PRACTICAL SOLUTIONS
Extending Choice-Based Conjoint on the Web
This article is based on a paper presented by Joseph Curry of Sawtooth Technologies at the Tenth Sawtooth Software Conference, April 2003. The complete paper can be downloaded from the Sawtooth Technologies Web site at www.sawtooth.com.
Choice-based conjoint is a popular research technique for collecting data for making product design and pricing decisions. Choice-based conjoint data collection took a giant leap forward with the introduction of "off-the-shelf" interviewing software designed specifically for collecting choice data via the Web.
Much of the power of Web-based choice software results from the fact that it makes the technique accessible to a wide range of researchers while limiting its potential for misapplication. But in doing so, the software sets limits on the choice designs it can create. Although these limits are not significant in the majority of studies, most frequent users have bumped up against them.
Two such frequent users are Richard Miller of Consumer Pulse and Dirk Huisman of SKIM. Both get around the data collection limits imposed by the "off-the-shelf" choice software by creating extensions of the choice-based interview approach using Sensus Web.
Miller's clients need to test prices ranges that are conditional on other attribute levels that make up the product concepts being tested. Figure 1 shows an example of the type of conditional pricing table he would employ for high-definition televisions (HDTV's). Here the specific price ranges used in constructing HDTV concepts would depend on brand, technology and screen size.
Figure 1: Price Table for HDTV's, where price ranges are conditioned on brand, technology and screen size
Using Sensus, Miller can implement conditional pricing on the Web. He can also randomize tasks within sets of choice tasks, and randomize choice task sets across respondents. As a result, Miller states that he can deliver more realistic and precise price information to his clients.
Huisman creates choice tasks for testing the impact of attributes that convey non-verbal information. He uses Sensus to construct the tasks and to dynamically vary all of the attributes that appear in the choice concepts. An electric toothbrush example he created is shown in Figure 2.
Figure 2: Choice task for testing non-verbal attributes
Huisman plans to use this approach to test whether including non-verbal information in choice tasks leads to better share predictions, less sensitivity to price, and a higher impact of promotions.
What do they do with the Sensus data once it's collected? Both Miller and Huisman take their Sensus data, reformat it, and then import it into their "off-the-shelf" choice-based conjoint packages where they estimate utilities and run market simulations.
By using Sensus, Miller and Huisman are able to generate choice-based conjoint designs that match their clients' markets and better mimic reality in situations where they run up against the current limits of the "off-the-shelf" systems. Their overall goals are to accommodate the needs of their clients and to produce results that have greater predictive validity.
[Back to Top]
IN BRIEF
New International Rep
We are pleased to announce the appointment of a new sales and support representative for Europe, Great Britain, and Ireland: PTL Computer Services. PTL Computer Services has many years of experience with our CATI products. They will be replacing SKIM, whom we would like to thank for their many years of service.
To schedule demonstrations or training, or to receive information or support, contact:
Paul Lamkin
PTL Computer Services
6 Hadrian Gardens
Kingsnorth Ashford
Kent TN23 3PH
United Kingdom
