LE4 - preference measurement + conjoint

• independent from each other -> every manufacturer of phones could offer different case materials

• mutually exclusive -> phone manufactured by only one manufacturers

• as precise as possible declaring the description

• with a similar number of attribute levels -> otherwise attributes with many levels are overvaluated

• attrubute levels which are combined in a way that no impossible combinations of attribute levels are generated, but unlikely combinations can occur

• influencable by company manufacturing the product

• compensatory relationship to each other

• designed so that they have realistic attribute range

• not unacceptable

Most important factors is the determination of attributes and levels

• approaches not based on pre study

  • researchers or managers instinct, product reviews

• approaches based on qualitative pre studies

  • process tracking

  • personal interviews

  • Focus groups

• approaches based on quantitative pre study

  • direct rating of attribute importances

  • repertidy grid technique

=> how many attributes should be included?

• compositional => self explicated

• decompositional => conjoint analysis

• hybrid => tradition hybrid conjoint

Decompositional

• Dekompersitionelle Ansätze erfordern von den Befragten eine Bewertung ganzer Produktalternativen und zerlegen diese Bewertungen in Nutzenwerte für die einzelnen Attribute. Diese Methoden nehmen an, dass die Befragten rationale Entscheidungen treffen und alle Informationen und Abwägungen berücksichtigen, was zu einem hohen kognitiven Aufwand führt und viele Bewertungen erfordert, um genaue Präferenzen zu ermitteln.

Compositional

• Kompositionelle Ansätze erfordern, dass die Befragten direkt ihre Präferenzen für die einzelnen Attributsstufen einer Alternative angeben. Diese Methoden nutzen häufig Bewertungsskalen und werden dafür kritisiert, dass sie die Abwägungen zwischen verschiedenen Attributen nicht erfassen. Der kognitive Aufwand für jede Bewertung bleibt konstant, unabhängig von der Anzahl der zu bewertenden Produkte und Attribute, was zu einer vereinfachten und möglicherweise unrealistischen Bewertungssituation führt.

• Predict the market share of a proposed new product, given the current offerings of competitors

• Predict the impact of a new competitive product on the market share of any given product in the marketplace

• Determine consumers’ willingness to pay for a proposed new product

• Quantify the trade-offs customers or potential customers are willing to make between the various attributes or features being considered in the new product design

The Idea:

• modify attribute levels to construct different product concepts

• rank or rate (traditional) or choose between (CBC) theses product profiles

• find out the preference of customer

Design product and attribute levels of profiles -> responders evaluate utitlity of each profile -> identidy responders preference

• Bewertung einzelner Produktprofile

• profile method

  • combination of one level of each attribute

  • shows whole product profile

• trade of method

  • combination of only two attributes

  • shows only 2 profiles where you can decide between (trade off)

combination from attribute levels of the product

• reduce number of stimuli to a practical subset of stimuli

• 3 Attributes with 3 levels each => 27 Stimuli

• fractional factorial design

3³ = 27 Stimuli

utility values (ß) schould fit the ranks as best as possible

• MONANOVA or LINMAP for ordinal scale

• OLS for interval scale

Befor: Analysis of utility values on the individual level => Aim: compare utility of several products => normalize utilities

ßgesamt = ßgesamt - ßmin // Σ ßmax - ßmin

to calculate w => w = ßmax // Σ ßmax

utility function = U =

• respondents choose between product concepts (like in reality)

• discrete choice modelling

-> multinomial logistic regression is needed

Pro

• realistic decision making situation

• experimental design is pooled across respondents -> more flexibel designs

• allows a “none” option

Con

• low information density for each participant

  • CBC surveys less information

• most realistic form of preference measurement

• A no-choice parameter can be estimated capturing low purchase intention

• flexibility in the design of choice stimuli

• new und better way to incentive aling respondents?

CBC: bezieht sich auf Wahlentscheidungen zwischen ganzen Produkten, praxisnah

Conjoint: Attribute werden verglichen

1. determine attribute + levels

2. design experiment -> choice tasks

   1. determine number of choice tasks

   2. design of stimuli

   3. explicating the decision context

   4. test the efficiency of design

3. respondents answer choices

4. estimate partworth utility

5. evaluate validity of the model

6. interpret results

7. perform market simulations

• How many concepts (alternatives) per tasks?

  • generally 2 to 5 concepts

  • graphical representation affect the decision

• How many tasks per survey?

  • respondents take about 7 minutes to answer 20 tasks -> 20 second / task

  • minimum is one task

  • generally 12 to 18 tasks

• Include a no-choice option?

  • pro: not forced to choose, information about interests in product

  • cons: much less information for utility, difficult to interpret

• Include a no-choice option?

  • pro: not forced to choose, information about interests in product

  • cons: much less information for utility, difficult to interpret

• model fit -> how plausible are the estimated utility values

Utility function composes of a deterministic (= value of attributes) and a stochastic (latent, not directly observable) component

𝑢ℎ,𝑖 = 𝑣ℎ,𝑖 + 𝛿ℎ,𝑖

S. 52 einfügen?

CBC is most used

• number of attributes

  • <8 = cbc

  • >8 = ACA, ACBC, PCPM, ASE

• mode of interviewing

  • require computer: ASE, ACA, ACBC

  • pen and paper: CBC, TCA, PCPM

• sample size

  • no use of CBC by small sample size (<100)

  • ACA, PCPM, ASE possible for small sample size

• interview time

  • few minutes = CBC

  • more interview time needed = ACA, PCPM, ASE

• pricing research

  • CBC and ACBC

• menus

  • Sawtooth Software has created Menu based conjoint for this purpose