 2 4955+ 4 2 4
Socio-economicA 0 2 1B 12 4 14C 69 83 72D 0 4 4E 19 7 9
Marital StatusSingle 73 89 69
Married 19 7 26Separated / divorced 3 0 3
Co-habiting 5 4 2
preference for Brand Bleast liked Brand D
preference for Brand Bleast liked Brand A
preference for Brand Aleast liked Brand B
Knowing your product – the importance of brand profiling
Know
ing
your
pro
duct
– th
e im
porta
nce
of b
rand
pr
ofilin
g
Can allow you to understand why the consumer likes a product. Many factors can have an effect on initial and continual purchase.
Extrinsic – e.g. price, packaging, glassware.
Intrinsic – e.g. colour, carbonation, flavour.
Today we will focus on flavour
0
1
2
3
4
5
6Sulphury
Sweet
Bitter
Spicy Hop
Astringent
Floral
Sour
Warming
Drying
Smooth
Carbonation
FruityPremiumStandardVolume
0
1
2
3
4
5
6Sulphury
Sweet
Bitter
Spicy Hop
Astringent
Floral
Sour
Warming
Drying
Smooth
Carbonation
FruityPremiumStandardVolume
Brand profilingKn
owin
g yo
ur p
rodu
ct –
the
impo
rtanc
e of
bra
nd
profi
ling
Spider PlotsFlavour often analysed using between eight and sixteen characters.
Spider plotsKn
owin
g yo
ur p
rodu
ct –
the
impo
rtanc
e of
bra
nd
profi
ling
Advantages……. • Simple requirements – panel training requirement low• Between 8 – 16 terms required• Little time required to generate• Easy to compare brands directly
Disadvantages……..• Beer has more than eight to sixteen flavours • ‘Constructs’ rather than ‘attributes’ used e.g. sulphury, fruity
Constructs vs Attributes
Know
ing
your
pro
duct
– th
e im
porta
nce
of b
rand
pr
ofilin
g
An example -
0
1
2
3
4
5
6Sulphury
Sweet
Bitter
Spicy Hop
Astringent
Floral
Sour
Warming
Drying
Smooth
Carbonation
0
1
2
3
4
5
6Sulphury
Sweet
Bitter
Spicy Hop
Astringent
Floral
Sour
Warming
Drying
Smooth
Carbonation
FruityFruity
EthylHexanoateIsoamyl acetate
Ethyl Butyrate
Acetaldehyde
Brand Flavour Fingerprinting
Know
ing
your
pro
duct
– th
e im
porta
nce
of b
rand
pr
ofilin
g
The Taste and Mouthfeel Spider Plot shows the relative contribution of taste and mouthfeel characters to the brand identity (Scale is 0-10).
Origins Plot: a summary plot that shows the origin of the brand’s main flavour characters (Contribution (%) to overall beer flavour). OFI: Overall flavour index.A low score represents a lightly flavoured beer, and a high score represents a beer with very intense flavour characters.
The Icon Plots break down the main areas of flavour contribution into their component flavour attributes (Maximum contribution =100%).
Brand
A Summary Description that uses consumer-friendly terms while providing a technically accurate description of the brand flavour character.
A list of the flavour attributes that are to be expected in this brand.
A list of the flavour attributes that shouldn’t be found in this brand.
How do you fingerprint ?
Know
ing
your
pro
duct
– th
e im
porta
nce
of b
rand
pr
ofilin
g
So what do we need?• Highly trained taste panel of 10 or more. • Three samples of each brand of different batches.• Correct sensory procedures.• Statistical analysis and interpretation.• A knowledge of beer production. • A knowledge of alternative descriptors.
FingerprintingKn
owin
g yo
ur p
rodu
ct –
the
impo
rtanc
e of
bra
nd
profi
ling
Benefits -• Scaling of attributes allows accurate profile• Constructs can be made from the attributes• Information available to many levelsDisadvantages –• High level of tasting competence required
Product MapKn
owin
g yo
ur p
rodu
ct –
the
impo
rtanc
e of
bra
nd
profi
ling
A1
A3A2
B2B1
B3
C2
C1
C3D1
D3
D2E3
E2E1
F2
F1
F3 G1
G2
G3H1
H2 H3I1 I2
I3Complex
Subtle
Tropical Fruits
Spicy
Malty
BiscuityGrainy
Fruity AppleBanana
Green Apples
Toffee
Hoppy
Bitter
Trial 1
Trial 2
Astringent
Preference MappingPr
efer
ence
Map
ping
Why perform preference mapping ?To understand the market:-
CONSUMERS PRODUCTSBy using Internal and External preference mapping.
Internal preference map
Pref
eren
ce M
appi
ng
Principal component 1 (45%)
Prin
cipal
com
pone
nt 2
(35%
)
Brand 2
Brand 4
Brand 3
Brand 1
A fizz of fruit – big bananas and dessert apples. Subtle fresh hop characters with whispers of grain. A fine bitterness and a clean, refreshing finish.
Subtle fruit, with hints of hard-boiled egg, give way to warm winey notes. Dry, with an iron-filings after-taste.
A deliciously aromatic beer, brimming with spice. An assertive bitterness and subtle hops blend with coconut water and grain.
Light tropical fruit notes banana, mango and pineapple - a clean, refreshing flavour. Light hop notes and a fast finish on the palate.
Adam, single,
30 yrs, 106 kg
regular consumer
A simple example
Adam
Bill Rich June Simon MEAN
Brand A 9 6 2 6 2 5Brand B 6 9 9 2 6 6.4Brand C 2 6 9 6 9 6.4Brand D 6 2 2 9 6 5Brand E 6 6 6 6 6 6
5 consumers’ likings for 5 products
Pref
eren
ce M
appi
ng
Preference spaceFind a direction in space that reflects Adam’s scores.
A B
CD
E
AdamA 9B 6C 2D 6E 6Pr
efer
ence
Map
ping
A B
CD
E
A simple internal preference mapNow, let’s look at the rest
Pref
eren
ce M
appi
ng
Adam 9 6 2 6
6Rich
2 9 9 2 6
Bill6 9 6 2
6
June6 2 6 9 6
Simon2 6 9 6 6
Principal Components Analysis
Generating a mapAdam
Bill Rich June Simon MEAN
Brand A 9 6 2 6 2 5Brand B 6 9 9 2 6 6.4Brand C 2 6 9 6 9 6.4Brand D 6 2 2 9 6 5Brand E 6 6 6 6 6 6
A B
CD
E
•Fred •John
•Kate
•Pat •Sue
• Mean
-1.5
-1
-0.5
0
0.5
1
1.5
-1.5 -1 -0.5 0 0.5 1 1.5
- - a xis F 1 (6 6 %) - ->
Adam Bill
June Simon
RichMean
A B
CD
E
•Fred •John
•Kate
•Pat •Sue
• Mean
-1.5
-1
-0.5
0
0.5
1
1.5
-1.5 -1 -0.5 0 0.5 1 1.5
- - a xis F 1 (6 6 %) - ->
Adam Bill
June Simon
RichMean
Variables and observations (axis F1 and F2:99%
Pref
eren
ce M
appi
ng
Overlay sensory dataInclude the sensory properties on the consumer preference map
Variables and observations (axis F1 and F2: 99 %)
A B
CD
E
•Fred •John
•Kate
•Pat •Sue
• Mean
-1.5
-1
-0.5
0
0.5
1
1.5
-1.5 -1 -0.5 0 0.5 1 1.5
- - a xis F 1 (6 6 %) - ->
Adam Bill
June Simon
RichMean
Variables and observations (axis F1 and F2: 99 %)
A B
CD
E
•Fred •John
•Kate
•Pat •Sue
• Mean
-1.5
-1
-0.5
0
0.5
1
1.5
-1.5 -1 -0.5 0 0.5 1 1.5
- - a xis F 1 (6 6 %) - ->
Adam Bill
June Simon
RichMean
Variables and observations (axis F1 and F2:99%
Variables and observations (axis F1 and F2: 99 %)
A B
CD
E
•Fred •John
•Kate
•Pat •Sue
• Mean
-1.5
-1
-0.5
0
0.5
1
1.5
-1.5 -1 -0.5 0 0.5 1 1.5
- - a xis F 1 (6 6 %) - ->
Adam Bill
June Simon
RichMean
Variables and observations (axis F1 and F2: 99 %)
A B
CD
E
•Fred •John
•Kate
•Pat •Sue
• Mean
-1.5
-1
-0.5
0
0.5
1
1.5
-1.5 -1 -0.5 0 0.5 1 1.5
- - a xis F 1 (6 6 %) - ->
Adam Bill
June Simon
RichMean
Variables and observations (axis F1 and F2:99%
FruitySpicy
MaltyHoppy
Pref
eren
ce M
appi
ng
To recapPr
efer
ence
Map
ping
Internal Preference Mapping allows consumers to be put in a ‘people space’, relating to their preference. The procedure does not show why they like it, but only how much. Level of discrimination for the consumer is higher from internal preference mapping, compared with external preference mapping.
External Preference Map
Pref
eren
ce M
appi
ng
Uses PCA applied to sensory data and results in a product map, according to sensory attributes
A1A1
A3A3A2A2
B2B1
B3
B2B1
B3
B1
B3
C2
C1
C3C2
C1
C3
D1
D3
D2D1
D3
D2E3
E2E1
E3
E2E1
F2
F1
F3
F2
F1
F3 G1
G2
G3
G1
G2
G3H1
H2H3
H1
H2H3
I1 I2
I3
I1 I2
I3Complex
Subtle
Tropical Fruits
Spicy
Malty
BiscuityGrainy
Fruity AppleBanana
Green Apples
Toffee
Hoppy
Bitter
Trial 1Trial 1
Trial 2Trial 2
Astringent
Then correlate the consumers Allows an ideal point to be generated.
12 beer examplePr
efer
ence
Map
ping
5.995.80
5.695.645.635.55
5.325.28
5.165.06
4.75
5.55
Brand AVariant 1Variant 3Variant 2Brand BBrand C
Standard ProductBrand DBrand EBrand F
Variant 4Variant 5
>100 consumers rate their preference on a 1-8 scale. From Greenhof& MacFie1994
5.995.80
5.695.645.635.55
5.325.28
5.165.06
4.75
5.55
Brand AVariant 1Variant 3Variant 2Brand BBrand C
Standard ProductBrand DBrand EBrand F
Variant 4Variant 5
>100 consumers rate their preference on a 1-8 scale. From Greenhof& MacFie1994
Remember the consumer testing we performed on our own brand, some variants, and our competitors?
We also had the same samples subjected to sensory evaluation.
PCA resultsPr
efer
ence
Map
ping
From Greenhoff & MacFie 1994
OverallLike
veg .aroma
cabbagelike
soapy
bitter
hoppy
alcoholic
fragrant
fermentingfruit
maltytoffee
sweet
body
citrus rancid
Br. E
V3Br. C
V1
Br. D
Br. B
StdV2 Br. A
V5
V4Br. F
drainy
We obtained the following results from Principle Components Analysis of the sensory data
……and the consumers ?
Pref
eren
ce M
appi
ng
From Greenhoff & MacFie 1994
OverallLike
veg .aroma
cabbagelike
soapy
bitter
hoppy
alcoholic
fragrant
fermentingfruit
maltytoffee
sweet
body
citrus rancid
Br. E
V3Br. C
V1
Br. D
Br. B
StdV2 Br. A
V5
V4Br. F
drainy
Internal vs External mapping
Pref
eren
ce M
appi
ng
Consumer data
PCA – people map
Correlate in sensory
dataPresent Plot
Internal mapping
Sensory data
PCA – product map
Correlate in consumers
Present Plot
External mapping
A suggested approach
Pref
eren
ce M
appi
ng
Perform cluster analysis- gives you some groupings - charts useful
Perform internal preference mapping- picture of consumers and brands- overlay clusters and sensory – do they fit?- compare preference map with sensory data
Perform external preference mapping- place consumers in
The importance of terminology
The
Impo
rtanc
e of
Te
rmin
olog
ySuccessful new products can be produced if we match consumer expectations.
“ to delivering that product unchanged to the
consumer...”
“to product...”
“to recipe...”
“ From consumer preferences to product
flavour specifications...”
Plant Production
Innovation
Expert Panel
Consumer Testing
Distribution & Product Support
Principal component 1 (45%)
Princ
ipal c
ompo
nent
2 (35
%)
Brand 2
Brand 4
Brand 3
Brand 1
Principal component 1 (45%)
Princ
ipal c
ompo
nent
2 (35
%)
Principal component 1 (45%)
Princ
ipal c
ompo
nent
2 (35
%)
Brand 2
Brand 4
Brand 3
Brand 1
Brand 2
Brand 4
Brand 3
Brand 1
New product development
Links
to B
rewi
ng
Prac
tice
?
From an internal preference map you may see a gap where many consumers are not catered for.
You may want to develop a beer for these consumers.
The right flavourLin
ks to
Bre
wing
Pr
actic
e
From an external preference map, you can see that fragrant, fruity beers cater for the most consumers.
OverallLike
veg.aroma
cabbagelike
soapy
bitter
hoppy
alcoholic
fragrant
fermentingfruit
maltytoffee
sweet
body
citrus rancid
Br. E
V3Br. C
V1
Br. D
Br. B
StdV2 Br. A
V5
V4Br. F
drainy
OverallLike
veg.aroma
cabbagelike
soapy
bitter
hoppy
alcoholic
fragrant
fermentingfruit
maltytoffee
sweet
body
citrus rancid
Br. E
V3Br. C
V1
Br. D
Br. B
StdV2 Br. A
V5
V4Br. F
drainy
Is this the beer that you will sell most of ?
Check your cluster analysisWhat about Drinkability ?
Common Pitfalls Co
mm
on P
itfal
ls
…… in preference measurement and analysis Requires the selection of correct consumer and productDetailed conduct of exercise – e.g. order of presentation
service temperaturedark or clear glassconsistent sample size
Many considerations should be madee.g. the environment :- in-Hall or Bar ?
Common Pitfalls Co
mm
on P
itfal
ls
…… in sensory evaluation• Insufficient training of ‘expert’ sensory panel e.g. narrow range attributes inability to scale correctly• Insufficient tasters• Incorrect use of terminology• Bias from incorrect procedure• Lack of replication• Sensory fatigue
Some ExamplesSo
me
Exam
ples
Example 1 – Fresh beer and drinkabilityAgeing flavours could decrease drinkability by up to 15%. The consumer preferred beer approximately 4 months old. The beer that the consumer wants will reduce the volume of sales. Market research was used to determine how large the dislike of fresh beer was. The consumer dislike was judged small enough to be overcome by ‘educating the consumer’.
Some ExamplesSo
me
Exam
ples
Example 2 – A change in brewing practice. A brewing company moved from a ‘traditional’ brewing practice to a new site. The beers brewed in the new brewery showed differences in profile, as determined by BFF. Paired comparison tests were performed, and it was shown that the difference in flavour profile was not reflected in a significant difference in consumer preference.
Some ExamplesSo
me
Exam
ples
Example 3 – New Product DevelopmentOn-line research showed that consumers enjoyed the concept of a low carbohydrate beer. Assess the preferred brand in the market. The preferred beer was profiled, and minor alterations made – the ideal product.Pilot batches of the new lo-carb beer with the preferred flavour profile was brewed. Further consumer testing was performed in the lo-carb market which showed that the new product was preferred.
SummarySu
mm
ary
To get the optimal benefit from market research - Correct measurement of consumer preference is imperative.Understand your brand in detail.Preference mapping lets you understand the consumer.To deliver the correct marketing mix.
Questions please ?Qu
estio
ns
Thank you all for listening Please feel free to ask questions now, or later.
adam.fenton@flavoractiv.comwww.flavoractiv.com