Advanced Conjoint Analysis Methods: Intelligent Design Equals Impactful Results

1
Advanced Conjoint Analysis MethodsIntelligent
Design Equals Impactful Results

• PBIRG WorkshopJanuary 28, 2008

2
Structure of the presentation

• Introduction to SKIM Analytical Healthcare
• Types of marketing challenges through the product
  lifecycle
• Introduction to Conjoint Analysis
• Designing a Conjoint Study for different
  marketing challenges Intelligent Design Equals
  Impactful Results
• Summary of Conjoint Designs

3
A combination of expertise falls under the
umbrella of the SKIM Group
SKIM Group (EU and US based)
SKIM Software Services
SKIM Analytical
SKIM Research services
Consumer
SKIM Analytical Healthcare
4
SKIM gathers a broad spectrum of expertise,
knowledge and cultures
Biochemistry
Econometrics
Business Administration
Pharmaceutical sciences
Industrial Engineering Marketing
Drug Development
Europe (Rotterdam Geneva)
US (New York)
Team
International business
Pharmaceutical sciences
Economics
Economics
Supportive Consultants (software, programming,
qualitative unit, fieldwork collaborators)
Marketing
Business Administration
Psychology
And a variety of cultural backgrounds
5
Prior experience and knowledge of several
indication areas ? better understanding of your
business issues
Cardiology Atrial fibrillation Hypertension Strok
e ICH VTE Dermatology Atopic
dermatitis Psoriasis Medical Devices Blood
glucose meters Coagulation devices Insulin pens /
pumps Medical systems Device testing GI
tract Crohns disease GERD IBS Peptic
ulcer Haemostatic system Coagulation
testing Coagulation Dialysis
Hormones and Urology Male hormone
therapy Erectile dysfunction Uterine
fibroids (Low dose) HRT Contraception Metabolics
Diabetes/(Inhaled) insulin Metabolic
syndrome GLP 1 Islet cell Tx Musculoskeletal /
Immune systemOsteoarthritis Osteoporosis Multipl
e Sclerosis Pagets disease Pain
(Acute/chronic) Rheumatoid arthritis
Transplantation and Ix
Neurology Alzheimers Disease Epilepsy
Migraine DAT / Parkinsons disease Oncology Non
-Myeloid cancers Leukemia- NSCLC Colorectal
cancer Pancreatic cancer Bone metastases Prostate
cancer Non-Hodgkins Psychiatry Bipolar
disorder Depression Schizophrenia Respiratory CO
PD Asthma
6
Structure of the presentation

• Introduction to SKIM Analytical Healthcare
• Types of marketing challenges through the product
  lifecycle
• Introduction to Conjoint Analysis
• Designing a Conjoint Study for different
  marketing challenges Intelligent Design Equals
  Impactful Results
• Summary of Conjoint Designs

7
Research throughout the product lifecycle
PRE LAUNCH
POST LAUNCH
Introduction
Growth
Phase 3
Phase 2
Phase 1
Maturity
Forecasting
Positioning
Product optimization
Product opportunity identification
Segmentation
Sales
-6
-3
0
2
4
Time
8
Conjoint as tool throughout the product lifecycle

• Conjoint analysis can be used to get insightful
  and impactful answers to the different marketing
  questions, e.g.
• What features are valued by patients /
  physicians?
• What can you do if a competitor launches a new
  product?
• How to position / re-position your products?
• How can the market be segmented?
• What will be the potential of your new product?
  And what happens with the potential if the
  efficacy increases?

9
Structure of the presentation

• Introduction to SKIM Analytical Healthcare
• Types of marketing challenges through the product
  lifecycle
• Introduction to Conjoint Analysis
• Designing a Conjoint Study for different
  marketing challenges Intelligent Design Equals
  Impactful Results
• Summary of Conjoint Designs

10
What is conjoint analysis?

• Technique initially developed by psychologists in
  early 70s.
• Academics were interested in understanding how
  people made decisions/choices.
• By just asking, people tended to say what
• they thought the interviewer wanted to hear
  (politically / socially correct answers)
• was top-of-mind
• So the answers didnt reflect what they actually
  would do / choose / buy.
• It was noticed however that choices involved
  trade-offs and compromises.

This can be solved by conjoint analysis! An
example.
11
Conjoint choice example
All else equal which of these 2 beer bottles
would you buy? Click on beer of choice (blue
square)


2
2
12
Conjoint choice example (contd)
All else equal which of these 2 beer bottles
would you buy? Click on beer of choice (blue
square)


1
2
13
Conjoint choice example (contd)
All else equal which of these 2 beer bottles
would you buy? Click on beer of choice (blue
square)


2
1
14
Conjoint choice example (contd)
Index of added value
Carlsberg Heineken 2
1

• The choices indicate that
• S/he prefers Heineken.
• Offering a price reduction of 1 is NOT enough to
  change his / her mind.
• The added value of Heineken is LARGER than the
  added value of a price reduction of 1.

Next slide
15
Conjoint choice example (contd)
Index of added value
Carlsberg Heineken 2
1

• The choices indicate that
• S/he prefers Heineken.
• Offering a price reduction of 1 is ENOUGH to
  change his / her mind.
• The added value of Heineken is SMALLER than the
  added value of a price reduction of 1.

Next slide
16
Conjoint choice example (contd)
Index of added value
Heineken Carlsberg
2 1

• The choices indicate that
• S/he prefers Carlsberg.
• Offering a price reduction of 1 is NOT enough to
  change his / her mind.
• The added value of Carlsberg is LARGER than the
  added value of a price reduction of 1.

Next slide
17
Conjoint choice example (contd)
Index of added value
Heineken Carlsberg 2
1

• The choices indicate that
• S/he prefers Carlsberg.
• Offering a price reduction of 1 is ENOUGH to
  change his / her mind.
• The added value of Carlsberg is SMALLER than the
  added value of a price reduction of 1.

Next slide
18
Basic idea of Conjoint Analysis

• Basic idea of Conjoint Analysis

• Mimic the actual choice process
• A number of products, which differ from each
  other with respect to a number of features (like
  price, brand, size, efficacy etc.) are shown to a
  respondent.

• 2. Respondents will show their actual choice
  behavior
• The respondent then chooses the product which
  s/he finds most attractive.
• This choice process is repeated several times

• 3. Determine the importance of, and preferences
  for different product features by analyzing the
  respondents choice behavior
• After a series of choices, the decision rules
  or drivers of choice of the respondent can be
  determined

19
In conjoint analysis products are seen as a
combination of features, which can be adapted

• The key to using Conjoint Analysis is to think
  about products as a collection of different
  features
• E.g. Mobile Phone
• Brand Stand-by time Games Price Design
• Or

Price
Package type
Administration
Efficacy
Brand
Safety
20
Example 1 Give physicians trade-offs / choices
between different products, defined on features
21
Sensitivity analysiswhat level of efficacy must
be achieved?
of physicians who would Rx to patient i
22
Share of choice for different product scenarios
23
Simulator Example
Simulator Example
24
It is the option to simulate that makes it
appealing

• The Simulator helps you determine
• The value of product features
• The value of modifications to an existing product
• The potential of product line extensions do we
  cannibalize our own share or take mostly from our
  competitors?
• Demand curves (willingness to Rx/price)
• Important input into demand forecasting models
• Benefit segments driven by the same features
  (latent class analysis)
• And has validity beyond the end of the study

25
Structure of the presentation

• Introduction to SKIM Analytical Healthcare
• Types of marketing challenges through the product
  lifecycle
• Introduction to Conjoint Analysis
• Designing a Conjoint Study for different
  marketing challenges Intelligent Design Equals
  Impactful Results
• Summary of Conjoint Designs

26
Adapt conjoint design to marketing objective

• The conjoint design depends on your marketing
  objective and stage in lifecycle of the product

PRE LAUNCH
POST LAUNCH
Phase 2
Introduction
Growth
Phase 3
Phase 1
Maturity
Forecasting
Positioning
Product optimization
Product opportunity identification
Segmentation
Sales
-6
-3
0
2
4
Time
27
Case study product opportunity identification
PRE LAUNCH
POST LAUNCH
Phase 2
Introduction
Growth
Phase 3
Phase 1
Maturity
Product opportunity identification
Sales
-6
-3
0
2
4
Time
28
Product opportunity identification

• Situation when looking for a product opportunity
• In early stage of development
• Many uncertainties
• Not much knowledge about the market or customers
• Appropriate techniques
• Preference models (Adaptive Conjoint Analysis)
• Case study Stadium Hospitality Package
  Composition
• Maximum Difference Scaling

29
Stadium Hospitality Package Composition (ACA)

• Research objective determine the optimal
  composition of box and seat hospitality offers
  for a stadium
• Since drivers of decisions are not known, an
  Adaptive Conjoint Design was utilized to
  investigate sensitivities to a larger set of
  attributes

30
ACA design 19 different attributes

• Seat Location
• Restaurant Access
• Catering Fees
• Transport Links
• Contract Term
• Concerts Cultural Events
• Seat Specification
• In-Seat Services
• Conference Facilities
• Climate Control
• Multi-purpose venue
• Etc.
• What would you guess to be the most important
  drivers of choice?

31
Step 1 of ACA Rating

• 90 levels

32
Step 2 of ACA Importance

• 19 Importance ratings

33
Step 3 of ACA Pair-wise comparisons

• 30 pair-wise comparisons

34
Analysis by inspecting average utilities

• This average importance of this attribute is
  slightly over average/expected importance
  (100/19)
• What does that mean?
• The average utility for Heating and Air
  Conditioning is 39.5
• What does that mean?
• Number of respondents used 200

35
Stadium Hospitality Package Composition (MaxDiff)

• Maximum Difference Scaling (MaxDiff) can also be
  used to investigate the sensitivity for many
  items
• This can be done within attributes (e.g. climate
  control) and across attributes (e.g. climate
  control versus transport links versus seat
  location)

36
The Discriminating Power of this tool originates
from the use of choice data (best and worst from
a set)
The respondent gets a choice set of 3 or 4 items
chosen from the complete set of items
From each set, the respondent chooses the best
and worst item in accordance with her opinion.
This choice task is repeated a couple of times
with different choice sets chosen from the
complete set of items, in accordance with an
experimental design
37
MaxDiff delivers actionable information

• We get two types of information
• Ranking of items, to show what is liked most (and
  less)
• Distances between items in the rank, to show how
  far apart items are (strength of preference)
• We consider this approach superior to rating and
  ranking, because of
• Its discriminating power
• Across items and respondents because of the
  forced choice situation
• Its independence of culture
• Respondents can always indicate their most and
  least liked items, even if they know how to use a
  rating scale
• Free from scale use bias

38
Results for seat location
France, UK, Spain
Half-way line (100)
Most Preferred
Wings (90)
Germany, Canada
Country name indicates most preferred level for
that country
Near standing buffet / lounge (65)
Near large restaurant (59)
Near the car parking (49)
Near rail links (23)
Least Preferred
Corners (2.1)
Behind the goal (0)
Scores have been rescaled to a 0-100 scale
39
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

40
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

41
Case study Product optimization
PRE LAUNCH
POST LAUNCH
Phase 2
Introduction
Growth
Phase 3
Phase 1
Maturity
Product optimization
Sales
-6
-3
0
2
4
Time
42
Product optimization

• Market description
• Hyperphosphatemia elevated serum phosphate
  levels
• 80 of all stage V renal disease patients have
  hyperphosphatemia
• All products in market (phosphate binders) are
  efficacious, but compliance is an issue
• A new product Y will be introduced
• How can compliance be increased?

43
Product optimization Compliance model

• Which product would you choose to get the HIGHEST
  RATE OF COMPLIANCE?

Physicians choice task

• Compliance potential simulator

44
Case study Forecasting (2 examples)
PRE LAUNCH
POST LAUNCH
Phase 2
Introduction
Growth
Phase 3
Phase 1
Maturity
Forecasting
Sales
-6
-3
0
2
4
Time
45
Forecasting (example 1)

• Three new oral anticoagulants currently in phase
  III may represent a threat for current injectable
  products
• The primary objective of this study is to gain an
  insight into the possible threat by the launch of
  new orally available therapies within the
  anticoagulants market and the emotional drivers
  that will be guiding the choices for
  anticoagulants within the future competitive
  landscape.
• The company needs to make the following business
  decisions
• What are the consequences of the introduction of
  oral anticoagulants on product share and
  perception?
• What current unmet needs do the new oral
  anticoagulant answer? How does our current
  product stand in comparison to this?

46
Conjoint to understand drivers of choice

• Conjoint to understand drivers of choice
• Feature based
• Discrete choice approach
• Patient (profile) specific
• Provides in detail insight in the drivers of
  choice and increases the in depth understanding
  (particularly in hybrid research setting quali
  quant)
• Enables Benefit segmentation

47
Explanation slide
Highest level
Lowest level
non-inferior
The importances are calculated based on the
distance between the best and worst levels of
each attribute for each individual
inferior
superior
oral QD
SC BID
IV
SC QD
oral BID
The distance between two levels reflects the
added value
higher
lower
same
elevation
no elevation
transcient
no antidote
antidote
The blocks are the level of Product X
risk
no risk
biological
synthetic
lt6h / 6-12h / gt12h
NB fictitious results
48
Different product benefits are recognized
throughout the product lifecycle

• The conjoint design depends on the way physicians
  look at the different products in the market

PRE LAUNCH
POST LAUNCH
Phase 2
Introduction
Growth
Phase 3
Phase 1
Maturity
Long term values
Emotional benefits
Functional benefits
Rational benefits
Product intrinsic
Sales
-6
-3
0
2
4
Time
49
Example of conjoint for forecasting of new HC
product

• Conjoint for forecasting of new HC product
• A combined Holistic and feature based conjoint
  (see example)
• A constant sum approach

50
Example of conjoint for forecasting of new HC
product
Holistic brands / product categories
New product is feature based

• Good for forecasting of new HC product in a
  mature market!
• Combining the feature base new products with
  holistic existing products offers a more
  realistic choice setting

51
Simulator for forecasting of new HC product
Product A Product B Product C
52
Forecast for new product
Product A Product B Product C
Product A Product B Product C
53
Conjoint Market Simulation Assumptions

• All attributes that affect prescriber/patient
  choices in the real world have been accounted
  for
• Equal availability (distribution)
• Respondents are aware of all products
• Long-range equilibrium (equal time on market)
• Equal effectiveness of sales force
• No out-of-stock conditions
• All on formulary
• Reimbursement

54
The triple A paradigm
55
Forecasting (example 2)

• In some markets, the patient base of physicians
  is too diverse to ask the physician for a single
  choice of a product (heterogeneous patients)
• Often, pre-defined patient types are used for
  which physicians have to make a choice
• However, in some indication areas patients are
  too heterogeneous to define them in 3 or 4
  patient types
• Constant sum conjoint (dividing 10 prescriptions
  over the products) is one solution yet, this
  leads to rationalized answers
• Showing individual patient types during the
  conjoint exercise is the best solution
• First, a physician describes 4-5 different
  patient types (e.g. last 4-5 patients seen)
• These patient types are randomly shown during the
  conjoint exercise
• Physician needs to choose one product for each
  patient type (Considering this patient, which
  product would you choose?)

56
(No Transcript)
57
Refined analysis of patient cases and conjoint

• As the patient cases that are shown are
  representative to the patient population, the
  choices made and the simulated shares of the
  products will reflect the real product shares
• Analysis can be more refined by segmenting the
  different patient types that are shown, and
  looking into the physicians sensitivities for
  the different product features for different
  patient types

58
Case study Segmentation
PRE LAUNCH
POST LAUNCH
Phase 2
Introduction
Growth
Phase 3
Phase 1
Maturity
Segmentation
Sales
-6
-3
0
2
4
Time
59
Segmentation

• Latent class to segment the market and find
  segments that offer the greatest potential for
  the new product
• How many physicians are driven by characteristics
  that my product has?
• Which physicians to target?

60
Latent Class

• Latent Class is an analysis technique to group
  respondents based on the similarity in their
  choice pattern. In other words, respondents who
  are driven by the same features are grouped in
  the same segment.
• In the following example, the market is
  heterogeneous with regard to the sensitivities of
  insulin pump users. A part of the market will be
  sensitive to the new pumps another part might
  never consider the new pumps. In order to target
  the pump users that might be interested in a new
  pump, the market has been segmented.
• Preceding the analysis, the number of segments
  targeted are defined. In this survey, we run 6
  analyses in order to identify 2, 3, 4, 7
  segments. In a latent class analysis, the
  probability of each individual to belong to each
  of the groups (segments) is estimated. In an
  iterative process, the sensitivities for each
  group is re-estimated and the probability of each
  individual to belong to each of the groups is
  adapted. In the next phase, the various segment
  solutions are compared.

What?
Why?
How?
61
Latent Class analysis 3 groups of pump users
based on their revealed sensitivities
3 ml (pump capacity) Insulin1 Prefilled (3ml)
PumpA lt 50 g Very small Insulin2
35
21
44
Prefill 3ml PumpB lt 50 g Very small Insulin1
3 ml Insulin1 lt 50 g 3ml Prefill
PumpA Insulin2 lt 50 g
Prefill Remote C 3 ml Insulin1
33
22
25
20
Prefill PumpB lt 50 g Very small
PumpA Insulin2 lt 50 g
3 ml Insulin1
62
Segment 1 Insulin1 users with large cartridge
needs (3 ml)
3 ml (pump capacity) Insulin1 Bolus
Recommendation Prefilled (3ml)
PumpA lt 1.8 Oz Very small Insulin2

• Who are in this segment?
• Use a large capacity (3.0/3.15ml)
• Mostly Insulin1 users

43
22
44
3 ml Insulin1 lt 50 g 3ml Prefill
Prefill 3ml PumpB lt 1.8 Oz Very small Insulin1
PumpA BR Insulin2 lt 1.8 Oz
This segment is one that sees most benefit in 3ml
cartridges containing Insulin1
25
20
3 ml Insulin1
Prefill Remote C 3 ml Insulin1
38
22
Prefill PumpB lt 1.8 Oz Very small
PumpA BR Insulin2 lt 1.8 Oz
!! This is the battle ground for PumpB. These are
users that are not particularly linked to a
certain pump brand. The potential is large.
63
Segment 2 PumpB loyalists that want very small
and light pump

• Who are in this segment?
• Use mostly large capacity (3.0/3.15ml)
• Mostly Insulin1 users
• Mix of PumpA/PumpB users
• More French

3 ml (pump capacity) Insulin1 Bolus
Recommendation Prefilled (3ml)
PumpA lt 1.8 Oz Very small Insulin2
35
22
35
Prefill 3ml PumpB lt 50 g Very small Insulin1
3 ml BR Insulin1 lt 1.8 Oz 3ml Prefill
PumpA BR Insulin2 lt 1.8 Oz
This segment is one that sees most benefit in an
PumpB pump that is very small and light
Prefill PumpB lt 50 g Very small
3 ml BR Insulin1
BR Prefill Remote C 3 ml Insulin1
19
21
22
33
PumpA BR Insulin2 lt 1.8 Oz
!! This could be a potential for the new
generation pumps.
64
Segment 3 PumpA and Insulin2 loyalists

• Who are in this segment?
• Half 1.8ml and half 3.0ml users
• Half are Insulin2 users (50) and 40 Insulin1
• PumpA users
• Mostly Germans

3 ml (pump capacity) Insulin1 Bolus
Recommendation Prefilled (3ml)
PumpA lt 1.8 Oz Very small Insulin2
43
35
PumpA Insulin2 lt 50 g
Prefill 3ml PumpB lt 1.8 Oz Very small Insulin1
3 ml BR Insulin1 lt 1.8 Oz 3ml Prefill
21
This segment is one that is strongly PumpA and
Insulin2 driven
PumpA Insulin2 lt 50 g
38
3 ml BR Insulin1
BR Prefill Remote C 3 ml Insulin1
19
21
Prefill PumpB lt 1.8 Oz Very small
22
!! There is no real potential here for company
because these are PumpA driven users.
65
Case study Positioning
PRE LAUNCH
POST LAUNCH
Phase 2
Introduction
Growth
Phase 3
Phase 1
Maturity
Positioning
Sales
-6
-3
0
2
4
Time
66
Positioning

• Objective to segment the contraceptive market
  based on driver of choice (method, effect on
  acne, effect on bleedings, etc) in order to
  determine the positioning of a new product
• What do women value?
• Who will be interested in the new product?
• How to target the right groups?
• Conjoint analysis is used to understand the
  contraceptive choice behavior of women and to
  segment women in the market

67
Adapted design to individual consideration set of
women

• In the last year, new modern methods have been
  introduced in the market next to the pill IUS,
  Implant, vaginal ring
• Many women only accept a few methods, and
  strongly reject other methods
• If all methods are shown in a choice task, women
  would purely choose product based on method
  method would be dominating the choice behavior,
  and no further insights into the sensitivities of
  the other product characteristics are shown
• To avoid that, adapt the choice situation to the
  individual consideration set of each woman
  exclude non-compensatory items

68
Adapted conjoint to individual consideration set
(1)
69
Adapted conjoint to individual consideration set
(2)

• 12 choice tasks with 2 most preferred methods

We used partial profile for the additional
attributes (benefits) in order to make the
choices easier for women
70
Example Conjoint application to segmentation

• Latent Class Analysis can group women based on
  sensitivities to different product characteristics

Method
NB fictitious results

• Where to find the segments?
• E.g. women driven by a positive effect on
  bleedings are mainly older women, currently not
  using any hormonal contraceptive, very active
  women. Mainly married. More in Germany.

71
Who are they?

• CHAID analysis has been used to describe the
  market segments that were defined by using Latent
  Class
• E.g. to answer the question who are the women
  that are mainly driven by benefit A?
• CHAID analysis searched in the data for factors
  that differentiates women in the different
  segments (besides choice behavior)

• CHAID analysis results in a structured (layered)
  depiction of the most differentiating factors
  between the different segments

72
Described classes
E.g. acne driven women ( of women) are
primarily young women, having no kids yet, mainly
living in Europe, having an outgoing lifestyle,
getting information from the Internet, visiting
the GP twice a year, suffering from acne, and
having varying sexual relationships
73
Key learning

• Make clear assumptions about the choice situation
  of women, and adapt the individual choice tasks
  to the individual mind set of each woman

74
Verifying the assumptions afterwards

• We also asked the women to score all methods on
  level of acceptability for verifying the
  assumptions afterwards

75
Verifying the assumptions

• After cross-checking the results
• Products that are ranked 1st or 2nd most
  preferred, are likely to be accepted
• Products that are ranked 3rd, 4th, 5th, or 6th
  have an increased likelihood to be rejected
• However, pill is longer acceptable
• Consideration we could also have included the 3
  most preferred methods

NB fictitious results
76
Structure of the presentation

• Introduction to SKIM Analytical Healthcare
• Types of marketing challenges through the product
  lifecycle
• Introduction to Conjoint Analysis
• Designing a Conjoint Study for different
  marketing challenges Intelligent Design Equals
  Impactful Results
• Summary of Conjoint Designs

77
Conjoint variations (Starting point)
Increased realism for an individuals choice
setting
Increased realism for the market
Standard conjoint
? least complex
78
Conjoint variations (1st dimension)
Increased realism for an individuals choice
setting
? Increased complexity
Limit choices to individual consideration set
Show individual patient cases
Always include respondents current product in
the choice set
Different modules for different choice situations
(e.g. patients)
Constant Sum
Standard conjoint
Not fully developed yet
79
Conjoint variations (2nd dimension)
? Increased complexity
Increased realism for the market
Prohibitions / Conditional relationships
Holistic conjoint
Standard conjoint
Add some realistic fixed choice tasks
80
Conjoint variations (3rd dimension)
Increased realism for the marketand the
individual
Choosing product with optional features
Product packages (e.g. combination therapy)
81
Conjoint adapted to real life choice context
Choosing product with optional features

• Customized design measuring sensitivities for
  different features , feature prices (separately),
  and total model price. To answer business
  questions like
• Which features are most appealing?
• Which features should be standard and which
  should be optional
• What is the role of feature price and model
  price in the choice
• What is the maximum price for a certain feature?
  What is the optimal price?

82
Summary Intelligent Design Equals Impactful
Results
Increased realism for an individuals choice
setting
Limit choices to individual consideration set
Increased realism for the marketand the
individual
Show individual patient cases
Always include respondents current product in
the choice set
Increased realism for the market
Different modules for different choice situations
(e.g. patients)
Constant Sum
Prohibitions / Conditional relationships
Holistic conjoint
Standard conjoint
Add some realistic fixed choice tasks
83
ACA as decision support system

• ACA can also be used as a decision support system
  for patients or physicians e.g. which insulin
  pen to choose for a child?
• As ACA provides direct individual results on
  importance and sensitivity of different product
  characteristics, it can be used to advice the
  respondent on a certain product / service choice
• We ask patients / physicians to trade off the
  treatment outcomes
• And provided them with feedback

84
Conjoint Analysis Preference and choice models
Preference models
Choice models
Respondents rate degree of preference among
product options or rank product options
Respondents choose the product they prefer among
several product options
85
Preference vs.. Choice
Preference models
Choice models

• -Multiple concepts (none option)
• Max. 10 attributes
• Max. 15 levels per attribute
• -Choice
• Interaction possible
• Aggregate utilities

• -2 concepts per task (or rank)
• -Max. 30 attributes
• Max. 9 levels per attribute
• -Rating or ranking
• -No interaction possible
• -Individual utilities

More rational, Less complex (less stimuli) More
tasks needed
Mimics reality better, More intuitive When
rational choice is made more complex
86
Strengths Weaknesses of CBC

• Strengths
• Questions closely mimic what buyers do in real
  world they choose from available products Can
  investigate interactions, alternative-specific
  effects Can include None alternative, or
  multiple constant alternatives
• Weaknesses
• - Limited ability to study many attributes no
  more than about six
• - Less efficient interview low ratio of
  information gained per respondent effort
• - Sample sizes slightly larger than with ACA and
  traditional conjoint

87
Strengths of ACA

• Strengths
• Ability to measure many attributes, without
  wearing out respondentRespondents find
  interview more interesting and engagingEfficient
  interview high ratio of information gained per
  respondent effort
• Weaknesses
• -Partial-profile presentation less realistic
  than real world
• -Respondents may not be able to assume attributes
  not shown are held constant
• -Typically not appropriate for pricing research
• -Tends to understate importance of price, and
  within each respondent assumes all brands have
  equal price elasticities

88
For questions about this presentation, please
contact
Jeroen Slappendel Regional Director SKIM
Analytical Healthcarej.slappendel_at_skimgroup.com
Nelson Silva Senior Project Manager SKIM
Analytical Healthcare n.silva_at_skimgroup.com Phone
973-643-0722 E-mail healthcareUSA_at_skimgroup.com
www.skimgroup.com