Editor’s note: Lee Smith is president and chief operating officer of InsightExpress, a Stamford, Conn., research firm.

Historically, researchers and marketers have addressed issues of market segmentation, brand strength, packaging and pricing, and competitive positioning independently. However, as markets become increasingly competitive, the impact of making a single incorrect decision among the linked set of thousands of choices to launch or reposition a product is often catastrophic.

Conjoint analysis provides researchers with the ability to measure the impact of individual product features and attributes without requiring consumers to evaluate each separately. Through a simple rating, ranking or selection by consumers, conjoint analysis captures the essence of a product as experienced in the real world. Importantly, conjoint analysis decomposes the product, service or offer into its component parts, enabling researchers to: quantify the importance and value of each element; recombine elements to create products of interest to target audiences; and accurately predict share among a competitive set.

Going mainstream due to the ability to develop sophisticated online surveys, conjoint analysis has become an important research tool because, as reality dictates, consumers purchase products - not individual features such as the color red or brand ABC. Using conjoint analysis, practitioners can present products, services or offers with thousands of potential combinations to respondents in a user-friendly manner. While attribute batteries and other traditional techniques were used in the past to understand basic consumer needs, these approaches are comparatively limited in scale and do not enable the researcher to effectively balance elements consumers find desirable (i.e., specific functionality, etc.) with those they do not (i.e., higher prices, etc.).

While marketers in the advertising and consumer products sectors have been conducting conjoint research for years, additional industries are embracing this technique. For instance, many financial service firms are using conjoint solutions to craft the “perfect” credit card solicitation by customer segment, balancing, for example, cash-back or frequent-flier mile reward systems with higher annual fees or interest rates.

Developing an online conjoint survey is not difficult, but it does require planning. At the heart of any conjoint analysis is the design, which is most commonly presented as a matrix of attributes and levels. The attributes represent the dimensions of the product, service or offer. Consider a cell phone manufacturer seeking to introduce a new product composed of a monthly fee, included minutes, scope of coverage, applicable roaming fees, whether or not the phone has Internet access, and brand of the company providing the service; these are the attributes of the design, of which there are six in this example (Table 1).

When establishing the design, it is important that the attributes be independent to the greatest extent possible. That is, removing one attribute from the design does not impact how the respondent will likely evaluate the levels of other attributes. To further develop the design, features that are typically represented as checkboxes in traditional research should be defined as an attribute with two levels - either yes/no, included/excluded, or other similar choices.

By necessity, nearly all conjoint designs include a Price or Fee attribute enabling the respondent to trade-off price for various combinations of features. The presence of a Price attribute is suggested if elasticities are to be calculated or should revenue/profit optimization be conducted. To measure a brand’s impact upon the purchase decision or to conduct competitive what-if scenarios, a Brand attribute should be included in the design as well.

Once the researcher has identified the attributes, they need to turn their attention to defining the levels. There are two or more levels for each attribute. Levels represent the potential values or enumerations for an attribute. In the above example, the Coverage attribute has three levels - local, regional and national. While levels for any given attribute may overlap, they must be in the same category.

A common challenge in developing a conjoint design is managing its size and complexity. A large design results in many potential products, increasing the number of offerings, or “cards,” a respondent must evaluate. While the above example represents 4*5*3*3*2*5=1,800 unique cell phone offerings, conjoint analysis software reduces the number of cards a respondent must evaluate to a manageable level. To determine the minimum number of cards a design requires, subtract the number of attributes in the design from the sum of all levels and then add one. The above example requires a minimum of (4+5+3+3+2+5)-6+1=17 cards. (Note: It may often be desirable to employ more than the minimum number of cards to heighten the robustness of the design.The use of an Ideal-Point or Vector attribute may further reduce the number of required cards.) To minimize respondent fatigue and drive research quality, it is often recommended that a survey not contain more than 25 products or cards.

When developing the design, a researcher should keep an eye toward the challenges the conjoint model will be used to address. As such, it may be desirable to exclude or replace one or more of the cards generated by the conjoint analysis software. This should only occur when a card representing a completely unviable offering is generated. Despite the temptation to eliminate all marginally marketable products, be aware it can be advantageous to have both highly desirable and undesirable offerings for survey participants to evaluate. Retaining these products tends to maintain the distributions of levels while increasing the quality of the conjoint design, accuracy of importance and utility calculations, and robustness of simulations.

Additionally, most conjoint analysis solutions provide the ability to select the “type” of attribute - either Part-Worth, Ideal Point, or Vector - as indicated in Table 2.

Defining a continuous attribute as an Ideal Point or Vector provides substantially greater flexibility when conducting what-if simulations - enabling levels not originally included in the design to be used in subsequent simulations. For example, if Monthly Fee was defined as Part-Worth, the researcher would be restricted to simulations employing only the four levels specified in the design. However, if Monthly Fee was defined as either Ideal Point or Vector, any value between $19.99 and $149.99 could be entered into the simulation - enabling additional business opportunities to be evaluated without the need to re-field the survey or collect additional data.

Without question, the power of any conjoint initiative resides within the simulation capabilities - driven by the underlying design. Online simulators (Figure 1) provide interactive decision support, enabling the practitioner to explore and understand the impact of consumer preferences beyond many traditional reporting methods. The conjoint analysis design principles outlined above have been developed over years of conducting conjoint studies across many industries. We will discuss interpretation of results in a subsequent article.

In summary, conjoint analysis - enabled by the Internet - is a powerful tool enabling researchers and marketers to better understand traditionally hard-to-measure consumer opinions without the need to inquire upon each facet of a product, service or offer.

Conjoint analysis is an excellent solution to identify feature sets for new products, create the best offers and messages, and quantify price and brand impact within the decision making process. And in developing the proper conjoint design, a researcher can identify the optimal product, conduct real-time what-if simulations and substantially increase the probability of success in the marketplace.