|
||||||
Technical Papers| Home |
Solutions |
Products |
Services |
Education |
Support |
| | ||
|   SS Summer 1999Eighth Conference Planned for East CoastThe eighth Sawtooth Software Conference will be held on March 21-24, 2000 on Hilton Head Island, South Carolina. Hilton Head is an upscale island resort community off the southern coast of South Carolina. This is the first time we have held our conference on the East Coast. We hope that those of you who have been waiting for us to meet on the East Coast will take advantage and attend.Our conferences are more down-to-earth than most quantitative market research conferences. While a few academics attend, most presenters and attendees are consultants/practitioners. The aim is to give practitioners useful, relevant and understandable instruction in quantitative market research techniques without infomercials from consultants or sales presentations of our own software. Our last conference covered such practical topics as data collection, Internet research and improving response rates. About half of the papers presented reported case or methodological studies using various conjoint and choice techniques. Other methods covered in past conferences include perceptual mapping, segmentation techniques, customer satisfaction models and neural nets. We value informality, interaction and good food when we meet. If you've never come to our conference before, you are in for a treat. You can get a feel for the topics presented by reading the abstracts from past conferences in our News area of our home page at www.sawtoothsoftware.com. Call for Papers Papers presented at previous Sawtooth Software conferences are cited frequently in journal articles. We are once again looking for exceptionally strong papers. If you'd like to be on the program, please respond promptly with a one-page abstract describing your proposed paper, with special attention to what the audience will "take away" from the presentation. To be accepted, a paper must show promise of being sufficiently practical to be of use to the least sophisticated members of the audience, while having enough substance to be of interest to the most sophisticated members. Please Email abstracts to bryan@sawtoothsoftware.com. In addition to giving an oral presentation at our conference, presenters are expected to produce a polished written paper suitable for publishing in the conference proceedings. For examples of papers published in past volumes, please see our Technical Papers Library on the web (www.sawtoothsoftware.com). Scaling Conjoint Part Worths: Points vs. Zero-Centered DiffsFor over a decade, a re-scaling method called "Points" has been the default for reporting conjoint part worths (utilities) in Sawtooth Software's conjoint simulators. This approach gives equal weight to each respondent by making the total sum of part worths equal to the number of attributes times 100. Since conjoint part worths are scaled to an arbitrary constant within each attribute, it seemed to us early on that setting the worst level of each attribute for each respondent to zero was as reasonable a decision as any. This held promise of making it easier to present part worths to clients, since the values would generally fall in a comfortable range of about 0 to 100 (the values can actually exceed 100). However, we now suspect that fostering familiarity has sometimes led to misinterpretation. With the benefit of years of hindsight, we think that Points was a mistake.We are abandoning the Points re-scaling method in all future versions of our software for the following reasons:
We do not plan to offer the Points re-scaling method in future versions of our conjoint software systems. Rather, zero-centered Diffs will be the default method. (That method has been the default for our CBC Latent Class Module.) We have noticed only one small benefit of Points or Diffs relative to zero-centered Diffs. If the worst level of an attribute has an average utility of zero or near-zero, it means that nearly all of the respondents agree that this level is the least preferred. If respondents disagree about the order of preference (say, for brand), the worst level of an attribute is significantly greater than zero. When viewing average zero-centered Diffs, one cannot gain a feel for the degree of disagreement regarding attribute levels. However, a new option available in the next generation market simulator captures that information in a different way by displaying average attribute importances. If you are using individual-level or latent-class part worths, the attribute importances can help you gain a feel for the level of agreement or disagreement regarding level order. Assume you are studying two brands and the average part worths are nearly tied at zero (zero-centered Diffs). If brand also has a significant attribute importance, you would know that respondents really aren't indifferent about brand. Rather, there are distinct groups that feel strongly, but oppositely, about which brand is preferred. We don't expect that making the change from Points (or Diffs) to zero-centered Diffs will come easily to those used to seeing part worths scaled as Points. Some may argue that the zero-centered Diffs are not as easily interpreted or useful as Points utilities. We contend that zero-centered Diffs communicate the real information available from the part worths and their use minimizes the opportunity for misinterpretation. Since our software always makes part worths available in ASCII files, users bent on using one of the previous methods can always re-scale the data to their preference using other analysis software. Hierarchical Bayes Estimation and ICE RevisitedAs you have probably noticed, hierarchical Bayes (HB) has been the subject of several favorable journal articles and many presentations at the more technical market research conferences. We expect that the amazing advances in the speed of computers will hasten the adoption of hierarchical Bayes algorithms not only for conjoint problems, but for other market research applications as well.Until now, there have been two problems with HB: it can take a lot of computer time, and user-friendly software has not been generally available. We think that we have, by in large, solved both of those problems with the CBC/HB Module we recently released. CBC/HB is easy to use and runs rapidly, so answers for medium-sized problems are available in a few hours. Those familiar with our ICE (Individual Choice Estimation) software may wonder how HB compares to ICE. What is our position on these two pieces of software that both achieve individual-level estimates from CBC data? At the 1998 Advanced Research Techniques (ART) conference, Joel Huber of Duke University presented results from three different studies. He and co-authors Neeraj Arora (Virginia Tech) and Rich Johnson (Chairman, Sawtooth Software) found that HB and ICE both performed about equally in terms of hit rates and share predictions for holdout choice tasks. The three data sets that Huber et al. examined had a minimum of 18 tasks each. In the ICE documentation, we suggested that about 20 tasks or more should be available for ICE estimation. Since Huber's presentation at ART, we have seen examples involving data sets that contain less information than suggested for ICE in which ICE has performed poorly, but for which HB has performed favorably. Indeed, the superiority of HB in achieving useful individual-level estimates with as few as six choice tasks is a point of differentiation we had not previously recognized. ICE offers at least two unique benefits. It can be significantly faster than HB for very large problems. Also, if respondents conform to the assumptions of the latent class model (high degree of homogeneity within groups with large separation between groups) and enough choice tasks are available, ICE has the potential to produce results superior to HB. We will continue to market ICE, though for most users we will recommend CBC/HB. Those who own either ICE or CBC/HB receive a $500 discount on the purchase of the other (ICE and CBC/HB each cost $2,000). Announcing Composite Product Mapping (CPM)Perceptual Mapping has been a useful method for condensing what can often become a very large amount of ratings information (such as a matrix of brand ratings on different attributes) into a visual picture:
 The farther a brand is positioned in the direction of an attribute vector, the more highly associated that brand is with that quality. Brands that are located near one another are perceived as relatively similar by the market. The first version of APM was released in 1986 and has never received a major update. We are pleased to announce the successor to the APM system, called CPM (Composite Product Mapping). APM's maps were based solely on respondents' perceptions (ratings) of brands (objects) on different attributes. While these maps were very good at describing perceptual differences in the market, they were often not very good at representing differences in preference. Many researchers and academics have been interested in market maps that are better tied to preferences but that still have the interpretability of discriminant- or factor-based perceptual maps. At our most recent Sawtooth Software Conference, Rich Johnson demonstrated "composite maps" that use both product perceptions and preferences. These maps often look a good deal like APM's discriminant maps, but they are more successful in representing differences in preference. Rich suggested that composite maps offer insurance against poor selection of attributes. If there are attributes that are useful for discriminating among brands in terms of perceptions but that have no explanatory power with respect to preferences, CPM's composite maps will largely ignore those attributes. One of the greatest shortcomings of the APM system has been the unavailability of a Windows-based mapping program. CPM provides such a module. You can control the fonts, styles, sizes and colors in the map. You can "drag and drop" labels to reposition those that might overlap, add titles and annotations, or selectively suppress attribute vectors or products. CPM can create three types of maps:
In the years since APM's introduction, it has become apparent that conjoint analysis provides a more powerful method for "what-if" simulation than mapping-based simulators. Rather than a simulator that tries to predict the result of changing a product on specific attributes (which is better left to conjoint analysis) CPM provides a method for estimating the density of demand at each point in the product space. CPM's Plot module generates Density of Demand plots like the following:
 Darker areas of the density map represent relatively higher average preference. CPM comes in different sizes, based on the number of attributes and products in the map. The prices are as follows:
CBC Version 2 Now AvailableAfter more than two years of development, we are pleased to announce the release of Choice-Based Conjoint (CBC) v2 for Windows. Choice-Based Conjoint (sometimes called Discrete Choice) has gained significantly in popularity over the years. Rather than having respondents rank-order or rate product concepts (defined on multiple attributes) as with traditional conjoint analysis or ACA, respondents perform a task that better mimics what they do in the real world: they make choices within sets of available products. CBC has proven very useful for problems that involve about six or fewer attributes. CBC has particularly found use in pricing research studies, where it can estimate unique price sensitivities for distinct brands.Until recently, the use of CBC has faced a significant methodological shortcoming: it generally could not be used to estimate stable individual-level part worths. Rather, CBC users pooled data by estimating aggregate logit models. Pooling data had the advantage of being able to manage larger models that, for example, could accommodate interaction terms. However, aggregation resulted in simulation models that could provide misleading results due to problems with IIA (Independence from Irrelevant Alternatives), also known as the Red Bus/Blue Bus problem. With the recent availability of estimation techniques that achieve individual-level utilities from CBC data (ICE and HB), this issue has largely been resolved. CBC v2 is the first component within the planned Suite of Marketing Research Tools (SMRT Suite). Other software packages (including ACA) will later integrate within this suite and share many common dialogs (e.g. entering attributes) and functions (e.g. market simulator) with CBC. CBC v2 offers significant improvements to nearly every aspect of the system, including: Questionnaire Development
You can download a free demo version of CBC v2 at www.sawtoothsoftware.com/trial.htm. You can use this software to develop questionnaires and analyze data for a sample study. Since this is a large file to download (roughly 6 MB), we can send you a CD-ROM in the mail if you would prefer. Please call us at 360/681-2300 to request the free CD-ROM. Ci3 TechQuestion: How Can I Remove Duplicate Respondent Records from My Data File?Ci3 v2.5 provides two ways to do this. If you are certain that the respondent numbers are unique in your file, you can simply choose Analysis | Utilities | Remove Duplicates. Ci3 discards any duplicate records based on respondent number. If it is possible that some cases have the same respondent number but really represent different respondent records, you can use the Field | Accumulate Data... option. Copy your questionnaire.dat and questionnaire.idx files from your study directory into another directory on your hard drive (for example, C:\TEMP). Next, delete those two files from the study directory. Accumulate the questionnaire.dat file from the C:\TEMP directory into your study directory with the Check for Duplicates box checked. As Ci3 processes the data, it checks for duplicate respondent numbers. If it detects duplicate numbers, it then does a thorough check of all the information contained in the suspect duplicate record. If all of the information matches perfectly, Ci3 skips the duplicate record. If Ci3 detects any difference, it asks you whether you want to delete the case or include it within the data set. Question: Can Multiple Interviewers Work from a Common Field Disk over a Network? The NETWORK ON instruction placed in the prequestionnaire section of your interview script lets you set up a Field Disk in an interviewing directory on a network drive. Multiple respondents can access the interview at the same time and the data are saved into the common questionnaire.dat file. This is also a simple way to set up a "Poor Man's CATI" system. Your Ci3 license lets you distribute and install Field Disks with no limitations. You can have as many telephone interviewing stations as you like working from the same Field Disk on a network drive. Ci3 currently doesn't provide an interface with databases (that might include customer names and telephone numbers), so you would probably manage your sample and schedule callbacks on paper. You may soon outgrow this "Poor Man's CATI System" and find need for the advanced capabilities of the Ci3 CATI system available from Sawtooth Technologies (847/866-0870). The Ci3 CATI system provides sophisticated capabilities for managing sample from a common database, scheduling callbacks, quota control, and automatic dialing of telephone numbers.
|
|
Technical Papers
|
About Us | Contact Us | Privacy | FAQ's | Forum |
|
© 2008 Sawtooth Software, Inc. All rights reserved. |