How to choose market share techniques for a new product forecast

How to choose market
share techniques for a
new product forecast
Vision in Business – Lisbon, Portugal
By Rafaat A. Rahmani
President
[email protected]
Thursday 15th November 2007
Copyright Lifescience Dynamics Ltd. 2007
Agenda
About
About us
us
Introductions
Introductions
Quick
Quick ‘n’
‘n’ Dirty
Dirty for
for early
early phase
phase
Analogue
Analogue based
based
Monté
Monté Carlo
Carlo Simulation
Simulation
External
External expert
expert opinion
opinion
Fixed
Fixed product
product profile
profile
Variable
Variable product
product profile
profile
Tips,
Tips, sources,
sources, wrap-up
wrap-up
Copyright Lifescience Dynamics Ltd. 2007
2
Agenda
About
About us
us
Introductions
Introductions
Quick
Quick ‘n’
‘n’ Dirty
Dirty for
for early
early phase
phase
Analogue
Analogue based
based
Monté
Monté Carlo
Carlo Simulation
Simulation
External
External expert
expert opinion
opinion
Fixed
Fixed product
product profile
profile
Variable
Variable product
product profile
profile
Tips,
Tips, sources,
sources, wrap-up
wrap-up
Copyright Lifescience Dynamics Ltd. 2007
3
We offer comprehensive
and end-to-end services
An overview of Lifescience Dynamics
Market
Research
Competitive
Intelligence
Valuation
& Modelling
Strategic
Consulting
Emotional &
rational
Scientific &
commercial
Market simulation
& scenario analysis
Strategic advice,
review & validation
Qualitative &
quantitative
Policy markers &
influencers
Drug and disease
area forecasting
Business planning
& implementation
Solutions based on comprehensive data,
exhaustive analytics, years of experience &
multi-disciplinary team efforts
4
Copyright Lifescience Dynamics Ltd. 2007
Agenda
About
About us
us
Introductions
Introductions
Quick
Quick ‘n’
‘n’ Dirty
Dirty for
for early
early phase
phase
Analogue
Analogue based
based
Monté
Monté Carlo
Carlo Simulation
Simulation
External
External expert
expert opinion
opinion
Fixed
Fixed product
product profile
profile
Variable
Variable product
product profile
profile
Tips,
Tips, sources,
sources, wrap-up
wrap-up
Copyright Lifescience Dynamics Ltd. 2007
5
A quote to set the scene…
“Forecasting is a complex modelling exercise...so
why should we make it more complex! "I try to make
things as simple as possible, but not simpler"
Albert Einstein
6
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Predicting fails, but forecasting is an
essential feature of any business
ƒ Forecasting is more art than science, therefore, is inexact by
definition.
ƒ Forecasting is about predicting future events............so uncertainty
can be reduced but can not be eliminated.
ƒ Forecasting is a set of techniques for understanding markets at a
fundamental level.
7
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Forecasting is a decision-making
tool
ƒ
ƒ
ƒ
ƒ
In order to grow, a company must consistently make good business
decisions (decisions that increase the value of the company).
Development of pharmaceutical products is expensive and risky, and it
often involves long time horizons.
In order to make quality investment decisions, it is useful to know as
much as possible about the future value that can be expected from the
investment.
We create models in an attempt to:
•
•
•
•
•
•
8
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
predict market evolution
predict the revenue
reduce risk and increase accuracy
go / no go decisions about new and existing projects?
what indications should be targeted?
how much can be spend on promotion?
What drives sales?
ƒ Start by asking simple questions
•
•
•
•
•
How many diagnosed patients are out there?
How many of them are/will be treated?
What are the trends in current therapy?
What could happen in the future to impact therapy class?
Where does our new product fit in?
ƒ
ƒ
1st line, 2nd line, 3rd line
Type of patients: mild, moderate, severe
• How much of it will the patient consume?
ƒ
ƒ
ƒ
Duration of treatment
Dosage
Compliance
• What is the price?
ƒ Some not-so-simple answers
• What will be the uptake of this product
9
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
The challenges of forecasting
Influence
Advocacy Groups
Opinion Leaders
Purchasers
Patients
Trusted Colleagues
Carers/Relatives
Advertising Publications
Demand
Awareness
Colleagues
Sales reps
Decision
Evaluation
s
S to c k
Point
of
use
Generics
Formularies
bility
a
l
i
a
v
A
Hospitals
Supply
10
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Wholesalers
Protocols of care
Decision making
Algorithm:
Competition, Rx,
Alternatives &
Substitutes
A robust model incorporating key
drivers for market penetration
•
•
•
•
•
•
Unmet needs
Order of entry
Competition - generic
Competition - branded
Protocols of diagnosis
and treatment
•
•
•
•
•
•
•
Therapy
area
•
•
•
•
•
•
•
11
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
•
Resistance to switch
Proof/believability of
offer for new molecule
Budgetary pressures
Reimbursement
Formulary approval
Healthcare reform /
policy
Advocacy Groups
Patients’ requests DTC
Market
access
Product profile
Product perception
Compliance
Number of indications
Number of markets –
US/EU/JP
(Brazil/China/India??)
Price
Product
related
Company
related
•
•
•
•
•
Promotion
Field force
effectiveness
Image of company
within
the therapy areas
Strength of therapy
franchise
Types of long term forecasting
Inputs
Features
12
Implications
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Scenario
Planning
Advanced
model – What
model
if analysis
Super
express
model
Express
model
Secondary research
9
9
9
9
Epidemiology/procedure-based
9
9/2
9
9
2
9/2
9
9
2
2
9/2
9
Rx-based
2
9
9
9
Analogue used
2
9
9
9
Competitive intelligence
2
2
9
9
Multiple products
2
9/2
9
9
Monté Carlo
2
9/2
9
9
Dynamic (VB based)
2
9/2
9
9
Multiple
product profiles
2
2
9/2
9
Web based
2
2
2
9
Speed
+++
+
-
---
Flexibility
---
+
++
+++
Cost implications
+++
+
--
---
Primary research
Juster Scale with PRF
Primary research
Conjoint with PRF
Agenda
About
About us
us
Introductions
Introductions
Quick
Quick ‘n’
‘n’ Dirty
Dirty for
for early
early phase
phase
Analogue
Analogue based
based
Monté
Monté Carlo
Carlo Simulation
Simulation
External
External expert
expert opinion
opinion
Fixed
Fixed product
product profile
profile
Variable
Variable product
product profile
profile
Tips,
Tips, sources,
sources, wrap-up
wrap-up
Copyright Lifescience Dynamics Ltd. 2007
13
Market share estimation is most
important aspect of forecasting
ƒ
ƒ
ƒ
ƒ
ƒ
Market share estimate cause most discussion and challenges
Typically market shares & prices tends most sensitive to final revenues
Generally market share estimates are based on guess work and
therefore, the least robust number in a forecast
Market share is function of many variables – some in management
control but largely controlled be competitors and underlying marketing
issues.
There are some rule of thumb for market share estimations as listed
below.
•
Order of entry
ƒ
•
Promotion
ƒ
•
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
The more you promote, the greater your market share but there is a limit
Acceptance rates
ƒ
14
First product gets the largest share in a new class of drugs
As new competitor enters a new market, it is accepted more quickly than the
previous one
Challenges facing market share
calculations
Time delay /
Response time
Non-linear
responses
Feedback
Residual /
Carry on
sales
15
Copyright Lifescience Dynamics Ltd. 2007
[email protected]om
Cause &
effect
Model development is an
iterative process
Primary and
secondary
research
analysis
Experience &
Intuition
Define model structure
(patient flow)
Build model framework /
Refine model framework
All inputs are variable
and can be changed:
• When new data become
available
• To model different
scenarios
• To assess the impact of
small changes
16
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Insert new data / alter
variables
Interpretations
Review
outputs
Insert
data
Typical modelling steps
5
4
3
Run live data & Check against known disease areas, back of envelop
Calculations and other brokers estimate, sales data
Sanity check
Input dummy data such as 1 sufferer per 1,000;
Check the logic & test put some outliers
Workout a mathematical
relationship
Epi pop.'s x % diagnosed x % drug-treated x
patient share % (for drug) x drug price x
days or cycles of therapy x compliance rate
Understand the interaction and identify what
2
1
17
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Observe relationships & data need
data would be required
Understand the disease area, patient flow
Define the problem
patient segments, referral, treatment algorithms
Refine the model and improve
data based on initial results
5
4
Re-visit
problem based
on results?
3
Run live data & Check against known disease areas, back of envelop
Calculations and other brokers estimate, sales data
Sanity check
Input some dummy data such as 1 sufferer per 1,000
Check the logic & test Put some out layers
Workout a mathematical
relationship
Epi population's) x % Diagnosed x % Drug-treated x
Patient share % (for drug) x Drug price x
days or cycles of therapy x compliance rate
Understand the interaction and identify what
2
1
Observe relationships & data need
Data would be required
Understand the disease area, patient flow
Define the problem
18
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
patient segments, referral, treatment algorithms
Refine
data/
assumption
/ seek
an alternative
Patient based model should be
built around patient flow
Funding
Seek Treatment
GP / PCP / FP
Diagnosis
Tests &
Diagnostics
OTC
Hospital
Rx Filling
( 20 Care
Pharmacy)
Chronic Vs. Acute
Rx Algorithm
Dosing
Compliance
S/E Complications
On going maintenance
1.
2.
3.
4.
19
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
It will go away
Try home remedy
OTC
See GP or A&E
(Hospital)
Initial Treatment
Prescription
Rx Filling ( 10 Care
Pharmacy)
Cure Vs. Treatment
Funding
Not feeling
well / symptoms
lat
ed
re
t
ke
ar
M
Rx data
20
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Price
ts
en
ev
CI &
Analysis
re
tu
Fu
Market Opportunity
+ Treated Patients
Reconciliation
Reconciliation &
&
Trend
Trend analysis
analysis with
with
curve
curve fitting
fitting
re
lat
ed
Final Forecast
Pr
od
uc
t
Patients/Epi
Patient consumption
Schematic overview of modelling
process and data flow
Robust approach marries
epidemiology, Rx, primary data
For example:
Population Growth Rate
Prevalence, Incidence, Diagnosed Rates
Target
Indication &
Epidemiology
Product profile
Vs Competition
Æ Peak Shares, Cannibalisation Factors by segment
Treatment Parameters
DOT / Rx
Rx Market Share
projected for each
product class
$/Day for Cox2i
$/Day for each
$/Rx for NSAIDs
product class
& Analgesics
21
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Historical (to date)
Rx Data by Drug
Rx 000’s
IMS Rx Data
Conjoint & primary research data
Forecast Modelling
Peak year share
Years to peak
Shape of the curve
Curve Fitting, Uptake Curves,
Bass Diffusion
Forecast
Rx
Pricing
Market Shares
converted back to
Rx 000
$
Rx Converted to
Gross & Net
Revenue
Share modelling and curve fitting
ƒ Availability of data and access to computers has meant that
econometric methods are readily available to model and forecast
market share.
• However, controversy exists over their usefulness
ƒ Market researchers have historically had a number of tools like
focus groups and surveys to gather intelligence for new product
acceptance.
ƒ Most companies have few systems in place allowing managers to
incorporate marketing assumptions, industry knowledge, market
research, and prior product performance in a quantitative long run
forecasting framework.
ƒ One solution is the development of a process based upon
diffusion theory and the product life cycle.
22
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Projection of World-Wide PC Demand,
1999-2010-Data From Bill Gates,
Newsweek
Actual Worldwide PC Shipments, 1981-1999 and Fitted and Projected
Shipments, 1981-2010, m=3.384 Billion, p= .001, q= .195
Peak
2008
Shipments Includes Replacements
(Upgrades)
180
160
697 Million Units
Shipments through
1999
M illions of Units
140
120
100
80
60
40
20
Year
World Wide PC Shipments
23
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Fitted World Wide PC Shipments
09
20
07
20
05
20
03
20
01
20
99
19
97
19
95
19
93
19
91
19
89
19
87
19
85
19
83
19
19
81
0
There are a number of curve
fitting algorithms
ƒ
To perform curve fitting
•
Define a function
ƒ
ƒ
•
•
•
ƒ
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
The function is then minimised to the smallest possible value with respect to
the parameters.
The parameter values that minimise the function are the best-fitting
parameters.
In some cases, the parameters are the coefficients of the terms of the model
Types of curves
•
•
•
•
•
24
Which depends on the parameters
That measures the closeness between the data and the model
Diffusion Curves
Gompertz Curve - A growth curve
Linear / multiple-linear regression
Exponential curves such as Weibull
Gaussian (Normal) Distribution
Choose models that will be easy
to administer
ƒ If there are too many variables, or could not define a function then
choose a model-free fit
• Neural networks
• Cubic splines
ƒ Choose model based on the following criteria
•
•
•
•
•
25
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Predictive value
Ease of implementation
Efficiency
Ease of explanation
Defensibility
New market models
(Bass-type Models)
ƒ
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
ƒ
Innovators: those ‘very likely’ to take up a concept immediately - measures used
vary, but elite ‘top box’ scorers
Imitators: non-rejectors, who will eventually acquire
ƒ
ƒ
ƒ
ƒ
ƒ
Required input:
Year 1 penetration (y=innovators, from ad hoc research)
Final penetration (x=innovators+imitators - from ad hoc)
Coefficient (z1) for innovators (0.005 = half percent)
Coefficient (z2) for imitators (0.4 = typical value)
ƒ
ƒ
ƒ
ƒ
ƒ
Equation (actual numbers):
Year 1 = innovators
Year 2 = (y*z1) +
([z2-z1]*previous year total) ([z2/y] * previous year total2)
The Diffusion Curve
ƒ
ƒ
The Bass Diffusion Model describes the diffusion of new technologies
into consumer markets and is being used extensively in Pharma as well.
In the Bass Model, there are two characteristic types of consumer:
•
•
Innovators - make purchasing decisions based on own evaluation of the pros
and cons of the new technology. Described as being “Internally Influenced.”
Imitators - make purchasing decisions based on the example of others, and
who only adopt a new product when their contemporaries have deemed it
valuable. Described as being “Externally Influenced.”
100%
ƒ Each type of consumer has a
characteristic uptake curve.
ƒ The overall adoption curve
(diffusion) is a blend of the two
curves, based on the relative
preponderance of each
consumer type.
90%
70%
60%
50%
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Launch
Peak
Share
40%
Years to peak
30%
20%
10%
0%
2005
27
Type of market
80%
2010
2015
2020
CI helps in improving input for
modelling future events
ƒ
A event as anything which can affect the course of a market. Events include:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
ƒ
28
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Changes in labelling for existing products
Changes in price
Changes in promotion
Diagnostics
Loss of exclusivity
New class of drug
New clinical study publications
New dosage – E.g. OD to once a week
New indications
New presentation
New product launches
New types of drug - combination
Off-label use
Restricted use
Vaccine
Events are described by an index date and a set of parameters determining the time evolution
of the system. The mathematical function that models the process is referred to as a
Diffusion Curve.
Key inputs required
ƒ
Diffusion theory uses a number of equations that produce the S-shaped
curve resembling a product life-cycle
•
•
•
ƒ
Introduction Phase
Growth Phase
Maturity Phase
One popular formula is called a Logistic Curve.
•
•
The curve is nonlinear.
Its shape is based upon three pieces of information.
ƒ
ƒ
ƒ
Saturation level > Peak share at 100% awareness level
Location of the inflection point (that point in time where the growth rate is
maximised)
Intensity of the introduction Phase. > Unmet needs
– Described as a delay factor, this intensity usually has a numerical value ranging from zero
to one. A factor close to zero implies a significant amount of pre-selling - i.e. pent up
demand.
– A factor close to one means that sales might be delayed because of distributional
considerations, a tight advertising budget, or delays in training the sales force.
29
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Typical sales curves
EU 5 - Market Share & Gross Revenue
(Revenue on Right Hand Scale)
$150,000
100%
90%
80%
70%
$100,000
$ 000
% Share
60%
50%
40%
$50,000
30%
20%
10%
0%
$0
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030
30
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Order of entry
ƒ Order of entry effect is proportional to first two entrants
(current market)
• e.g., 4th entrant = 14 / (46 + 23) = 14/69 = 0.203
i.e., calibrated share estimate multiplied by 0.203
ƒ Generic order of entry model
Nos of
competitors
Share
2nd
3rd
4th
5th
6th
31
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
1
2
3
4
5
6
100%
67%
53%
46%
41%
37%
33%
27%
23%
20%
18%
20%
17%
15%
14%
14%
13%
12%
11%
10%
9%
Agenda
About
About us
us
Introductions
Introductions
Quick
Quick ‘n’
‘n’ Dirty
Dirty for
for early
early phase
phase
Analogue
Analogue based
based
Monté
Monté Carlo
Carlo Simulation
Simulation
External
External expert
expert opinion
opinion
Fixed
Fixed product
product profile
profile
Variable
Variable product
product profile
profile
Tips,
Tips, sources,
sources, wrap-up
wrap-up
Copyright Lifescience Dynamics Ltd. 2007
32
Modelling uptake based on an
analogue
ƒ
Identification of key drivers and then ranking and rating them
•
•
Hospital vs. community product
Presentation
ƒ
•
•
•
•
•
•
ƒ
Oral, injection, IM and Sub-cut
Dosage
Unmet needs
Chronic aymptomatic
Reimbursement (currently well reimbursed)
Co-pay differential
Disease severity?
Model on similar products e.g. in CNS
•
Usually difficult to find a good analogue
ƒ
? Anti-psychotic drugs
– Risperdal & Zyprexa
ƒ
? Anti-Parkinson's drugs
– Requip & Celance
ƒ
33
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
IMS commercial Analogue database
Agenda
About
About us
us
Introductions
Introductions
Quick
Quick ‘n’
‘n’ Dirty
Dirty for
for early
early phase
phase
Analogue
Analogue based
based
Monté
Monté Carlo
Carlo Simulation
Simulation
External
External expert
expert opinion
opinion
Fixed
Fixed product
product profile
profile
Variable
Variable product
product profile
profile
Tips,
Tips, sources,
sources, wrap-up
wrap-up
Copyright Lifescience Dynamics Ltd. 2007
34
Key features of MC simulations
ƒ
Monté Carlo simulation is a form of simulation
•
•
•
ƒ
Monté Carlo Simulation Monte Carlo simulation helps make better
decisions
•
•
•
ƒ
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
models interdependencies
sensitivity analysis simulations
without simulation, a spreadsheet model will only output a single number
Suppliers
•
•
35
randomly generates values for uncertain variables over and over to simulate a
model
without simulation, a spreadsheet model will only reveal a single (most likely)
outcome
automatically analyses the effect of varying inputs on outputs of the modeled
system.
@Risk, www.palisade.com
Crystal Ball, www.crystalball.com
MC works on the uncertain
variables in the model
ƒ Each uncertain variable is defined with a probability distribution.
• Peak share in worst scenario
• Peak share in most optimistic scenario
• Peak share in most likely scenario
ƒ Distribution of uncertainty is an equation that describes shape and
range
36
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
The Monté Carlo Simulation
ƒ
ƒ
For each input the model user defines a distribution (in terms of minimum, most likely and
maximum values)
The Monté Carlo simulation can be run 5000 times from which a mean result can be
calculated with 90% confidence intervals.
Patients
Peak Share
Result
Launch date & Others...
37
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Monté -Carlo
Simulation
Forecasting
Model
Total GW406 sales until 2020 £ 000s
1.200
1.000
0.800
Probability
Price
Normal Distribution
0.600
Total Global GW406 Sales until 2020
0.400
0.200
0.000
16,035,000
17,435,000
18,835,000
20,235,000
21,635,000
Revenue
Contribution
The model provides a user friendly
tool for forecasting the sales of key
new products
INPUTS
ƒ Total sales per product are derived from:
•
a) patient epidemiology and Rx to calculate total drug treated patients
•
b) market environment and product characteristics to calculate total prescriptions per product
•
c) price per day to calculate the total sales per drug
ƒ The market share calculations are based on Discrete Choice peak shares derived from primary
market research and benchmarks are used to model key competitors
38
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
ƒ
Excel style database, and
viewed and edited using an
Excel front end
ƒ
Allows the user the ability to
create a series of different
scenarios in different databases
ƒ Key outputs (per product)
from the models include
Rx, Rx share, revenue per
product and revenue share.
ƒ
Offers a transparent approach
for storing data and
assumptions used to generate
the forecasts
ƒ The outputs can be
reported to any level of
segmentation included in
the model.
OUTPUTS
The flexibility of the model allows
assessment of numerous
scenarios
ƒ
Ability to run multiple scenarios and to save scenarios powered by Crystal Ball
Monte Carlo Simulation
•
•
•
ƒ
ƒ
ƒ
39
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
prioritisation of key variables
generation of Tornado Diagrams
ability to view sales report for any year based on cumulative probability
All the key drivers are assigned a range of variability to account for error margins:
the lowest, the highest and the mid-point from the primary market research.
These three elements will be used to simulate market scenarios using ‘Monte
Carlo Simulation’.
Key outputs from ‘Simulation’ will be sensitivity analysis and stretched ‘S’ curve of
cumulative total sales
Capability to analyse not only the historical market for each therapeutic area, but
also to predict the evolution of the market with accuracy by incorporating a
multitude of future market events
Agenda
About
About us
us
Introductions
Introductions
Quick
Quick ‘n’
‘n’ Dirty
Dirty for
for early
early phase
phase
Analogue
Analogue based
based
Monté
Monté Carlo
Carlo Simulation
Simulation
External
External expert
expert opinion
opinion
Fixed
Fixed product
product profile
profile
Variable
Variable product
product profile
profile
Tips,
Tips, sources,
sources, wrap-up
wrap-up
Copyright Lifescience Dynamics Ltd. 2007
40
The hypothesis underpinning
primary research for uptake
*
Standard errors
Studying a
small sample
Amplifying and
applying it
on the universe
*
*
0
50
100
Sample size
41
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
150
200
Overview of a typical forecasting project
with external expert opinions
Market
Survey:
Utility
Model:
Market
Model:
Market
Model:
Front End
Data,
Peak share
Presentation:
of
Insight
Putative profiles
Drivers
&
Diagnostics
Competitors,
gains & losses
42
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Data
Longitudinal
share
visualisation
of
&
Putative
profiles & key
competitors
Simulation
Tool
Choice of method(s) depends on
product and therapy area
considerations
Of which these are just a few…
New
product?
No
Adaptive conjoint
• Early stage products
Yes
No
Bass diffusion
based modelling
Fixed
product
profile?
Yes
43
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Statistical
modelling
Choice task
• Established therapy areas
• Later stage products
Juster Scale
• Cost effective
Process from primary data to
market model
Process flow
Discrete
Discrete choice
choice or
or ACA
ACA
output
output
Utilities
Utilities
Preference
Preference shares
shares from
from
Discrete
Discrete choice
choice or
or ACA
ACA model
model
Peak
Peak market
market share
share
Market
Market model
model
44
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Process overview
Primary Market Research to capture choices doctors will be making based on product profiles
via patient record forms`
Yes / No choices made by the doctors will be converted into utilities`
Utilities converted to produce preference shares for key indications
Brand preference is the sum of the utilities (the part-worth) for each component attribute level
Consumers have the greatest preference for products with the highest utility and should
choose the product with the highest utility most often
Preference shares by indication converted into patient share
Market share determined for real and/or hypothetical products among the respondents surveyed
Baseline market configuration established
Hypothetical product offering profiled
Rx are reconciled by re-weighting via IMS data to reflect off-label use in each market
Modelled with future events using Bass Diffusion model
Time to peak via IMS Analogue
Agenda
About
About us
us
Introductions
Introductions
Quick
Quick ‘n’
‘n’ Dirty
Dirty for
for early
early phase
phase
Analogue
Analogue based
based
Monté
Monté Carlo
Carlo Simulation
Simulation
External
External expert
expert opinion
opinion
Fixed
Fixed product
product profile
profile
Variable
Variable product
product profile
profile
Tips,
Tips, sources,
sources, wrap-up
wrap-up
Copyright Lifescience Dynamics Ltd. 2007
45
Establishing peak market uptake
via primary market research
We often use a patient-based approach, where physicians are asked to
complete patient record forms, detailing the characteristics and therapy of
the last 5 patients they have seen, and then predict their likelihood of
using the new product in these specific patients
•
Increases realism of responses
•
Avoids discussion of the ‘average patient’
•
Provides positioning in patient segments
•
Highlights where the new product is
perceived to fit in the therapeutic
armamentarium
A score of 8 + on the Juster scale has been shown to be a
reliable indicator of future behaviour
ƒ
ƒ
ƒ
ƒ
46
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
The Juster scale has been used to predict purchase
rates for a range of items, in different product classes
and in all cases, has proved to be a better predictor
than purchase intention scales*
The Juster Probability Scale
10
9
8
7
6
5
4
3
2
1
0
Certain, practically certain
Almost sure
Very probable
Probable
Good possibility
Fairly good possibility
Fair possibility
Some possibility
Slight possibility
Very slight possibility
No chance, almost no chance
* Marketing Bulletin, 1994, 5, 47-52
Agenda
About
About us
us
Introductions
Introductions
Quick
Quick ‘n’
‘n’ Dirty
Dirty for
for early
early phase
phase
Analogue
Analogue based
based
Monté
Monté Carlo
Carlo Simulation
Simulation
External
External expert
expert opinion
opinion
Fixed
Fixed product
product profile
profile
Variable
Variable product
product profile
profile
Tips,
Tips, sources,
sources, wrap-up
wrap-up
Copyright Lifescience Dynamics Ltd. 2007
47
Conjoint approach is used to if the
product profile is flexible
•
Conjoint analysis and Multidimension scaling (MDS) were developed for
measuring ‘human’ perceptions and preferences
ƒ
ƒ
ƒ
•
•
Well suited for measuring the human psychological judgements
Decomposition of a set of overall responses to factorially designed stimuli
ƒ
•
•
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Breaking down a decision of a purchase into many attributes and each attribute
with different level of performance
Conjoint analysis attempts to explain consumers’ overall choice/evaluation of
a marketing stimulus, in terms of the value of its constituent attributes
The Conjoint models are used to simulates consumer’s future buying
decision based on their current behaviour
ƒ
48
Developed by psychologists in the 60’s and refined and adapted by marketers ever
since
Widely used and validated techniques
Multivariate
E.g. physician's Rx decision
The hypothesis underpinning
Conjoint Analysis
Symptoms
History
Cough
Wheezing
Bronchospasm
Breathless on exertion
Tight chest
Tests
•X-Ray
•Peak flow
sis
gno thma
Dia
As
ate
der
Mo
Choices of drugs
A,B,C,D,E
reading etc.
Inhaled steroid
Combination
Short or Long Acting
Leukotriene Antagonists
[email protected]
•
•
Replicating choices made by
individual doctors to the universe
Choices made by today’s doctors
will not be that different in the
future, given the underlying
market assumptions remain the
same
Rationale:
Beta Agonist
Choices
Copyright Lifescience Dynamics Ltd. 2007
The theory
I should Rx Drug A
Inhaled Corticosteroids
49
ƒ
Is effective against the Cough
& Wheezing
Good safety profile
Competitively priced
Obtaining importance weightings:
Conjoint Models
ƒ There are two original types of true conjoint:
• Pair-wise
• Full Concept
ƒ and a series of hybrids...
ƒ Computer-based:
• ACA
• CBC
ƒ Paper-based:
• Scalar Conjoint
• CVA (Value analysis)
50
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Agenda
About
About us
us
Introductions
Introductions
Quick
Quick ‘n’
‘n’ Dirty
Dirty for
for early
early phase
phase
Analogue
Analogue based
based
Monté
Monté Carlo
Carlo Simulation
Simulation
External
External expert
expert opinion
opinion
Fixed
Fixed product
product profile
profile
Variable
Variable product
product profile
profile
Tips,
Tips, sources,
sources, wrap-up
wrap-up
Copyright Lifescience Dynamics Ltd. 2007
51
Historical Data
ƒ
ƒ
Two well known & standard sources - IMS and Scott-Levin.
Types of Data that they provide :
•
•
•
Volume (IMS / Scott-Levin)
IMS’s NSP, NPA (US)
IMS’s MIDAS (Global)
ƒ
Total Sales (DOL TOT)
– Currency sales
– Grams, kilo or IU, SU
ƒ
ƒ
•
Allocation (IMS-NDTI, Varispan, Scott-Levin - VONA)
ƒ
ƒ
•
ƒ
ƒ
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
By year, Quarter, Month, or Week (For long term forecasts – Yearly is preferred)
By Country (U.S. data is best, other countries have variable quality)
Alternative include:
ƒ
52
Uses by ICD-10 Code
Concomitant Uses (i.e. using with other medications)
Availability
ƒ
•
Total Scripts (TRx) - Scripts written
Average Daily Use (ADU / DaCON)
SEC Filings
News archives
Historical Data (cont.)
ƒ End goal is converge historical sales data and the relevant target
patient populations
ƒ This entails converting “Sales Volume Data” into patient #s
• 1st Step – To agree on Patient-Day of Therapy (DOT)
= # of tablets/ml / patient / day
• DOT = EUTRx / DaCon (or ADU) or
= TRx x Average Days Per Rx.
• Assumption on a Compliance* Level
ƒ
Compliance varies from <50% for less critical drugs to >90% for life saving
medicines say insulin
• Patient-Year Equivalents = DOT / (DOT/Yr x Compliance).
Compliance Compliance (or Adherence) refers to a patient both agreeing to and then undergoing
some part of their treatment program as advised by their doctor. Most commonly it is whether a
patient takes their medication e.g., if a 30-day script is completed by a patient in 60 days, his /
her compliance is 50%.
53
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Epidemiology data sources
ƒ
Syndicated reports
•
•
ƒ
General Sources
•
•
•
ƒ
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
ƒ
ƒ
ƒ
Therapy monitors
Syndicated Patient Diaries
ƒ
ƒ
Current Treatments – Ad-hoc
•
•
54
USRDS, OMIM
ƒ
ƒ
Current Treatments - Syndicated
•
•
ƒ
PubMed
Lancet, NEJM
Medscape
Disease specific sites:
•
ƒ
Decision Resources
Datamonitor
ƒ
Primary market research
Custom Patient Diaries
ƒ
CDC - Center for Disease
control - USA
Italian Statistics Institute
Krebsregister Saarland Germany
Ministry of Health, Labour and
Welfare - UK
National Cancer Institute
National Statistics http://www.statistics.gov.uk/
Robert Koch Institut
UNAIDS - United Nations
Programme of HIV/AIDS
World Health Organisation
Forecasting data sources
ƒ
Level of unmet needs,
concomitant diagnoses, their
trends, concomitant treatment
or procedures
•
•
ƒ
Future competition
•
•
55
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Primary market research
Custom Patient Diaries
Premier healthcare informatics
Primary market research
IDdB, PharmaProject, R&D
Focus
Ad-hoc competitive intelligence
Marketed treatment regimens
•
•
•
•
•
•
Hospitalisation
•
•
ƒ
ƒ
ƒ
Physicians Desk Reference (US)
British National Formulary (UK)
Rote Liste (DE)
Hipocrates’ Vademecum (ES)
Vidal (FR)
L’Informatore Farmaceutico (IT)
Price
•
•
•
•
•
•
Redbook (US)
MIMS (UK)
Rote Liste (DE)
Hipocrates’ Vademecum (ES)
Vidal (FR)
L’Informatore Farmaceutico (IT)
Forecasting do’s and don’ts
ƒ
Challenge key assumptions
•
•
•
•
Price
Duration of therapy
Compliance
Uptake / penetration rates
ƒ
ƒ
ƒ
ƒ
ƒ
ƒ
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Historical / past trends
Uptake curve
Years to peak sales
Current & future competition
Generics
Specific epidemiology
Order of launch
Price
Specific drug use
Therapeutic area
•
•
•
•
ƒ
ƒ
Risks (greater in early phases)
General spreadsheet related errors
•
•
•
•
ƒ
Acute
Chronic
Cycles of therapies for IVs, Subcut
Diagnostic / Device
Stage of Development
•
Country
•
•
•
•
56
ƒ
Data entry errors i.e. typo etc.
Logic
Errors in linking of cells etc.
Calculations / maths etc.
Sanity check the data & outputs
•
•
•
Back of envelope calculations
Broker's reports
Similar past / current products
Market share techniques based
on secondary data
Techniques
Order of entry models
Analogues
Internal PM opinions
Monte Carlo
Combination of internal expert
Opinion with Monte Carlo
57
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Benefits &
Challenges
9 Quick, easy to use and cost effective > Very good for BD&L opportunities
8 Drugs are not same in efficacy and safety (even if small)
8 Based on past, uses past as the future predictor and ignore market events
9 Quick, easy to use and cost effective > Very good for BD&L opportunities
8 Very difficult to find a good match
8 Based on past, uses past as the future predictor and ignore market events
9 Quick, easy to use and cost effective > Very good for BD&L opportunities
8 Biased and very difficult to defend
8 Based on past, uses past as the future predictor and ignore market events
9 Quick, easy to use and cost effective > Very good for BD&L opportunities
8 Lower and higher limits are based on biased view
8 Very difficult to explain the underlying maths and algorithms
9 Increase robustness and cost effective > Very good for BD&L opportunities
9 Lower and higher limits are based on expert view and bias taken out
8 Very difficult to explain the underlying maths and algorithms
Market share techniques based
on direct estimation
Techniques
Benefits &
Challenges
Quantitative direct
estimation
9 1 to 5 years before market launch
9 Paradigm shifts can be studied and anticipated
8 Expensive and time consuming and quality depends on market researchers
9 Build consensus, obtain group thought
8 Dominated by one or two doctors and rest gravitate to their numbers
8 Relative small sample size
9 Helps to understand where the new product will fit Rx Algorithms
8 Inability to recall & review a range of patients, accurately summarising
8 Responses tend to cluster around certain numbers (10%, 20%, and 25%)
9 Robust numbers and defendable
8 Inability to recall & review a range of patients, accurately summarising
8 Responses tend to cluster around certain numbers (10%, 20%, and 25%)
Self-completion
forms
9 Robust numbers and defendable
8 Inability to recall & review a range of patients, accurately summarising
8 Responses tend to cluster around certain numbers (10%, 20%, and 25%)
Expert Interview/KOL
Focus Groups
Qualitative
doctor interviews
Errors arising from physicians inability to rapidly
recall and review a range of patients and then
accurately summarising the results
58
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Market share techniques based
on in-direct estimation
Techniques
Conjoint
Juster Scale
Conjoint
with patient diary
Juster Scale
with patient diary
Delphi
techniques
59
Copyright Lifescience Dynamics Ltd. 2007
[email protected]
Benefits &
Challenges
9 Defendable, well established
9 Incorporates product features and claims
8 Expensive and time consuming and quality depends on market researchers
9Defendable, well established
9 Incorporates product features and claims
8 Less time consuming and quality depends on market researchers
9 Defendable, well established
9 Most closest to real prescribing decisions
8 Responses tend to be higher share than the reality and P&R element missing
9 Defendable, well established
9Most closest to real prescribing decisions
9 Responses tend to be higher share than the reality and P&R element missing
9 Easy and cost effective
8 Inability to recall & review a range of patients, accurately summarising
Lifescience Dynamics (Worldwide Headquarters)
The Oriel – Thames Valley Court
185 Bath Road
Slough
Berkshire SL1 4AA
England, UK
O: +44 (0) 1753 205 126
F: +44 (0) 1753 205 127
Lifescience Dynamics (USA)
304 Park Avenue South, 11th Floor
New York, NY 10010
USA
O: +1 (212) 926-9290
F: + 1 (347) 523-9639
[email protected]
www.lifesciencedynamics.com
Thank you
Questions?
Copyright Lifescience Dynamics Ltd. 2007
60