Two New Features in Discrete Choice Experiments to Improve - - PowerPoint PPT Presentation

Two New Features in Discrete Choice Experiments to Improve Willingness to Pay Estimation that Result in SDR and SADR: Separated (Adaptive) Dual Response

Forthcoming in Management Science Christian Schlereth

WHU – Otto Beisheim School of Management

Bernd Skiera

Goethe University Frankfurt

Carson et al. (1994); Dhar (1997); Louviere et al. (2000); Haaijer et al. (2001); Vermeulen et al. (2008)

Choice-based conjoint nowadays one of the most important method to measure willingness to pay

A B C

Do not purchase any of the three Attribute 1 Attribute 2 Attribute 3 …

Choice-based conjoint

• Free-choice questions only, i.e., each choice-sets contains a no-purchase option
• No-purchase option provides:
• Clear reference point
• Realistic experimental setting
• Allows prediction of market penetration

2

Information gained when choosing a product

A B C

Do not purchase any of the three Attribute 1 Attribute 2 Attribute 3 …

Purchase decision Selection decision B > 0 Product B provides sufficient utility for a purchase B > A; C

Choice-based conjoint

3

Information gained when choosing no-purchase option

A B C

Do not purchase any of the three Attribute 1 Attribute 2 Attribute 3 …

Not enough data to learn about individual preferences Purchase decision Selection decision 0 > A; B; C None of the products provide sufficient utility for a purchase No information about relative attractiveness

• f attributes

Choice-based conjoint

4

Dual response: Selection decisions are also observed when no- purchase option is chosen

• Dual Response

A B C

Attribute 1 Attribute 2 Attribute 3 … Purchase most preferred Do not purchase most preferred Forced choice question Free choice question

Purchase decision Selection decision 0 > A; B; C From free choice question From forced choice question  Selection decision is always observable; thus: more accurate estimation of preferences  But higher cognitive effort for a respondent due to double amount of questions B > A; C

Dhar & Simonson (2003); Dhar & Nowlis (2004); Brazell et al. (2006)

Dual response

5

Shortcoming: Context effects in choice-based conjoint

Examples:

• Attraction Effect:

 No-purchase option is chosen less frequently, if a dominant product alternative exists

• Similarity Effect:

 No-purchase option is chosen more frequently, if similar attractive product alternatives exists, as an “easy way out”

Huber, Payne, and Puto (1982); Tversky and Shafir (1992); Dhar (1997); Rooderkerk, Van Heerde, and Bijmolt (2011);

A B C

Do not purchase any of the three Attribute 1 Attribute 2 Attribute 3 …

6

• Whether products provide sufficient utility for a purchase is not the only reason for a respondent

to pick the no-purchase option

• Context effects typically neglected in estimation

Shortcoming: Context effects in dual response

Other empirical findings

• Higher share of chosen no-purchase option (Dhar and Simonson 2003; Dhar and Nowlis 2004; Brazell et al.

2006)

• Artificial time delay between selection decisions and purchase decisions reduces no-purchase share

(Dhar and Simonson 2003) A B C

Attribute 1 Attribute 2 Attribute 3 … Purchase most preferred Do not purchase most preferred

7

• Context effects also exist for dual response
• As a result, willingness to pay estimates are substantially lower compared to choice-based conjoint

Shortcoming: Extreme response behavior

Extreme response behavior (Gensler et al. 2012)

• Respondent always chooses no-purchase option
• No information when respondent will start buying
• WTP might be estimated too low
• Respondent never chooses no-purchase option
• No information, when respondent will stop buying
• WTP might be estimated too high

Extreme response behavior in previous studies (if reported) :

Choice-Based Conjoint

• 58% in Gensler et al. (2012)
• 64% in Parker and Schrift (2011)
• 22% in Wlömert and Eggers (2014)
• Up to 56% in our studies

Dual Response

• 31% in Wlömert and Eggers (2014)
• Up to 36% in our studies

8

Shortcoming: Impact of purchase probability on measurement accuracy

9

Implication: Companies estimate willingness to pay more accurately for a respondent who does not intend to buy their product If a choice-set contains more than one alternative, likelihood increases that a respondent compares a rather attractive alternative against the no-purchase option  purchase decisions are less informative for a respondent with high purchase probability

Aims of paper

• Development of SDR: “Separated Dual Response“, which

1. Avoids context effects by imposing a strict separation between all forced and free choice questions

• Development of SADR: “Separated (Adaptive) Dual Response“, which also

2. Avoids extreme response behavior by imposing a strict separation between all forced and free choice questions through an adaptive mechanism that captures heterogeneity in willingness to pay 3. Ensures similar accuracy in measured willingness to pay, independent of a respondent‘s purchase probability

10

Agenda

11

 Mechanism of SDR and SADR  Simulation study to analyze dependence between willingness to pay accuracy and purchase probability  Tests of statistical and behavioral endogeneity  Insights from three empirical studies

SADR (Separated Adaptive Dual Response): In addition to feature 1 („strict separation“), we use decisions in forced choice questions to adaptively identify fewer, but more informative free choice questions

Two new features for discrete choice experiments resulting in SDR and SADR

12

Feature 1: Strictly separating forced and free choice questions Feature 2: Adaptive mechanism to select fewer, but more informative, free choice questions SDR (Separated Dual Response): We ask all forced choice questions first and then all free choice questions. Thus, we introduce a time delay between a forced and a free choice question

Separated Dual Response (SDR) ... A1 B1 C1

Buy Selected1 Do not buy

A2 B2 C2

Buy Selected2 Do not buy

... Separated Adaptive Dual Response (SADR) ... A1 B1 C1

Buy A 1 Do not buy

A2 B2 C2

Buy A 2 Do not buy

...

Adaptive mechanism of SADR – Separated Adaptive Dual Response

13

Information gap

1. Forced choice block

 Use efficient choice design for all respondents (e.g. D-optimal)  Use linear probability model to approximate individual preference order

(Heckman & Snyder 1997)

 Simulate preference order of all products in full factorial design

2. Free choice block

 "Smartly“ select n products for inclusion in purchase questions  Use binary logit model to select next area, which provides most information about purchase decision making  Iterate m times

Select A, Select B, Select C Select A, Select B, Select C …

SADR

Purchase Presented, Purchase None …

Estimate preliminary preference order

Most preferred product Least preferred product

75% 50% 25%

Preference order

0% (= No Purchase)

Probability of a purchase

100% (= Purchase)

A respondent‘s perspective - screenshots of SADR

Forced Choice Block:

(“pick one of the products“)

j forced choice questions A1 B1 C1 A2 B2 C2 ... A3 B3 C3 A4 B4 C4 A5 B5 C5

Free Choice Block:

(“buy or not buy product“)

Block 1 of n free choice questions Buy D1' Do not buy D1' Buy D2' Do not buy D2' Buy ... Do not buy ... Buy ... Do not buy ... ... Block 2 of n free choice questions

14

Summary of studied discrete choice experiments

15

Choice-Based Conjoint (CBC) Dual Response (DR) Separated Dual Response (SDR) Separated Adaptive Dual Response (SADR) A1 B1 C1

Do not buy

A1 B1 C1 ...

Buy Selected1 Do not buy

A2 B2 C2

Do not buy

A2 B2 C2

Buy Selected2 Do not buy

... A1 B1 C1

Buy Selected1 Do not buy

A2 B2 C2

Buy Selected2 Do not buy

... ... A1 B1 C1

Buy A 1 Do not buy

A2 B2 C2

Buy A 2 Do not buy

... Mechanism Shortcomings Context effects Extreme response behavior Impact of purchase probability

• n accuracy of WTP

X X X X X X X X

Estimation

Scale-extended model  DR-2Max-model (Diener, Orme, and Yardley 2006)  Extended to account for differences in consistency between selection and purchase decisions (Swait and Andrews 2003) Force choice questions Free choice questions Estimation  Multinomial logit model using Hierarchical Bayes  All models implemented in Matlab

A B C

Attribute 1 Attribute 2 Attribute 3 …

Buy product D Do not buy product D

A B C

Attribute 1 Attribute 2 Attribute 3 …

A B C

Attribute 1 Attribute 2 Attribute 3 …

Buy product D Do not buy product D Buy product D Do not buy product D

     

 

 

h,i,j h,i ',j' j j'

d d 1 h,i 2 h,i' h j J i C j' J' i' C 2 2 h,i' 1 h,j J

exp V exp V L exp V exp V exp V     

   

                         

  

16

Agenda

17

 Mechanism of SDR and SADR  Simulation study to analyze dependence between willingness to pay accuracy and purchase probability  Tests of statistical and behavioral endogeneity  Insights from three empirical studies

Monte carlo simulation study

18

Set-up based on: 18 choice sets with 4 attributes and 4 levels each Setup similar to:

• Aurora and Huber (2001)
• Toubia et al. (2004)

Experimental Conditions Number of Levels Values Types of Discrete Choice Experiments 7  Choice-Based Conjoint ( 0 separate free choice questions)  Dual response, SDR (18 separate free choice questions)  SADR [m=1, n=9] ( 9 separate free choice questions)  SADR [m=9, n=1] ( 9 separate free choice questions)  SADR [m=3, n=3] ( 9 separate free choice questions)  SADR [m=2, n=2] ( 4 separate free choice questions)  SADR [m=4, n=4] (16 separate free choice questions) 4  4  5  6  7 2   = .5 (low accuracy)   = 3 (high accuracy) 2  σ² = .5 



(low heterogeneity)  σ² = 3 



(high heterogeneity) 3  γ = .6 (low: ~10% no-purchase decisions)  γ = -.8 (medium: ~30% no-purchase decisions)  γ = -1.75 (high: ~50% no-purchase decisions) Number of conditions 4 ∙ 2 ∙ 2 ∙ 3 = 48 Number of types of discrete choice experiments 7 Number of replications 5 Total number of studies 48 ∙ 7 ∙5 = 1,680 Notes: 100 respondents; SADR = separated adaptive dual response; m = number of iterations in free choice block of SADR, each of which consists of n free choice questions.

Comparison of ability to recover constant in utility function

19

Purchase Probability Share of No- Purchases Choice- Based Conjoint Dual Response & SDR SADR [m=1,n=9] [m=9,n=1] [m=3,n=3] [m=2,n=2] [m=4,n=4] LOW HIGH .56 .58 .71 .70 .70 .83 .61 MIDDLE MIDDLE .63 .68 .73 .71 .72 .85 .62 HIGH LOW .82 .89 .74 .72 .73 .86 .63 Mean .67 .72 .73 .71 .72 .85 .62

Notes: RMSE = root mean squared error; lower values indicate better ability. SDR = separated dual response; SADR= separated adaptive dual response; m = number of iterations in free choice block, each of which consists of n free choice questions.

• Ability to recover constant in utility function (RMSE)
• serves to predict the error of the probability that a respondent will buy a product or not
• varies with purchase probability for choice-based conjoint and dual response, but not for

SADR

Agenda

20

 Mechanism of SDR and SADR  Simulation study to analyze dependence between willingness to pay accuracy and purchase probability  Tests of statistical and behavioral endogeneity  Insights from three empirical studies

Examination of endogeneity

Violation of assumption of independence of choices, because design of free choice questions depend upon previous choices and therefore on realizations of error term

Hauser and Toubia (2005); Liu, Otter, and Allenby (2007)

Statistical Perspective Behavioral Perspective Adaptive nature of free choice questions might affects a respondent’s choices and cause anchoring or framing effects

DeShazo (2002); Alberini, Kanninen, and Carson (1997); Hanemann, Loomis, and Kanninen (1991)

Xforced choices β, σ² Forced Choice Block Free Choice Block Yforced choices Xfree choices Yfree choices

Examined using approach of

Alberini, Kanninen, and Carson (1997)

Examined using approach of

Liu, Otter, and Allenby‘s (2007)

21

500 1000 1500 2000 2500 3000 3500 4000 4500 5000

• 2.6
• 2.4
• 2.2

500 1000 1500 2000 2500 3000 3500 4000 4500 5000

• 2.6
• 2.4
• 2.2

SADR with adaptive design SADR with fixed design

MCMC iteration MCMC iteration Estimate of constant Estimate of constant

Statistical concerns of endogeneity can be ignored

• Mechanism of creating Xfree choice can be ignored if (Liu, Otter, and Allenby‘2007):
• estimation method adheres to likelihood principle
• Xfree choice does not contain information that is beyond Yforced choice and Yfree choice

Demonstration

• Comparison of recovery accuracy of SADR with

adaptive free choice question design and a fixed (D-optimal) design

• 18 Choice-Sets, 44-design, 16 free choice

questions

• Sampled parameter values of 1000 consumers

drawn from normal distribution

• Mean (StdDev) of constant: -2.40 (2.45)
• 2.40 (1.98)
• 2.43 (1.93)

22

Behavioral concerns of endogeneity can be ignored

Indications of behavioral concerns of endogeneity, studied in the double bounded dichotomous choice literature

• Downward shift in WTP through follow-up free choice questions (e.g., from $250 to $150)
• See further Alberini, Kanninen, and Carson (1997), Hanemann, Loomis, and Kanninen (1991), and McFadden and Leonard (1995)

Recommendations to avoid behavioral endogeneity

 Well-balanced, symmetric designs result in very modest bias, even if anchoring is strong (Veronesi, Alberini, and Cooper

2011)

 Testing for structural shifts in WTP by estimating additional term δ·ln(pos∙) in utility function (Alberini, Kanninen, and Carson

1997)

Applying Alberini, Kanninen, and Carson’s (1997) test in our empirical studies detects no structural shift in WTP for later free choice questions

23

Agenda

24

 Mechanism of SDR and SADR  Simulation study to analyze dependence between willingness to pay accuracy and purchase probability  Tests of statistical and behavioral endogeneity  Insights from three empirical studies

Description of three empirical studies to compare (SDR and) SADR against choice-based conjoint and dual response

Study 2: Basketball tickets

• N = 880 (customers of market leader)
• 52·4·3·2 Balanced Design

Study 1: Tablets

• N = 459 (fans of a major league

basketball team)

• 43 Balanced Design

Choice- Based Conjoint Dual Response SADR Force Choice Questions

• 18

18 Free Choice Questions 18 18 9

Study 3: Video-on-demand

Choice- Based Conjoint Dual Response SADR Force Choice Questions

• 12

12 Free Choice Questions 12 12 6 Choice- Based Conjoint Dual Response SDR SADR Force Choice Questions

• 12

12 12 Free Choice Questions 12 12 12 6

• N = 1,425
• 4·32·23 Balanced Design

25

Set-up of questionnaire

Choice-Based Conjoint Dual Response SADR Introduction Measurement of cognitive effort

(Bettman et al. 1986)

Configurator-task and direct questions to measure willingness to pay Collection of demographic and socio-economic variables

Random assignment to one discrete choice experiment 100% 100% 100%

SDR (study 1 only)

26

Systematic differences of selecting no-purchase option

Choice-based conjoint Dual response SDR SADR Study 1: Tablets N=214 N=203 N=219 N=208 Share of no-purchase option 22.7% 46.2% 42.2% n.a. Extreme response behavior 1: Never no-purchase option 32.7% 9.4% 3.7% .5% Extreme response behavior 2: Always no-purchase option 3.3% 12.8% 3.7% 5.3% No extreme response behavior 64.0% 77.8% 92.7% 94.2% Study 2: Basketball tickets N=160 N=146

• N=153

Share of no-purchase option 15.8% 29.4%

• n.a.

Extreme response behavior 1: Never no-purchase option 56.3% 25.0%

• 9.5%

Extreme response behavior 2: Always no-purchase option 0.0% 0.7%

• 3.8%

No extreme response behavior 43.8% 74.3%

• 86.7%

Study 3: Video-on-Demand N=267 N=308

• N=305

Share of no-purchase option 48.4% 67.2%

• n.a.

Extreme response behavior 1: Never no-purchase option 16.9% 5.8%

• 1.0%

Extreme response behavior 2: Always no-purchase option 17.2% 30.8%

• 6.9%

No extreme response behavior 65.9% 63.3%

• 92.2%

27

• SADR with lowest share of extreme response behavior; Separation via SDR contributes mostly to this result
• Choice-based conjoint with highest share of respondents who would always purchase
• Dual response with highest share of no-purchase option

Summary of study results

Studie 1 Tablets Studie 2 Basketball tickets Studie 3 Video-on-demand Lowest share of extreme response behavior SADR SADR SADR Validity Best internal validity (hit rate, RMSE) SADR SADR SADR Best predictive validity (hit rate, RMSE) SADR, SDR SADR SADR Best convergent validity (RMSE) n.a. SADR SADR Best external validity (RMSE) n.a. SADR SADR Cognitive effort Lowest perceived difficulty No significant differences No significant differences CBC, SADR Lowest dropout rate in discrete choice experiment CBC (SADR lower dropout rates than dual response) CBC (SADR lower dropout rates than dual response) CBC (SADR lower dropout rates than dual response) Lowest duration CBC (SADR with about equal duration than dual response) CBC (SADR with about equal duration than dual response) CBC (SADR with about equal duration than dual response)

28

Deep dive on convergent and external validity for study 2 & 3

Study 2: Basketball tickets

• N = 880 (customers of market leader)
• 52·4·3·2 Balanced Design

Study 1: Tablets

• N = 459 (Fans of a major league

basketball team)

• 43 Balanced Design

Choice- Based Conjoint Dual Response SADR Force Choice Questions

• 18

18 Free Choice Questions 18 18 9

Study 3: Video-on-demand

Choice- Based Conjoint Dual Response SADR Force Choice Questions

• 12

12 Free Choice Questions 12 12 6 Choice- Based Conjoint Dual Response SDR SADR Force Choice Questions

• 12

12 12 Free Choice Questions 12 12 12 6

• N = 1,425
• 4·32·23 Balanced Design

29

Convergent validity – choice-based conjoint

Choice-Based Conjoint Dual Response SFFC Study 1: Video-on-Demand Aggregate-level comparison Mean WTP self-stated ↔ CBC 10.90 € ↔ 11.61 € self-stated ↔ DR 10.30 € ↔ 8.29 € self-stated ↔ SFFC 11.40 € ↔ 11.85 € Median WTP 10.00 € ↔ 7.82 € 10.00 € ↔ 6.82 € 10.00 € ↔ 10.40 € Max WTP 30.00 € ↔ 167.39 € 30.00 € ↔ 58.16 € 40.00 € ↔ 50.84 € Study 2: Basketball Tickets Aggregate-level comparison self-stated ↔ CBC self-stated ↔ DR self-stated ↔ SFFC Mean WTP 24,30 € ↔ 54,58 € 23,38 € ↔ 24,78 € 25,30 € ↔ 25,52 € Median WTP 23,00 € ↔ 28,94 € 24,00 € ↔ 20,66 € 25,00 € ↔ 23,25 € Max WTP 60,00 € ↔ 312,41 € 60,00 € ↔ 94,47 € 70,00 € ↔ 72,15 €

A B C

No- purchase

• ption

Attribute 1 Attribute 2 Attribute 3 …

30

Convergent validity – dual response

Choice-Based Conjoint Dual Response SFFC Study 1: Video-on-Demand Aggregate-level comparison Mean WTP self-stated ↔ CBC 10.90 € ↔ 11.61 € self-stated ↔ DR 10.30 € ↔ 8.29 € self-stated ↔ SFFC 11.40 € ↔ 11.85 € Median WTP 10.00 € ↔ 7.82 € 10.00 € ↔ 6.82 € 10.00 € ↔ 10.40 € Max WTP 30.00 € ↔ 167.39 € 30.00 € ↔ 58.16 € 40.00 € ↔ 50.84 € Study 2: Basketball Tickets Aggregate-level comparison self-stated ↔ CBC self-stated ↔ DR self-stated ↔ SFFC Mean WTP 24,30 € ↔ 54,58 € 23,38 € ↔ 24,78 € 25,30 € ↔ 25,52 € Median WTP 23,00 € ↔ 28,94 € 24,00 € ↔ 20,66 € 25,00 € ↔ 23,25 € Max WTP 60,00 € ↔ 312,41 € 60,00 € ↔ 94,47 € 70,00 € ↔ 72,15 €

A B C

Attribute 1 Attribute 2 Attribute 3 … Purchase most preferred Do not purchase most preferred

31

Attribute 1 Attribute 2 Attribute 3 … Attribute 1 Attribute 2 Attribute 3 … Attribute 1 Attribute 2 Attribute 3 … Purchase product D Purchase product E No purchase No purchase

Convergent validity – SADR

Choice-Based Conjoint Dual Response SFFC Study 1: Video-on-Demand Aggregate-level comparison Mean WTP self-stated ↔ CBC 10.90 € ↔ 11.61 € self-stated ↔ DR 10.30 € ↔ 8.29 € self-stated ↔ SFFC 11.40 € ↔ 11.85 € Median WTP 10.00 € ↔ 7.82 € 10.00 € ↔ 6.82 € 10.00 € ↔ 10.40 € Max WTP 30.00 € ↔ 167.39 € 30.00 € ↔ 58.16 € 40.00 € ↔ 50.84 € Study 2: Basketball Tickets Aggregate-level comparison self-stated ↔ CBC self-stated ↔ DR self-stated ↔ SFFC Mean WTP 24,30 € ↔ 54,58 € 23,38 € ↔ 24,78 € 25,30 € ↔ 25,52 € Median WTP 23,00 € ↔ 28,94 € 24,00 € ↔ 20,66 € 25,00 € ↔ 23,25 € Max WTP 60,00 € ↔ 312,41 € 60,00 € ↔ 94,47 € 70,00 € ↔ 72,15 €

32

Comparison of demand functions against the one derived from self- stated willingness to pay

33

Study 2: Basketball Tickets Study 3: Video-on-Demand

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 5 € 7 € 9 € 11 € 13 € 15 € 17 € 19 € 21 € Share of customers purchasing at price p Price for the self-customized VoD plan

Choice-Based Conjoint Dual Response SADR Directly stated

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 9 € 14 € 19 € 24 € 29 € Share of customers purchasing at price p Price for the self-customized basketball ticket Choice-Based Conjoint Dual Response SADR Directly Stated

External validity Study 2 – basketball tickets

Approach

• Comparison 1: Comparison of actual and predicted choice proportions in four ticket categories

in season before and after study

• Comparison 2: Comparison of actual and predicted number of viewers after price increase of 2 €

between seasons NOTE: Price categories were sold out in only 8.09 % of all games

CBC Dual Response SADR

Choice proportions in four price categories

N=160 N=146 N=153 RMSE season before study .099 .095 .059 RMSE season after study .102 .098 .061 Change in number of viewers after price increase RMSE .041 .038 .034

34

External validity Study 3 – video-on-demand

“with reference to external validity, (…) one should measure the attributes of real choice alternatives (e.g., real brands) faced by each subject and observe their reported (…) most recent choice” (Batsell and Louviere

1991)

 Comparison of each subject’s self-reported answer, whether they previously purchased VoDs and how much they paid for them on average.

Choice-Based Conjoint Dual Response SADR N=267 N=308 N=305 stated ↔ predicted stated ↔ predicted stated ↔ predicted Share of Paying Customers 67.04% ↔ 40.82% 67.86% ↔ 29.54% 69.18% ↔ 62.95% RMSE 0.55 0.59 0.50

35

Summary

• SDR and especially SADR with unique features to better measure willingness to pay

 Unique feature of SDR & SADR:  Avoidance of context effects through strict separation of forced and free choice questions into two blocks  Unique additional features of SADR:  Avoidance of extreme response behavior through adaptive free choice questions  Independence between purchase probability and accuracy in measuring willingness to pay through selecting free choice alternatives from the whole range of the preference order  Reduction of number of redundant purchase questions (about 20% less effort)

• Empirical findings

 SADR offers higher internal predictive, convergent, and external validity; large parts of increase in predictive validity stem from the separation feature as implemented in SDR  SADR requires less cognitive effort than Dual Response, but respondents spend the same amount of time  Better "balanced“ and more informative decisions than in choice-based conjoint and dual response

36

Thank you for your attention!

Christian Schlereth Chair of Digital Marketing WHU – Otto Beisheim School of Management +49 (0) 261 6509 455 +49 (0) 261 6509 509 christian.schlereth@whu.edu Bernd Skiera Chair of Electronic Commerce Department of Marketing Goethe University Frankfurt +49 (0) 69798 34649 +49 (0) 69798 35001 skiera@skiera.de

37

Demo: http://www.dise-online.net/demo.aspx

Implementation of SDR and SADR in DISE (Dynamic Intelligent Survey Engine) Part 1: forced choice block

38

<predefinedPages markID="1003"> <cbc percentageStart="25" percentageEnd="60"> <choiceSetQuestion>Which of the tablets would you prefer most?</choiceSetQuestion> <attributes> <attribute> <name>Brand and operating system</name> <isNominal /> <levels> <level><text>Apple (iOS)</text><baseValue>150</baseValue></level> <level><text>Samsung (Android)</text><baseValue>50</baseValue></level> <level><text>Smarttab (Android)</text><baseValue>0</baseValue></level> </levels> </attribute> <attribute> <name>Screen size</name> <isNominal /> <levels> <level><text>7 inch</text></level> <level><text>10 inch</text></level> </levels> </attribute> ... </attributes> <cbcDesign> 1,1,1,2,2,4; 3,2,2,1,1,3; ... </cbcDesign> <configuration> <noChoiceSetsPerPage>1</noChoiceSetsPerPage> <noProductsPerChoiceSet>3</noProductsPerChoiceSet> <hasNoChoice>false</hasNoChoice> <noSortedCompleteFactorialDesign>false</noSortedCompleteFactorialDesign> </configuration> </cbc> </predefinedPages>

Specify all attributes and levels (here, a brand specific base price is specified) Include choice design Specify number of alternatives per choice set as well that choice sets contain forced choice questions, i.e., without a no- purchase option Assign ID “1003“ to forced choice block

SDR & SADR – Separated (adaptive) dual response

Demo: http://www.dise-online.net/demo.aspx

Implementation of SDR and SADR in DISE (Dynamic Intelligent Survey Engine) Part 2: free choice block

39

<predefinedPages> <freeChoiceBlock percentageStart="65" percentageEnd="90"> <question>Would you actually buy the presented tablet?</question> <configuration> <cbcMarkID>1003</cbcMarkID> <noAttributesInCbc>6</noAttributesInCbc> <presentSelectedCbcProducts>true</presentSelectedCbcProducts> </configuration> </freeChoiceBlock> </predefinedPages> <predefinedPages> <freeChoiceBlock percentageStart=“65“ percentageEnd="90"> <question>Would you actually buy the presented tablet?</question> <configuration> <cbcMarkID>1003</cbcMarkID> <noAttributesInCbc>6</noAttributesInCbc> <noQuestionsPerIteration>2</noQuestionsPerIteration> <noIterations>3</noIterations> </configuration> </freeChoiceBlock> </predefinedPages>

SDR – Separated dual response SADR – Separated adaptive dual response

Create link to forced choice block Use mechanism that just shows the chosen alternatives from forced choice block Create link to forced choice block Assign n = 2 and m = 3

Demo: http://www.dise-online.net/demo.aspx