SLIDE 1
Adaptive discrete choice designs Presentation held at - - PowerPoint PPT Presentation
Adaptive discrete choice designs Presentation held at - - PowerPoint PPT Presentation
Adaptive discrete choice designs Presentation held at 2008 Boris Vaillant www.quantitative-consulting.eu Product attributes The goal of discrete choice analysis in marketing is to assess the influence of
SLIDE 2
SLIDE 3
www.quantitative-consulting.eu 3
Product comparisons In product comparison questionnaires, respondents indicate their preferred choice or purchase pattern in a series of ‘choice tasks’
Typical example: Doctors are asked to indicate the prescription share for each of the shown treatments (e.g. for the last 10 patients with the corresponding indication)
SLIDE 4
www.quantitative-consulting.eu 4
The preference of individual customers or customer groups for the different product elements is parametrised and can be used to calculate the preference for existing or new products
C a r e x a m p l e : S e d a n s e g m e n t
Utility parameters
MB BMW Audi
Brand
20 10
Utility model 1 Utility model 1
- 10
10 20 30 40 50 60 Total Tiptronic Medium 7 L 130 HP 34 KEUR Audi 51 7 4 18 12 10 10 L 9 L 8 L 7 L
Consumption
7 14 18 42 KEUR 38 KEUR 34 KEUR 30 KEUR
Price
4 12 24 High Medium Base
Equipment
10 4 Tiptronic Manual
Gearshift
7 190 HP 170 HP 150 HP 130 HP
HP
7 6 3
Utility model 2 Utility model 2
- 10
10 20 30 40 50 60 Total Manual Base 9 L 150 HP 42 KEUR MB 30 7 3 20
SLIDE 5
www.quantitative-consulting.eu 5
In the marketing practice, discrete choice data often feeds larger market models in which companies can test the outcome of different strategies
Customer preferences Demand curve and profit-optimal price Product information
- Target segments
- Description of existing and
future products Price information
- Most important
products in the market
- Changes in time
Market information
- Market size
- Market shares
Market simulation Simulation of customers choice in scenarios Price changes New products Competitive reaction
- Customer preferences as
measured via the discrete choice questionnaire
- Either individual or group
preferences
- Group-wise results or
aggregation of individual results
- Obtain market effects e.g.
for a price change
- Determine optimal new
price
Additional information from the survey
- Price barriers
- Brand value
- Strengths and weaknesses
Market simulation
SLIDE 6
www.quantitative-consulting.eu 6
Marketing applications: Example of a simulation tool
SLIDE 7
www.quantitative-consulting.eu 7
- Substitution effects strongly influence the optimal price decision
- Complexity of substitution decisions is best captured by models based
- n individual choice data
One practical challenge is the estimation of substitution (or portfolio-) effects. This often requires that the parameters of the model be estimated for each individual customer Marketing applications: Substitution effects
50 100 150 200 50 100 150 Demand including substitutes Price index Demand index Sales new product alone Effects from substitute A Effects from substitute B Effects from substitute C 50 100 150 200 60 80 100 120 Profit including substitutes Price index Profit index Profit new product alone Effects from substitute A Effects from substitute B Effects from substitute C Profit effect portfolio
SLIDE 8
www.quantitative-consulting.eu 8
Creating optimal bundles is another area which requires knowledge of the individual preferences
- Testing of bundling scenarios leads to
the best bundles to sell and their optimal price
- Individual preferences and WTP for the
different products are estimated in a discrete choice study
- From this, one can estimate which
products customers will purchase and in which combination
Marketing applications: Bundling
Demand for equipment X and Y WTP equipment X WTP equipment Y Effect of bundling Price of X Price of Y Price of bundle (X+Y) Customer buys more Customer buys less No change 2000 4000 6000 8000 500 1000 1500 Profit of equipment bundle Price of bundle Profit of bundle Profit equipment X Profit equipment Y Price X + Price Y Price of bundle (X+Y)
SLIDE 9
www.quantitative-consulting.eu 9
Goals of optimal discrete choice design The goal of discrete choice design is to minimise the error of parameter estimates while keeping respondent burden at a minimum
Respondents
- Reduce task complexity
- Reduce number of choice
tasks
- Avoid respondent fatigue
Consultant
Obtain reliable and valid results Have a variety of design options
Study sponsor
Reduce cost and
- btain reliable and
valid results
SLIDE 10
www.quantitative-consulting.eu 10
Discrete choice modelling
SLIDE 11
www.quantitative-consulting.eu 11
Classical design criteria (1)
SLIDE 12
www.quantitative-consulting.eu 12
Classical design criteria (2)
SLIDE 13
www.quantitative-consulting.eu 13
Bayesian adaptive design (1)
SLIDE 14
www.quantitative-consulting.eu 14
Bayesian adaptive design (2)
H(u) H(u|Z) H(Z) H(Z|u) I(Z,u)
SLIDE 15
www.quantitative-consulting.eu 15
Sequential Monte Carlo (1)
SLIDE 16
www.quantitative-consulting.eu 16
Sequential Monte Carlo (2)
(U1) Start (U2) Reweight (U3) Resample (U4) Move (U2) Reweight
...
SLIDE 17
www.quantitative-consulting.eu 17
R-implementation We created a test-suite in R to compare the different design strategies in simulations. Our tests are based on a typical scenario from pharmaceutical market research
This is a typical scenario from pharmaceutical market research. Here, doctors are asked to estimate the prescription share of each of the presented treatments. Compared to simple choice questions, this provides sufficient information for individual-level estimation
SLIDE 18
www.quantitative-consulting.eu 18
Results (1)
B1 Det B2 Max B3 Ent B5 Simple Fixed Random 0.0 0.1 0.2 0.3 0.4 0.5 RMSE for Utilities
M=0.248 P=2e-04 M=0.254 P=0.002 M=0.253 P=0.001 M=0.269 P=0.2 M=0.282 P=NA M=0.32 P=3e-05
5 10 15 20 0.3 0.4 0.5 0.6 0.7 0.8 RMSE vs. # questions # of questions RMSE B1 Det B2 Max B3 Ent B5 Simple Fixed Random
SLIDE 19
www.quantitative-consulting.eu 19
Results (2)
Det Max Ent Sim Fix Rnd 0.10 0.15 0.20 0.25 0.30 0.35 0.40 ML D-criterion: Avg SE
- 2
- 1
1 2
- 3
- 2
- 1
1 2 3 Bias of estimates 'True' values Estimated values Bias factors B1 Det : 1.108 B2 Max : 1.088 B3 Ent : 1.102 B5 Simple : 1.102 Fixed : 1.067 Random : 1.016
SLIDE 20
www.quantitative-consulting.eu 20
Performance
1000 5000 10000 20000 40000 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 RMSE vs. MC sample size and processing time MC sample size RMSE
Update=0.45 s Select=0.51 s Update=2.5 s Select=2.7 s Update=5.2 s Select=5.8 s Update=11 s Select=13 s Update=22 s Select=24 s M=0.278 P=NA M=0.248 P=0.002 M=0.263 P=0.1 M=0.241 P=1e-04 M=0.242 P=2e-04
SLIDE 21
www.quantitative-consulting.eu 21
References (1)
SLIDE 22