Burke uses the power of the Internet so that clients can take full advantage of the possibilities offered by conjoint and discrete choice research. COSMOS^®, Burke's hybrid conjoint approach, keeps complex projects manageable in length, while also allowing companies to derive utility scores at the individual respondent level. CHOICES™ is Burke's individual-level approach to discrete choice, in which respondents replicate the buying process by choosing from among a set of fully described options.

Burke emphasizes individual-level approaches because individual consumers differ in what features they consider and what they buy. Determining the relative attractiveness of products, brands and price for each individual Burke surveys leads to better targeted marketing efforts.

For both conjoint and discrete choice projects, use of the Internet makes it possible to fully randomize attribute levels in profiles and estimate interactions of variables in addition to main effects.

Conjoint research assumes a product or service is a bundle of features that are "considered jointly." It is distinguished from compositional approaches, which ask respondents to rate the importance and desirability of product/service features one at a time, then combine the scores to derive an overall rating for a particular product or service.

Conjoint research typically requires using experimental design to put together "full profiles" of product/service packages. Respondents will then provide an overall rating or ranking to unique collections of product/service features. Analysis can then uncover the main effect of each feature level.

Experimental design ensures orthogonality - or independence of features. Unfortunately, an orthogonal design can sometimes require an unwieldy number of profiles.

Burke developed COSMOS, which stands for "COmpetitive Strategy and Market Option Simulator," as a way to estimate utilities for large conjoint designs using only a limited number of full-profile ratings. It also incorporates self-explicated ratings of the desirability and importance of product/service features, then uses a series of regression techniques to derive utility scores at the individual respondent level for every level of each feature in the study.

A major financial services company wanted to measure the value of potential features for a new "smart" credit-card offering. It turned to Burke to conduct a COSMOS study with members of an online research panel. Burke also recruited respondents at malls to take the same survey, so that the company could check whether findings varied depending upon recruitment method. There was not a significant difference in general interest in the concept between the two groups, and the cost of recruiting people at the mall turned out to be more than three times as much per respondent compared to using the online panel.

Armed with findings from the study, the company knew how much various features would add to the overall perceived utility of its smart credit card offering. It was also able to identify segments based on the appeal of potential card features.

CHOICES™, Burke's proprietary extension of discrete choice modeling, is designed to overcome inflexibility and inaccuracy often seen in "total sample" or "aggregated" approaches.

CHOICES creates predictions at the individual respondent level, taking advantage of known information about sample members, such as category usage, attitudes and demographics. Because it works at an individual level, CHOICES enhances flexibility for performing market segmentation, subgroup analyses, and target market profiling.

The algorithm behind CHOICES can also produce overall results that are simply better fitting and more accurate. Because CHOICES uses more information, it can improve models' predictive power and reduce statistical error.

CHOICES has been applied in dozens of situations to:

• Set prices for maximum profitability.
• Evaluate whether a new concept has market viability.
• Configure a new product or service.
• Re-configure an existing product to increase profitability.
• Better understand what features drive customer preferences.
• Design line extensions that minimize "cannibalization" of existing products.
• Isolate effects of a change being considered in product or marketing strategy.

A telecommunications company wanted to learn more about the increasing convergence with entertainment services taking place in its industry. Burke programmed a Web-based CHOICES study, asking 1,500 online panel members to indicate their preferences from sets of individual and "bundled" services offered by national and regional providers of local, long-distance, and wireless telephone service, Internet access, and cable television and other entertainment content. The service choices in each set were randomized within a complex set of constraints, resulting in data on 9,000 unique choice sets .

Burke then developed a simulator that let the company analyze consumer preferences at an individual level. The company could customize the model for communities with different demographic profiles, and also adapt it as demographic and technology-usage patterns changed. The research thus helped the company with its immediate forecasting - and results were incorporated into a five-year model as well. The company's efforts were also cited by the American Marketing Association as part of its annual recognition of "exemplary performance and leadership in online research."