On Tues October 19th, 2010 at 9:00 AM PST SurveyAnalytics joined Nico Peruzzi, PhD of Outsource Research Consulting to present:
"Why You need to Why You Need to Use MaxDifference (MaxDiff) Scaling of Importance Over Rating Scales."
Over 57 people joined us as we answered the following:
1) What is MaxDiff Scaling and what are its benefits to using it in a research project?
2) How can you develop MaxDiff studies that provide unique data vs. using a standard rating scale, etc?
3) How MaxDiff studies differ from other trade-off analysis techniques.
MaxDiff (Maximum Difference Scaling) is an approach for obtaining preference/importance scores for multiple items:
*MaxDiff is also known as “best-worst scaling”
Rating scales make it very difficult to analyze nominal differences between different attributes. We cannot say that feature "A" with a rating number of 2.5 is 50% less preferred than feature "B" rated at 5. This is where the use of MaxDiff can make a big difference.
1) Determine the number of attributes you want to use (up to 30)
2) Add question text and features
3) Review and Change Settings: Set Max Attributes if you only want a certain number of attributes to show at one time.
4) Preview and send survey
5) Analyze data. MaxDiff scores are easy to interpret. They are commonly placed on a 0 to 100 common scale and sum to 100. Thus, when you see a “10” it has twice as much value as a “5.” Another way to analyze data is to use logit regression analysis to generate share of preference percentages.
Example of Feature Rating:
Example of Logit Regression Analysis Percentages:
Regardless of which one you choose to use, you can’t do this with rating scale results.
6) Follow up all MaxDiff questions by:
-Taking the top finishers and use for analysis
-Run Another MaxDiff question
-Add A Single select/ final choice question to determine which one attribute respondents prefer the most.
- Add A Ranking question in your survey
Question & Answers from Presentation:
Q: Which analysis or command do you use to run MaxDiff in SPSS or PASW?
A: MaxDiff can be designed using SPSS Orthoplan or Sawtooth Software’s MaxDiff Designer. Analysis commonly occurs via Logit or Hierarchical Bayes, which can be found in most robust stats packages such as SurveyAnalytics.
Video/Audio Available to view: http://link.brightcove.com/services/player/bcpid86057647001?bctid=644279154001
MaxDiff Slides from Outsource Research Consulting
About the Presenters:
Nico Peruzzi, PhD is a partner with Outsource Research Consulting, a provider of quantitative research and high-end analytics.
Nico Peruzzi, PhD is a partner with Outsource Research Consulting (www.orconsulting.com), a provider of quantitative research and high-end analytics. He has provided consultation on all aspects of the research cycle to organizations of wide-ranging size and industry. Conjoint analysis, database analytics, data mining, segmentation analysis and predictive modeling are some of his areas of expertise. His clients include Herman Miller, Cisco WebEx, TiVo, Data Robotics, and various smaller businesses, research companies, and consultants small and large. Dr. Peruzzi has a BA in Biological Sciences from UC Santa Barbara, graduating Summa Cum Laude, Phi Beta Kappa, and Golden Key National Honor Society. He holds his MS and Ph.D. in Psychology from Pacific Graduate School (now Palo Alto University).
Esther LaVielle is a Senior Account Manager at QuestionPro and Survey Analytics, which was started in 2002 in Seattle and is now one of the fastest growing private companies in the US. Prior to her adventure at QuestionPro she spent 3 years as a Qualitative Project Manager at the Gilmore Research Group.