The context of questioning is clearly
specified with respondents placed in realistic price/cost value situationsUp
to 35 attributes, each with up to 9 levels/ options can be accommodated, but
higher response quality is usually obtained with studies in the 20-25 attribute
range
The context of questioning is clearly
specified with respondents placed in realistic price/cost value situations
Priorities for product enhancement (and
degradation if desired) are carried out in the full view of all alternative
possible feature changes – hence SIMALTO: SImultaneous
Multi Attribute
Level Trade Off
Unambiguous, easy and relevant tasks
for the respondent, resulting in less random variation than from more difficult
questioning approaches
Cost/Price included in the priority
process at both the total product and individual option levels, so this
important influence is not left to respondents’ imagination, or ignored
The questioning discriminates between really important benefits and those
that are merely nice-to-have. Part
of this questioning invokes the homily that the buyer
can’t-spend-the-same-money-twice, relevant to many product/service choices in
practice.
The process prevents respondents voting for the best of everything –
achieving/exceeding affordable expectations is more the essence of the
questioning
|
|
Respondents analysed as individuals – no averaging of response |
|
|
Recognises importance to improve is different from importance to avoid degradation |
|
|
No single equation assumed to replicate the "customer preference model" |
|
|
Rule-based expert system approach minimises data distribution and correlation problems of regression based conjoint utility methods |
|
|
Unique reliability and accuracy measures of the preference simulation are provided |
|
|
On sample sizes over 200 a ‘hold out’ sample provides additional accuracy evidence |
|
|
Simulations based on both liking and value models |
|
|
Feature value and brand worth measured individually prior to their combination for full preference forecasting |
|
|
Recognises a performance options worth (utility) varies, depending upon what other options it is combined with in the final total product specification |
|
|
Allows for customer need based segmentation (clusters) |
|
|
Individual respondent price sensitivity can be incorporated if desired |
|
|
Incorporates and simulates activity measures, e.g. “use more”, “recommend to others”, “switch to” (including inertia factors) |
|
|
Users can sense-check forecast simulations with the meaningful input information to the models |
Incorporates up to 5 types of importance measurements
 |
Direct elicitation (stated) |
|
Derived (difference/correlations) |
|
Priority to improve |
|
Expectation - Current Perception Gap |
|
Willingness
to pay for improvements/ Changes in Satisfaction-Loyalty if made |
Parallels neural net logic/data mining procedures
|
learn from examples |
|
derive key measures |
|
deduce their relevance |
Over-determining predictions is reduced by building logic constraints into process. Examples of “rules” are:
|
first priority improvements more influential than secondary |
|
unacceptable options deter choice |
|
preference positively correlated with increased value |
Rules ensure sensible predictions from smaller samples than otherwise needed from random searching for causality
