Credentials

1
GfK Group
GfK HealthCare
Martin Hamblin GfK, Inc
May 2004
Click to edit Master title style

• Click to edit Master text styles
• Second level
• Third level
• Fourth level
• Fifth level

Improving Decision Making in Portfolio
Development Allan Bowditch, Chairman Alastair
Bruce, General Manager Martin Hamblin GfK,
Inc. PBIRG Annual Conference Meeting May 2004
San Diego Pharma Biotech Gaining the Best of
Both Worlds
2
The business enterprise has two and only two
basic functions marketing and innovation.
Marketing and innovation produce results all the
rest are costs. Peter Drucker
3
Historically, marketing has been driven by
IT
4
Focus

• Marketing is both an art and a science, and like
  any other investment activity, it must be
  grounded in research, planned carefully, and
  measured and evaluated based on return on
  investment.
• Guided by this belief, our philosophy toward
  portfolio optimization is focused on a careful
  balance of creative intuition and rigorous
  analysis of bullet-proof data.

5
To develop a model to help manage how a company
can develop an optimal product range in a given
market / franchise
Objective

• Identify the key treatment drivers, by disease,
  and assess the value of candidate products that
  offer these drivers
• Qualitative investigation Triads by physician
  specialty
• Discriminate between the relative value of
  differing levels of each attribute i.e.
  clinical end-points
• Adaptive Conjoint Analyses ACA
• Assign past and future values to these
  end-points, to permit future market modeling and
  forecasting
• Attribute Dynamic Evaluation ADE
• Create a market framework
• Epidemiological and other secondary data Time
  Series

6
Franchise Forecasting Model Rationale

• Any product can be described as a collection of
  clinical end-points
• Careful selection of end-points creates a
  lexicon, a vocabulary
• The vocabulary of utilities allows any product to
  be described
• An assemblage of end-points is, in effect, a
  surrogate for a product
• Assign a value to each end-point utility value
  allows that end-point to be compared
  quantitatively with any other, within a context

7
Franchise Forecasting Model Rationale

• Static preferences utility values are
  transformed into dynamic ones that vary over time
• Dynamic preferences are integrated with secondary
  data into a dynamic Franchise Forecasting Model
• Franchise Forecasting Model thus allows for
  market simulations of various scenarios

8
Franchise Forecasting Model Components
Market Framework
Utility Values
Dynamic Values
9
FFM Interview Process Overview

• Interviews conducted with relevant specialty
  groups e.g.
• Primary care physicians
• Internal Medicine
• Oncologists
• Cardiologists
• Etc
• Semi-structured face-to-face interviews (but
  online also possible depending on complexity of
  exercise.)
• Interviews include an Adaptive Conjoint Analysis
  (ACA) component
• Physicians also be asked to fill out
  self-completion sheets pertaining to the
  trade-off questionnaire within the
  disease/condition context.
• Ideally, interviews last no more than 45 minutes.

10
Franchise Forecast Modeling Study Flow
STAGE 1 Qualitative Identify clinical drivers
STAGE 3  Build ACA model in EXCEL to allow
product preference simulations, by disease stage
Franchise Forecasting Model ACA values
Dynamic values Market Forecast
STAGE 3  ADE commercial forecasting model, by
disease type
Identify variables correlated to DRIVERS
STAGE 2 Quantitative measurements of relevant
end points ACA
                                                 
             
11
Case Study Example
12
Sample Disposition Discussion

• The sampling design of this study was directly
  related with the number of attributes tested (42
  attributes).
• For technological, methodological and practical
  reasons, 42 attributes could not be tested in a
  single ACA exercise.
• a built-in limitation of the ACA system, which
  cannot evaluate more than 30 attributes at any
  one time.
• methodological and practical considerations an
  elevated number of attributes would require
  respondents to answer a list of trade-off
  questions so long it would be absolutely
  impractical to implement.
•  
• To overcome this limitation, Martin Hamblin GfK
  proposed a split sample design.
• Respondents were divided into two matching groups
  and each group evaluated half the attribute set
  with a minimum of 5 common attributes evaluated
  by both groups this design allowed for the
  measurement of all 42 attributes covering 135
  levels or utilities in an integrated model.

13
Sample Disposition (continued)

• Interviews were divided in 5 treatment groups
• Condition X
• Condition Y
• Treatments A
• Treatments B
• Treatments C
• In order to achieve stable ACA results (30) for
  all relevant specialties in all 5 treatment
  groups a minimum sample of n 780 was required.
  The chart below displays the sample design in
  detail.

Condition Overall
14
Adaptive Conjoint Analysis (ACA)
Allow to respond to the question which product
attributes customers value most ?
B
A
Step 1 list attributes
B3
A3
Step 2 define a range of levels for each
attributes
B2
A2
B1
A1
Step 3 physician is asked to make choices
between various attributes levels
For each attribute level Utility index lt
0 Utility index gt 0 Utility index 0 For each
attribute Utility index delta between
highest and lowest utility index
Output is a utility index for each attribute
level
15
Importance of Drug Attributes Current, Past
Future Top Ten Attributes By Time Frame
16
Detailed Process Overview

• IMS TRx 97 Q1 to 03 W2

• PATIENTS
• TRx x / DACON
• -------------------------
• 91 (days/quarter) x COMPLIANCE

• 6. Different PATIENTS for each EXISTING product
  group

• 14. Different PATIENTS for each EXISTING product
  group for 2003 to 2020 (DYNAMIC FORECASTS

• 20. Different PATIENTS for each EXISTING product
  group and NEW product for 2003 to 2020 (DYNAMIC
  FORECASTS)

• 8. Different PATIENTS for each EXISTING product
  group for 2003 to 2020 (STATIC FORECASTS)

• 2. IMS DACON YTD Jun 2003

• 3. COMPLIANCE for each EXISTING product group

• Verispan PDDA Users
• 97 Q1 to 03 Q2 Type 1, Type 2 IGT
  (Dysemetabolic Syndrome X)

• EPI Prevalence
• 1997 to 2020 Type 1, Type 2 and IGT

17
Detailed Process Overview

• 14.PATIENTS for each EXISTING product group for
  2003 to 2020 (DYNAMIC FORECASTS

• 20.PATIENTS for each EXISTING product group and
  NEW product for 2003 to 2020 (DYNAMIC
  FORECASTS)

• 11. Patient types ACA Utilities for 1997 to 2020
  (24 levels dynamic)

• 13. Patient Types PREFERENCE SHARES for each
  EXISTING product group 1997 to 2020

• 9. ACA Utilities for 138 Levels (3 interpolated)
  over 42 attributes 2002

• 10. Relative importance for 24 selected attribute
  levels
• 1997 (-5 years), 2007 (5 years), 2012 (10
  years)

• 12. PROFILES for each EXISTING product group

• 17. PREFERENCE SHARES for each EXISTING product
  group NEW product for 1997 to 2020

18
Detailed Process Overview

• 17. Patient Types PREFERENCE SHARES for each
  EXISTING product group NEW product for 1997 to
  2020

• 15. PROFILES for each NEW product

• 16. UPTAKE for each NEW product for 2003 to
  2020

• 18. CANNIBALIZATIONS for each NEW product

• 19. COMPLIANCE for each NEW product

19
(No Transcript)
20
Gap Analysis Condition X
90
80
70
60
50
40
30
20
10
Importance
Satisfaction
0
Attribute I
Attribute F
Attribute J
Attribute L
Attribute T
Attribute A
Attribute B
Attribute C
Attribute D
Attribute E
Attribute G
Attribute H
Attribute K
Attribute M
Attribute N
Attribute O
Attribute P
Attribute Q
Attribute S
Attribute U
Attribute V
Attribute X
Attribute Y
Attribute W
21
Importance vs. Satisfaction Condition X
22
Condition Y Treatment A
PERCEPTUAL MAP
Type 2 - oral only
21
Product E
22
3
12
7
2
5
13
Product B
Product A
1
20
19
24
11
6
8
18
10
Product C
17
Product F
9
Product G
4
16
Product D
23
14
15
The Model
PBIRG Example ADE.xls
23
Example of ADE Model
24
Outcomes Discussion

• Identifies the level of satisfaction with
  existing treatments in certain market sectors
• Indicates the risk/benefit as a result
• Identifies the key future requirements
• Consequently provides an early means of screening
  in/out potential new product candidates (either
  internal or external) for specific sub markets
  within overall area
• Highlights the minimum market profile for the new
  product in the separate sub markets
• Models the effect of the new product entry on
  existing, and other newly launched products from
  the company.
• Highlights the optimal product mix
• Provides more specific information for making
  go/no go decisions
• Helps in market segmentation of specific patient
  and physician types

25
GfK Group
GfK HealthCare
Martin Hamblin GfK, Inc
May 2004
Click to edit Master title style

• Click to edit Master text styles
• Second level
• Third level
• Fourth level
• Fifth level

Improving Decision Making in Portfolio
Development Allan Bowditch, Chairman Alastair
Bruce, General Manager Martin Hamblin GfK,
Inc. 2004 PBIRG Annual Conference Meeting May
2004 San Diego Pharma Biotech Gaining the Best
of Both Worlds