Editor’s note: Lisa Wood is the director of the Decision and Market Analysis program at Research Triangle Institute (Research Triangle Park, N.C.), which conducts new product preference studies. Dean Hering is a senior engineer in the Center for Digital Systems Engineering at Research Triangle Institute. Mohan Bala is a management scientist in the Decision and Market Analysis Program at Research Triangle Institute. Todd Romig is a product manager at Volvo GM Heavy Truck, Greensboro, N.C.

When preferences for a new product are fairly well known (e.g., line extensions for consumer products), decisions about whether to offer the actual product often can be made with some confidence. However, when demand for a new product is highly uncertain and prototypes are expensive to develop - as with very new durable products - an approach where potential customers evaluate hypothetical products can be very useful. In this article, we describe how Research Triangle Institute (RTI) used a new tool - TradeOff VRâ„¢, which combines conjoint analysis and virtual reality (VR) - to capture the voice of the customer in the product design and planning process for a new refuse truck at Volvo GM Heavy Truck (VGHT).

Conjoint analysis is a quantitative market research technique for eliciting customer preferences for a set of product profiles where a product is defined as a set of features or attributes. Using conjoint analysis, we statistically estimate customer preferences for the product features or attributes based on customer evaluations of a set of product alternatives defined in terms of levels of different features. This technique has been used extensively by market researchers to gather information on customer preferences for new product features 1,2.

Conjoint analysis for new product design typically consists of the following steps:

  • Define the product in terms of its features or attributes, design conjoint interview, and use interview to collect data on customer preferences for the product features.
  • Statistically estimate a customer preference or utility function that quantifies the relationship between customer preferences and each of the product features.
  • Use the following results in the product design process:

    - Estimated weights for levels of each feature

    - Importance weights for each product feature

    - Preference shares for each hypothetical new product and most preferred product design

Figure 1 shows the basic steps in a conjoint analysis study.

Recently, the use of conjoint analysis to capture the voice of the customer in the quality function deployment (QFD) process has been introduced3. QFD begins with defining customer attributes or customer requirements and then estimating a relative importance weight for each attribute or feature 4. Conjoint analysis is one technique for statistically estimating the relative importance of each attribute and thereby capturing the voice of the customer in a consistent, structured, and quantitative way. In a traditional conjoint interview, product features are described to the respondent and the respondent is asked a preference question. For very new products or for unfamiliar products, it may be difficult for respondents to answer such questions. The addition of VR to the conjoint interview allows us to more realistically capture the voice of the customer in the product planning and design process. In particular, by using VR-based conjoint we can:

  • allow the customer to see and interact with a visual 3D virtual version of the product;
  • configure and gather customer reactions to multiple virtual products; and
  • design the customer’s most preferred product visually and in real time.

Example application

Figure 2 shows a set of features and levels for a hypothetical new refuse truck. In this example, we’ve defined six different features of the truck. Each feature has two or three levels or options. For example, the remote-mounted radiator can be in back of the cab or underneath the cab. Figure 3 shows a traditional conjoint interview question for the hypothetical truck in a trade-off format. In a traditional conjoint interview, respondents would respond to a series of such trade-off questions. The difficulty with such a series of questions is that the respondent has to visualize or imagine what these features actually look like. That’s where VR comes in. VR depicts objects in a 3D virtual environment and allows respondents to explore a product’s features in detail. The illustration on p. 34 shows an example of the interior of a hypothetical truck in a VR environment. In this case, the respondent can look around inside the truck, look out the windshield, turn around in the truck to look out the back window, and sit in the driver’s seat. The respondent can also click on a button to move outside the truck and view the exterior.

In addition to showing virtual products in a 3D environment, by using VR-based conjoint we can allow respondents to interact with the product features - push buttons, turn objects around, and try out some of the features. Also, when appropriate, we can immerse the respondent in an environment. For the refuse truck, it made sense to immerse respondents in the interior of the truck since being in the interior most closely mimics the real world environment for decision-making.

Producing the virtual environment for the VGHT truck began by optimizing VGHT’s existing preliminary truck CAD data to 3D computer models suitable for interactive virtual environments. Each feature of the truck occupies a separate model so that the truck may be built from the appropriate collection of models. We modeled features that did not yet exist in design form in consultation with VGHT visionaries. By modeling each feature as a separate 3D object, it is easy to build any combination of features into a truck necessary for gathering voice of the customer data. For example, for each trade-off question in the conjoint interview, the software configures the truck as a collection of features based on the conjoint interview design.

This modular design also allows for easy changes to the system. For instance, if VGHT changes or adds a feature, the model for that feature is changed or added and the system incorporates the change automatically. By changing a text file, the system can add or remove any number of features desired. This allows for an extremely flexible system for use in future customer data collection.

The same configuration system that we use for displaying the questions and allowing the user to interact with the virtual product is also used for configuring the final virtual product in real time based on the customer’s responses to the conjoint questions. This means that, following the interview, the system immediately builds the customer’s most preferred product and the customer can interact with the product in the same manner as when answering the trade-off questions.

By using existing design data, TradeOff VR can be easily integrated into the product design process as well as the market research process. Virtual prototypes can be developed as quickly as they are designed and at a substantially reduced cost from physical prototypes. Producing a physical prototype for each feature of a new refuse truck, for example, is not economically feasible; additionally, changes are time-consuming and costly. With a virtual prototype, we change the design and update the model, and the customer sees the new change in a matter of hours, rather than weeks or months. In addition, color, texture, and other cosmetic changes can be made almost instantly and incorporated into the system automatically.

Data collection

VGHT presented the 45-minute VR-based conjoint interview to groups of customers in a conference room setting, thereby allowing multiple decision-makers from a single company to respond to the interview. The interview was projected from a PC onto a screen and each customer responded to a series of questions about four different applications of the refuse truck using a wireless polling device. At the conclusion of the interview, the software configured the group’s most preferred product for each of the four applications based on statistical estimation of the responses to the trade-off questions. The system also allowed VGHT to connect a head-mounted display to the computer so that the customer could look around inside a typical truck interior.

Hypothetical results

Below we provide examples of the types of results that can be generated based on the information collected during the conjoint interviews. All of the results are hypothetical and do not represent the actual data collected by VGHT.

Figure 4 shows the hypothetical estimated weights for levels of two features for two applications - manual side-loader and commercial front-loader. The weight associated with a particular level of a feature is a measure of its value to the customer; hence, the y-axis represents the value or utility associated with the levels for each feature. For example, the results for the Interior Walkthrough feature suggest that customers prefer a hard walkthrough with switches on the engine tunnel over either of the easy walkthrough possibilities for the manual side-loader.

Figure 5 shows hypothetical importance weights for each product feature for the same two applications. These results show, for example, that price is the most important feature to customers for the manual side-loader. For the commercial front-loader, both price and remote-mounted radiator are important features. For both applications, the front windshield -- whether one piece or two piece -- is not that important.

Examining the interior features, we found that customers prefer a hard walkthrough with switches on the engine tunnel and a rear window with more wrap-around for the manual side-loader application. For the commercial front-loader, customers prefer an easy walkthrough with switches on the folding arm and a standard rear window. These results are consistent with the estimated weights for the levels of each feature as well as the importance weights.

Powerful tool

This study shows how companies can use VR-based conjoint to capture the voice of the customer in the product planning and design process. In particular, we demonstrated that combining a structured and quantitative preference elicitation technique such as conjoint analysis with visualization is a powerful tool for obtaining customer feedback on hypothetical virtual product features before the products or even the prototypes are ever developed. This type of tool can be integrated into the new product development process and can significantly reduce new product development times 5. The virtual prototypes are based on the CAD data generated for the real product design and allow for product visualization and feature evaluation early in the design phases, saving time and development costs. Because the system assembles the product from 3D models in real time, we can introduce changes, new product features, or entirely new products easily and economically. By allowing customers to interact with a virtual visual product, companies can approach the reality of market-testing true prototype products.

References

1 Cattin, Philippe, and Dick R. Wittink. "Commercial use of conjoint analysis: a survey." Journal of Marketing 46 (Summer):44-53 (1982).

2 Wittink, Dick R., and Philippe Cattin. "Commercial use of conjoint analysis: an update." Journal of Marketing 53 (July):91-96 (1989).

3 Gustafsson, Anders. "Customer focused product development by conjoint analysis and QFD." Linköping studies in science and technology, dissertation no. 418, Linköping University, Linköping, Sweden (1996).

4 Hauser, John R., and Don Clausing. "The house of quality." Harvard Business Review May-June (1988).

5 Rosenberger, III, Philip J., and Leslie de Chernatony. "Virtual reality techniques in NPD research." Journal of the Market Research Society 37(4):345-355 (1995).