New Product Forecasting Seminar

1
New Product Forecasting Seminar

• Prepared for
• Pharmaceutical Business Intelligence Group
• January 25, 2005
• Presented by
• Frank Cuoco
• DRSI Group

180 Mount Airy Road, Suite 103 Basking Ridge, NJ
07920 Phone [phone] Fax
908-696-8558 www.drsigroup.com
2
New Product Forecasting Techniques

• Forecasting Philosophy 101
• New Product Forecasting Methods
• Introduction to Patient Based Forecasting
• Defining the Market Size
• Using Growth Curves to Model Product and Market
  Events
• Applying Market Events Using Logistic Regression
• Market Penetration Techniques
• Case Study

3
FORECASTING PHILOSOPHY 101
4

• You can never plan the future by the past.
• -Edmund Burke

5

• The best way to predict the future is to invent
  it.
• - Alan Kay

6

• The future will be better tomorrow.
• - Dan Quayle

7

• I hate this stuff. I dont have the time to do
  forecasting when my marketing plan was due
  yesterday.
• Anonymous Pharmaceutical Product Director

8
Forecasting Impacts All Aspects of Your Business
9
Economic Implications of an Inadequate
Forecasting Process
10
When to Use Internal vs. External Forecasts
11
Market projections (forecasting) are in all
aspects of marketing analytics!

• Sales Analytics detail response analyses,
  sample response analyses (single
  brand/portfolio), segmentation, sales force size
  and structure
• Marketing Mix/ROI Analyses promotion resource
  allocation
• Simulated Test Marketing evaluation of new
  product opportunities
• Patient Flow Modeling identification of
  opportunities in the flow of patients from
  diagnosis through treatment
• Market Structure Analysis understanding of the
  market dynamics and the patient characteristics
  physicians take into consideration when making
  treatment decisions

12
DRSI Group Forecasting Philosophy

• Forecast/models must be
• Easy-to-use
• Transparent
• Explainable
• Modeling is an iterative process close
  collaboration with your internal clients
  throughout the forecasting process is vital
• Optimally a discussion of assumptions not the
  number
• Scrutiny and validation of assumptions is vital
  avoid GIGO
• Must balance desire for detailed assumptions in a
  model with potential for increased forecast error

13
NEW PRODUCTFORECASTING METHODS
14
Summary of Forecasting Methods

• Jury of Executive Opinion
• Sales Force Composite
• Delphi Approach
• - Historical Analogy

- Questionnaires - Discrete Choice / Conjoint
Analysis

• Behavioral
• Predictive
• - Growth Curves
• - Diffusion of Innovation
• - Analogy Methods

15
Introduction to Patient Based Forecasting
16
Simple New Product Forecasting Concepts
Time n
Time n k
Total Patient Population
Market
Diagnosed Population
- Diagnosis 1
- Diagnosis 2
- Diagnosis 3
- etc.
Market Events
- Generics
Treated Population
- New Technology
- OTC Switches
- Govt Impact
- etc.
Product
Product Events
Product Penetration
- New Products
- Generic Intrusion
Volume Penetration
- Managed care (formulary restrictions)
- etc.
17
Defining the Market SizeEpidemiology Approach
18
Defining the Market Size - Epidemiology Approach

• Advantages
• More transparent understanding of market dynamics
• Allows assessment not only of currently treated
  population, but also potential untreated patients
  which may be treated in the future
• Can further stratify by
• Acute vs. maintenance
• Diagnosis
• Age
• Gender
• Severity of disease
• Race
• Results in better understanding of changes in
  market by dynamics

19
Defining the Market Size - Epidemiology Approach

• Disadvantages
• Government sources used for prevalence data may
  be dated
• Must distinguish between treated vs. untreated
  patient populations
• More difficult and time consuming to collect

20
Sources of Epidemiology Information

• Government Data
• Organizations
• Journals
• Large Clinical Studies
• On-Line Databases / Internet
• Syndicated Studies
• Audited Data

21
Patient Estimates - Sources of Error

• Inflated patient estimates (from organizations
  seeking increased RD development)
• Under-estimation of patients
• Treated vs. untreated patient population errors
• Untreated patients due to lack of effective
  therapy
• "Invisible patients
• Patients derived through audited data

22
Defining the Market Size - Epidemiology Approach

• Estimating Incidence and Prevalence
• Exhaust available secondary sources
• NDTI/PDDA
• Online Search
• Syndicated Studies, etc.
• Conduct a targeted literature search
• Compile a list of incidence and prevalence values
  from the literature
• Prevalence can be stratified by age, disease,
  type of therapy, e.g., acute vs. maintenance

23
Defining the Market Size - Epidemiologic Approach

• Example - Estimate the prevalence of Barrett's
  esophagitis and esophageal strictures (sub -
  populations of reflux esophagitis)
• Step 1 - Consult available secondary sources
  for background information and mentions of
  prevalence
• NDTI/PDDA
• Online Search
• Other G.I. Reports
• Step 2 - Perform a targeted literature
  search
• search MEDLINE, EMBASE, Internet
• use key words - epidemiology, esophagitis,
  Barrett's, esophageal strictures

24
Defining the Market Size - Epidemiologic Approach

• G.I. Example (continued)
• Step 3 - Compile a list of prevalence values from
  the literature
• Estimate of prevalence treated of
  Erosive esophagitis (EE) 1.2 of population
  or 3.6 million
• Estimates of complications of Erosive
  esophagitis
• Complication Diagnosis Midpoint
• Barrett's esophagitis 3-12 7.5
• Esophageal ulcer 2
• Esophageal hemorrhage lt.2
• Esophageal perforation lt.2
• Esophageal stricture 8

25
Defining the Market Size - Epidemiologic Approach
G.I. Example (continued)
26
Defining the Market Size - Primary Research

• Advantages
• Can be used to define new or emerging markets
• Can be used when there are inconsistencies in the
  patient data
• Relatively easy to collect
• Can augment secondary data by collecting data not
  found in audits

27
Defining the Market Size - Primary Research

• Disadvantages
• May be costly
• Deadlines may not allow enough time to conduct
  market research
• May be difficult to recruit respondents

28
Defining the Market Size - Primary Research

• Select the Type(s) of Physician(s) to Interview

Standard
Research for
Market
Forecasting
Research
Event
Current
Future
Time
Time
-Practicing - Innovators
Physicians - Thought

Leaders

-Scientists
Current
Future
Measurement
Time
Time
29
Defining the Size of a New and Emerging Market

• Step 1 - Assess the Current Size of the Market
• How many pharmacologic stress testing studies do
  you conduct in an average month? 40
• Please distribute these studies among the patient
  types below
• Diagnosis of CAD (coronary artery
  disease) ____
• Pre-operative Coronary stenosis ____
• surgery Peripheral vascular disease ____
• Other
  pre-operative surgeries ____
• Post - operative Post PTCA (Angioplasty) ____
  
• surgery Post CABG (coronary artery bypass
  graft) ____
• Other
  post-operative surgeries ____

30
Defining the Size of a New and Emerging Market

• Step 2 - Define the Size of the Market in 2 - 3
  Years
• Earlier you stated that you conduct 40 studies in
  an average month with pharmacologic stress
  agents. Based on your current usage and
  understanding of these agents, please estimate
  the average number of pharmacologic studies per
  month you expect to be conducting two to three
  years from now. 85

31
Defining the Size of a New and Emerging Market

• Step 3 - Assess the Future Size of the Market
• Earlier you stated that you currently conduct 40
  studies in an average month with pharmacologic
  stress agents. Based on your current usage and
  understanding of these agents, please estimate
  the average number of pharmacologic studies per
  month you expect to be conducting five years
  from now. 93
• Please distribute these patients among the
  patient types below.
• Diagnosis of CAD (coronary artery
  disease) ____
• Pre-operative Coronary stenosis ____
• surgery Peripheral vascular disease ____
• Other
  pre-operative surgeries ____
• Post - operative Post PTCA (Angioplasty) ____
  
• surgery Post CABG (coronary artery bypass
  graft) ____
• Other
  post-operative surgeries ____

32
Defining the Size of a New and Emerging Market
Step 4 - Graph the data
33
Defining the Size of a New and Emerging Market
Step 5 - Do a Curve Fit to Convert the Data to a
Continuous Distribution
34
Defining the Size of a New and Emerging Market

• Step 6 - Compute Curve Fit Equation and Place in
  Spreadsheet Cells - Plot Data

35
Defining the Market Size - Combination Methods

• Most accurate approach, increases precision of
  estimates
• Uses various sources and market estimation
  approaches to assess market size
• Can use data from each method audits,
  epidemiology, and primary research together, or
  do an independent analysis to verify patient
  estimates
• Disadvantage is that it may be more costly and
  time consuming

36
Using Growth Curves to Model Market and Product
Events
37
New Product Forecasting

• Common Models Used
• S-Curve
• Logarithmic
• Logistic
• Linear
• Exponential

38
Growth Curves

• Linear Growth Curves
• Defines a Straight Line
• Variables Increase / Decrease the Same Each Year
• Can be Used to Describe Early Growth of a
  Product Some Types of Population Growth
• Relationship Between Dependent and Independent
  Variables Must be Linear

y dependent variable x independent variable a
constant (expected value of y if x is zero) b
expected change in y for each change in x
39
Growth Curves (cont.)

• Exponential Growth Curves
• Useful When Forecasting Variables That Appear to
  be Increasing / Decreasing at a Constant Compound
  Rate
• Can be Used to Model a New Product's Early Sales
  Growth
• Useful to Model the Decline Phase of a Product
  Due to a New Product Entry or Generic Intrusion

y dependent variable x independent variable a
constant (expected value of y if x is zero) b
expected change in y for each change in x
40
Growth Curves (cont.)

• Logarithmic Growth Curves
• Can be Used to Model Early Growth and Maturity
  Stages of a Product's Life Cycle

y dependent variable x independent variable a
constant (expected value of y if x is zero) b
expected change in y for each change in x
41
Growth Curves (cont.)

• Logistic
• Useful for Describing the Product Life Cycle
• Can also be Used to Describe Cycle of Market or
  Product Events
• Versatile - Depending on Coefficients, Logistic
  can Mimic Exponential Curves and Logarithmic
  Curves
• Maximum Point Must be Set by the User

y dependent variable x independent variable a
constant (expected value of y if x is zero) b
expected change in y for each change in x
42
Curve Fitting and Forecasting

• Year History

1
10
2
11
3
12
4
13
5
15
6
17
7
22
8
27
9
32
10
40
11
48
12
56
13
63
14
70
15
76
16
80
17
83
18
84
19
84.5
20
84.5
43
Curve Fitting and Forecasting (cont.)

• Year History

Many models may fit the data well but provide
poor forecasts!
44
Using Growth Curves

• How do I pick the right curve?
• Procedure
• Examine and Graph Historical Data
• Fit Data to as Many Curves as Possible
• Look for ...
• High R2 - Signifies Good Fit to Data
• Most Important Does the Forecast Make Sense?

45
Anatomy of an "S" Curve
Logarithmic
Linear
Exponential
46
Applying Market Events Using Logistic Regression
47
The Logistic Curve Background

• One in the Family of "S" Curves
• Today, it is often used to describe the product
  life cycle
• Also used to describe the growth and decline of
  products and population growth
• Components Maximum Point or Saturation Level,
  Constant, Slope, and Time

48
Why use Logistic Curves?

• Ability to Interpolate Between Points
• Automate Model
• Turn events on or off
• Phase in effects
• Make model responsive to other events
• Ability to Easily Make Dynamic Changes to Model
  Assumptions
• Change maximum point
• Change launch date
• Change rate of growth

49
Logistic Curve

• Components
• Max Maximum penetration
• Constant Determines curves position
• Slope Measure of time to saturation
• Time
• Logistic Equation
• Max/(110(Constant SlopeTime))

50
Applying Market Events Using The Logistic Curve

• Forecast Data

G.I. Market Patients (millions)
Year History Fit Forecast 1997 16.24 16.27
1998 16.94 16.98 1999 17.44 17.47 2000 17.79 1
7.80 2001 18.02 18.00 2002 18.18 18.15 2003 18.
28 18.24 2004 18.35 19.30 2005 18.33 2006 18
.35 2007 18.37 2008 18.37 2009 18.38 2010
18.38 2011 18.39
51
Applying Impacts Using the Logistic Curve

• Step 1 - Get Assessment of Market Impact - Rx to
  OTC Switches
• Maximum impact 10
• First year 2
• Third year 7.0
• Step 2 - Graph Fit the Data
• Maximum 10
• Constant 1013.769
• Slope -0.5053

52
Applying Market Events Using the Logistic Curve
(cont.)

• Step 3 - Compute Logistic Curve Fit Equation and
  Place in Spreadsheet Cells
• Logistic Equation - Max/(110(Constant
  SlopeTime))
• Maximum 10
• Constant 1013.769
• Slope -0.5053
• Maximum/(110(Constant SlopeTime))
• .10 /(110(1013.769 -.5053 Year))

53
Forecasting Future Events Using the Logistic
Curve - Summary
54
Market Penetration Techniques
55
Market Penetration Techniques - Analogy

• Process
• There are four steps needed to estimate the
  uptake of a product using the analogy technique
• 1) Select an analogy or determine if a pattern
  exists in the launch data
• 2) Convert launch curve to a continuous
  distribution
• 3) Use conjoint or other market research as the
  maximum point
• 4) Adjust uptake curve and peak share based on
  market and product events

56
Market Penetration Techniques Analogy (cont.)

• Example - New Proton Pump Inhibitor
• Step 1 - Examine Historical Product Launches
  select one average product

57
Market Penetration Techniques Analogy (cont.)

• Example - New Proton Pump Inhibitor (cont.)
• Step 2 - Fit Data to a Logistic Curve
• Maximum 4.20
• Constant 344.37
• Slope -.17

58
Market Penetration Techniques Analogy (cont.)

• Step 3 - Substitute preference share peak value
  or other maximum value from another study and
  place formula into a spreadsheet.
• Logistic Equation Maximum Patient
  Share/(110(constantslopeyear))
• Peak Market Share 8 (year 2009)
• Substituting values into the equation gives us
• 0.08/(110(344.37-.17years from launch))

59
Applying Product Impacts Using the Logistic Curve
(cont.)

• Step 3 - (cont.)
• Logistic Equation - Max/(110(Constant
  SlopeTime))
• Maximum 8
• Constant 344.7
• Slope -.17
• Maximum/(110(Constant SlopeTime))
• .08 /(110(344.7-.17 Year))

60
Market Penetration Techniques Analogy (cont.)

• Step 4 - Adjust preference share based on
• Coverage and detailing levels
• Distribution
• Indications approved
• MC formulary acceptance
• New product entries
• Generic entries
• Order of entry
• Price Sensitivity

Unadjusted Share
Adjusted Share
61
Applying Product Impacts Using the Logistic Curve

• Modeling The Impact of Competitive New Product
  Entries
• Step 1- Get Assessment of New Product Impact
• Maximum impact 5
• First year 1
• Third year 3.5
• Step 2 - Graph Fit the Data
• Maximum 5
• Constant 823.088
• Slope -0.41024

62
Applying Product Impacts Using the Logistic Curve
(cont.)

• Modeling The Impact of Competitive New Product
  Entries
• Step 3 - Compute Logistic Curve Equation and
  Place Spreadsheet Cells
• Logistic Equation - Max/(110(Constant
  SlopeTime))
• Maximum 5
• Constant 823.088
• Slope -0.41024
• Maximum/(110(Constant SlopeTime))
• .05 /(110(823.088 - 0.41024 Year))

63
Forecasting Future Events Using the Logistic
Curve - Summary
64
Modeling Complex Situations Using Combination
Approaches

• Jury of Executive Opinion
• Sales Force Composite
• Scenario Development
• Delphi Approach
• - Historical Analogy

- Questionnaires - Discrete Choice / Conjoint
Analysis

• Behavioral
• Predictive
• - Growth Curves
• - Diffusion of Innovation
• - Analogy Methods

65
Modeling Complex Situations Using Combination
Approaches

• Situation Need to determine the optimal
  promotion resource allocation across a product
  portfolio
• Solution Conduct promotion ROI analysis and
  physician level detailing response analysis for
  each brand and build a model to determine optimal
  promotion spend
• Situation Need to quantify the impact of 16
  market events in a CNS market for strategic
  planning purposes
• Solution Conduct event based quantitative market
  research and build a custom forecast model
• Situation Need to help RD prioritize 5
  potential life cycle indications
• Solution Conduct discrete choice quantitative
  market research and build a market model for each
  indication to model the product with different
  product profile/launch scenarios
• Situation Need to quantify the impact of
  potentially unfavorable class labeling on brands
  and market
• Solution Conduct quantitative market research
  and build a market model to simulate different
  labeling scenarios and their impact on the
  product/market.

66
Asthmarrest Forecast Case Study
67
Asthmarrest Case Study Product Description

• Asthmarrest is a leukotriene D4 receptor
  inhibitor indicated for the chronic treatment of
  asthma in adult patients. Clinical studies show
  that Asthmarrest has approximately 70 efficacy
  in treating chronic asthma. The recommended dose
  of Asthmarrest is one 10mg tablet taken once
  daily 1-2 hours after a meal. Asthmarrest is not
  indicated for the treatment of acute episodic
  asthma. Asthmarrest is contraindicated in
  patients who are hypersensitive to Asthmarrest or
  any of its inactive ingredients.

68
Asthmarrest Case Study Product Assumptions

• Launch date Q1 2006
• Cost/day vs. other products 2.50/day
• Oral solid
• Indicated for treatment in adult asthma only
• Slightly better efficacy compared to other
  leukotriene antagonists in clinical trials
• Side effects not distinguishable from placebo
• Overall slightly better product profile than
  "gold-standard"

69
Asthmarrest Case Study Issues to Consider

• Potential Market
• Order of Entry
• Pricing - Cost per day vs. competitors
• Comparison to other products

70
Asthmarrest Case Study Constant Sum Question

• Doctors were asked what percentage of their
  asthma patients they place on current asthma
  agents. Then, doctors were given the Asthmarrest
  concept statement and asked what percentage of
  their asthma patients they would place on
  Asthmarrest and the other asthma assuming it was
  FDA approved and available.

71
Forecasting Exercise
72
Appendix ALogistic Regression
73
Logistic Regression

• Preference Share MAX / (110(CONSTANT SLOPE
  Year))

Slope
Max
Constant
MAX represents the saturation point CONSTANT
determines the curve's placement SLOPE is a
measure of the time to saturation