Think Global, Act Local

1
Think Global, Act Local

• How Market Research Can Help Defining Local
  Markets, yet Keeping the Global Picture

Presenters Jeroen Slappendel PBIRG 2007
2
Todays workshop

• System differences and Cultural differences
  between countries
• Screening
• Questionnaire phase
• Analysis
• Bias in results
• How to prevent biased results
• Short introduction of methodologies Conjoint
  Analysis and MaxDiff
• Using Conjoint Analysis and MaxDiff to minimize
  cultural bias
• Implementation of results

3
Market research around the world

• Goal to segment the contraception market based
  on attitudes and behavior in contraception
• Goal Identify potential and positioning for new
  Stroke treatment
• Goal Evaluate brand promise around the globe

Differences?
Similarities?
System limitations?
Cultural bias?
4
How biased are you?
5
Brazilian woman
Interview takes 25 longer than in other countries
Everything is important no differentiation
between attributes
6
Korean woman
Short answers
Shy when talking about feelings
Avoids direct answers
7
American woman
Comfortable talking about feelings
Less comfortable talking about intimate matters
Reports more side effects and co-morbidities
8
Different methodologies, different cultural
effects

• Focus groups
• Not very effective in Asian countries when
  dealing with personal matters or emotions
• Almost impossible to do in Japan, as both
  patients and physicians do not like speaking in
  public, let alone disagreeing in public
• In Brazil, potential for heated and long
  discussions moderator very important!
• In-depth interviews
• Can take up to 40 more time in Brazil,
  especially with patients
• German physicians give concise answers much
  probing needed
• Very short answers in South Korea and China
• French physicians held in high regard
  interviewers may find it difficult to probe
• Quantitative interviews
• Scaling bias Brazilians tend to rate all
  attributes as important

9
Screening

• In study, women were screened on the street and
  on-line
• In Italy, we did a split-run analysis for women
  screened on the street and on-line
• Large difference in contraception use between
  women screened on the street and screened on-line
• Italian women who did not use contraception, did
  not want to talk about it when screened on the
  street
• Risk of bias in results because of
  (self-)selective sample

10
The sample based on current usage is not
homogeneous
N14382
N2392
N2377
N2397
N2278
N2392
N2546
11
Questionnaire design

• Account for differences in length of
  questionnaires in different countries
• Overall, 30-minute questionnaire
• Average of 22 minutes in South Korea, hardly any
  drop-outs
• 50 of women in Brazil needed over 45 minutes to
  finish the questionnaire
• 30 of Brazilian women who started the
  questionnaire did not it
• Be careful with sensitive subject and when
  necessary adapt questionnaires
• In countries like France, it is deemed
  inappropriate to ask about religion
• In the US and South Korea, more women will quit
  the questionnaire when they are asked about
  intimate subjects. If possible, offer a way out
  in the questionnaire in these countries

12
Think global or act local?

• Comparable results across the world?
• Standardize scaling questions
• Choose research methodologies that are bias free
• Discrete Choice Modeling / Choice Based Conjoint
• MaxDiff Scaling
• Emphasizing cultural differences?
• Focus groups
• In-depth qualitative interviews
• Comparable results while understanding cultural
  differences?
• Hybrid research

13
How likely are you to ask your doctor for this
new pill? (answered by all women per country)
14

Lifestyle factors per region
(Socially) actives
Extroverts
Order religion
'Worriers'
Variety seekers
Naturalists
Achievers
N14519
N2395
N7182
Low score
High score
N2393
Note these figures havent been corrected yet
for individual answer patterns
N2549
15
Key facts about greasy skin and/or greasy hair
(region - weighted)
N14519
N2395
N7182
N2393
N2549
16
Key facts about Painful menstrual bleeding
(region - weighted)
N14519
N2395
N7182
N2393
N2549
The trend becomes very clear!
17
How to deal with scaling bias across cultures

• Standardization of scales per country
• Calculate average score per respondent
• Subtract average score of each score per question

Score How important is reducing acne symptoms to
you?
Score Korean woman 4 Average score 1.6 Standard
ized score 2.4
Score Brazilian woman 4 Average
score 3.9 Standardized score 0.1

• At first sight, the score of the Korean woman and
  Brazilian woman are equal
• After standardization, differences in importance
  become evident

18
Two methods that exclude cultural scaling bias

• Conjoint Analysis
• MaxDiff scaling

19
Conjoint Analysis

• To measure the sensitivity of the physicians to
  the various product characteristics we used
  conjoint analysis. In conjoint analysis the
  respondent is confronted with different
  combinations of features and has to specify his
  preference or choice. By choosing he/ she reveals
  his/ her sensitivity to each product feature.
• Conjoint analysis helps us understand how
  important each product characteristic (attribute)
  is in the choice of a product and how sensitive
  the respondent is to the various product
  characteristics (attributes).
• Using conjoint we are able to simulate market
  preferences because we know
• the preference for attributes and their levels
• the choices individuals make between the
  attributes and levels

20
Whats so good about conjoint?
When prescribing drugs for Alzheimer's
disease, how important is to you ?
Not Important
Very Important
Brand Improved recognition of family Depression
improves Dosage (OID/BID) Drug is fully reimbursed
1 2 3 4 5 6 7 1 2 3 4 5 6
7 1 2 3 4 5 6 7 1 2 3 4 5 6
7 1 2 3 4 5 6 7
(Circle one per item)
21
Whats so good about conjoint?
More realistic questions
Would you prefer . . .
1 increase in bone mass density (BMD) Costs
5 per month
4 increase in bone mass density Costs 30 per
month
OR
1
2

• If choose left, you are driven by price more than
  efficacy. If choose right, you concentrate more
  on efficacy
• Rather than ask directly if you value Efficacy or
  Price we present specific tradeoff scenarios and
  infer preferences from your choices

22
Introduction to conjoint
23
Direct questioning shows regional differences in
preference scales
Mean score
24
When indirect methods are used to establish
drivers of choice, a more clear picture is
obtained
Choice driven by
25
Introduction to Maximum Difference Scaling

• The Maximum Difference Scalling also known as
  Best/Worst Scaling was introduced by Jordan
  Louviere in the early 1990s
• This method can be seen as an extension of paired
  comparisons techniques
• The method applies nicely to general scaling
  problems for multiple items

26
MaxDiff Task example
27
MaxDiff vs Rating and Ranking

• Trade off vs direct
• No scale use bias vs scale use bias
• Seeking differences is in the methodology vs no
  trade off/forced ranking
• Engaging, simple 2 click tasks vs rational tasks
  (ref. subset ranking)
• Individual ratio scaled weights vs essentially
  ordinal (distribution)
• Easy to design, advanced yet automated analysis,
  easy to interpret
• Very similar applications

28
MaxDiff vs Conjoint

• Holistic vs additive
• (product changes price sensitivity)
• Perception vs choice
• Descriptive vs predictive
• Seeking differences is in the methodology vs
  choice
• Engaging, simple 2 click tasks vs complex
  comparisons
• Directly available individual ratio scaled
  weights vs analysis after simulation

29
When to use MaxDiff?

• Because of the properties of ease of use,
  scale-free, emphasis on strong preferences,
• SKIM considers MaxDiff superior to both rating
  and ranking.
• If your research is more exploratory or
• your list of items does not constitute a proper
  product model
• then MaxDiff is also to be preferred over
  conjoint.

Understanding preference for Product
claims Advertising messages Drug
characteristics Brands Etc.
30
Dont stop just there

• Using Latent Class is very useful for getting
  insight in heterogeneity by finding different
  preference structures
• Find those cultural differences!
• You can combine different MaxDiff in one overall
  model (buckets)
• See the case study (next)
• You can use MaxDiff for testing only a subset of
  all possible conjoint concepts (as an alternative
  to conjoint)
• A continuous scale -- from classic MaxDiff via
  MaxDiff with semi-holistic concepts to Conjoint
  with fixed SKUs to classic Conjoint
• Add perception questions to conjoint
• Which type of packaging would you choose?
• Which brand is most appealing?

31
MaxDiff in positioning research

• Objective design optimal positioning concept for
  Product X
• 600 on-line interviews with 2 specialties in
  Canada, UK, Spain, Germany and France
• Using of MaxDiff to find most appealing messages
  and claims from a total list of 120 messages

32
Preference for claims behavior
Helps avoid behavioral problems. (100)
Most Preferred
Most liked in Canada
Helps reduce aggressive behavior. (83)
Keeps the patient calm and relaxed. (78)
Helps avoid irritability and agitation. (46)
Least liked in Spain
Helps improve or control anxiety. (36)
Least Preferred
Keeps the patient more alert. (0)
Scores have been rescaled to a 0-100 scale
33
Putting it all together Evaluating buckets

• In addition to understanding which statements are
  most appealing within each bucket we can
  understand which overall buckets are most
  appealing
• Accomplished by having consumers complete the
  exercise again with any combination of
  statements in each task
• Shows rating of each statement in comparison to
  all other statements

Behavior
Cognition
Function
General Efficacy
Care Burden
General Patient Care
34
Focus of positioning concepts
Example results Behavior was the most
preferred category in all countries but Germany
Cognition was most preferred there
Example results
Function
General Efficacy
Cognition
Care Burden
Behavior
General Patient Care
Most preferred
Least preferred
MaxDiff Results
35
Global picture
Local Differences
36
Understanding local differences while keeping a
global picture

• Hybrid Research
• Pulls together typically disparate phases into
  one
• Grounds the analysis through the use of
  perspective and understanding
• Provides a balanced reading of the business /
  strategic problem

Comparable results across countries
Understanding local differences
37
Uncover differences in healthcare systems

• Research to stroke and ICH (Intracerebral
  Hemorrhage) in Brazil, China, India, Russia,
  Saudi Arabia, South Africa, South Korea and
  Turkey
• 10 qualitative interviews with healthcare
  professionals in each country and a total of 700
  quantitative interviews

Maximum of 4 hour window of opportunity to treat
after attack
38
The delay incurred prior to hospital admittance
may represent important barriers in most
countries, yet, in Saudi Arabia, Turkey and China
higher proportion of patients are admitted within
3 hours
Pre-hospital summary
39
In most countries, the average time from
admittance to diagnosis is between 1 and 1.4
hours both in daytime and night-time, yet the
waiting time is much longer in Brazil, South
Africa and Russia
Initial assessment summary
Specialists in parenthesis are almost as
important as the main decision maker
40
In most countries the majority of hospitals have
CT Scans except India and Turkey where patients
are referred to stand-alone scanning centres if
CT Scan is not available
Diagnosis summary
Specialists in parenthesis are almost as
important as the main decision maker
41
Implementation of results
How to think global, but act local?

• Which market has the highest potential?
• Brazilian women are very receptive to new
  contraceptive product
• Korean women show least interest to ask physician
  for a new type of drug
• How to communicate?
• In Europe, do not exaggerate or use marketing
  language. With physicians statements like The
  first and only drug have a negative effect.
• US physicians are more used to marketing language
  and expect strong marketing language, otherwise
  they suspect may be wrong with the drug
• Local regulations
• DTC possible in US and Brazil, while in Europe
  and South Korea only indirect advertising can be
  used

42
Pulling it all together

• Choose the appropriate methodology do you want
  to have a global picture or emphasize local
  differences?
• For comparable results across the world, try
  using methodologies and techniques that ensure
  bias free scales like discrete choice modeling
  and MaxDiff scaling
• Consider a methodology that allows to get
  comparable results, while gaining local
  differences and cultural understanding
• Never forget to take the international context
  into account!

43
For more information, please contactJeroen
Slappendel or Nelson Silva