Develop: How do you do it?

The design of a new product requires a merging of a designer’s perception of what works physically and technologically with what customers want, need, and perceive is the right design. The former is an engineering perspective and the latter a user perspective. They do not have to agree, and most often will not since they reflect two different groups of people with different agendas. A new product failure results when a designers perspective dominates to the exclusion of what matters to customers. Since customer input is absent in the extreme case, the probability is high that the new product would not meet customer requirements and expectations or even solve any problem they have. There is a Design Failure. The probability of a design failure varies inversely with the amount of customer input as illustrated in Figure 3.1. This chapter focuses on ways to gather customer input into the design process to minimize the probability of a design failure.

This chapter is divided into five sections. The first provides an overview of product optimization to set the stage for the remaining sections. Conjoint analysis, a traditional methodology for measuring attribute importances, is described in the second section. Conjoint analysis is only useful in the early stage of product development, and should be viewed as a methodology that will be superseded by another in the next stage of product development. This second section is followed by a discussion of the Kansei approach to product design in the third section. This approach expands on conjoint analysis by incorporating customer emotions into the design process. Included in this section is a discussion about merging conjoint and Kansei analysis. Early-stage pricing is discussed in the fourth section. I point out that pricing is needed for business case analysis but it is too early to develop final pricing at this point in the product development. I expand on pricing in later chapters. The fifth section is a summary. Finally, there is a technical appendix that provides background on technical issues mentioned in the chapter.

FIGURE 3.1 The design failure rate declines the more customer input there is in the design process.

Product design optimization

A product is not an entity unto itself. It is actually a collection of parts that, when combined in the “correct” combination or proportions, results in a salable product. A product is a container. A smartphone is a container for a screen, a case, a camera, software, and so forth. The parts in the container are features or attributes defining the product. These attributes are themselves defined by features or characteristics which are usually called levels. The levels are the design parameters I mentioned in Chapter 2. So, the camera in a smartphone is an attribute of the smartphone with two levels: it is either built into the smartphone or not. If the camera is built in, then it has levels of pixel resolution, such as 20 megapixels, 40 megapixels, and 48 megapixels. This means a product, as a container, is actually a combination of attribute levels. See Arthur [2009] for a discussion of how items that are called technologies, such as the smartphone, are really containers of other technologies. The design issue is then two-fold:

• 1. Which attributes to include in the product container?
• 2. What level of each attribute to include in the product container?

A design failure occurs when the attributes and/or their levels are incorrect from the customers perspective so that the product is not optimized to sell in the market. Designers, working independently of customers who would eventually buy the product, may not get the right attributes or the most important ones that would motivate potential customers to buy. The product is not optimized. Designers really have to have input from customers at the product design stage, the point at which the attributes and their levels are specified and assembled to create the product.

Certainly, not all the attributes and/or the levels customers say they want are technologically doable; they may be too costly to implement; the segment of the market requesting or demanding those features may be too small to be financially profitable for the business; or their incorporation into the product would require an increase in the price point to a level that would make it unsalable. In short, marketing concerns must be balanced against technology possibilities when determining what to include in the product design.

A traditional way to optimize a product configuration using explicit physical, engineering attributes is conjoint analysis. I describe and illustrate conjoint analysis in the following subsection. Another method, less common in Western product research and design, is Kansci analysis which has been successfully applied in Japanese and Chinese new product development efforts for some time. Kansei analysis is described after conjoint analysis. A methodology for combining the two is also described.

Not to be overlooked and just as important as the attributes and their levels is the price point. Customers will, after all, make a purchase decision not only on the product itself, but also on whether or not they can afford it or if the price conveys a sense of quality. This latter quality-based decision is based on the price point. A price that is too high will restrict the market to only those either with a high income level or those who will overlook price because they must have that new product; that is, the early adopters. Demand for this group would be highly inelastic. For the market as a whole, however, a high price would most likely be more than can be tolerated so a large segment of the potential market would be unable or unwilling to buy. This segment would have an elastic demand. At the same time, a low price may be interpreted as a signal that the product is cheap and so not worth the money. Determining the price point is a difficult process but one that is made more difficult by the fact that customers rarely know how to value a new product before it has actually been introduced; that is, when it is still at the design stage. In my opinion, the best they can do is give a range of values they might and might not find acceptable. I discuss how to find a range in Section 3.4.

Conjoint analysis for product optimization

Conjoint analysis is a member of a family of choice methodologies designed to determine or estimate customer preference for one product versus another. This amounts to determining the optimal combination of attributes and their levels. This family is described in Paczkowski [2016] and Paczkowski [2018]. Another member of the family is discrete choice analysis which seeks to handle the same problem but the context of discrete choice differs from that of conjoint analysis. As members of the same family, they share similar features but differ in important ways. The common features are: ^{[1] [2]}

• 3. a simple evaluative question used in a survey; and
• 4. an estimation procedure applied to the data collected in the survey.

They differ in that, for a discrete choice customer survey, respondents are presented with several alternative products at once in sets, called choice sets, and are asked to select the product from each set they would most likely buy^[3]. The minimum size of a choice set is, of course, two. The reason for this choice approach is that it mimics the shopping behavior of customers: they are viewed as facing two or more products on a store shelf and they have to choose one of them. It is this interpretation of shopping behavior that makes discrete choice more applicable for final optimization and market testing than for product optimization at the design stage. Product optimization still occurs at the market testing stage of new product development but it is more product refinement than initial design. In the market testing stage, the focus becomes more on how the product will perform against competitor products. It is for this reason that I discuss discrete choice in Chapter 4 and not here.

In a conjoint customer survey, respondents are also presented with alternative products, but in sets consisting of a single item; they are presented one at a time. They are then asked to rate their preference or likelihood to purchase but they do not “select” or “choose” per se. You can only select or choose when you have several items to select or choose from. I describe the general conjoint study framework in the next section and the choice framework in Chapter 4.

Conjoint framework

Conjoint analysis involves presenting survey respondents only one alternative at a time so the choice sets are singletons. Rather than being asked to make a choice, which is not possible since there is only one product placed in front of them, customers instead are asked to rate their preference for the product, usually on a purchase intent scale ranging from 1-10, although other scales are certainly possible.

The singleton sets are created using experimental design principles. Typically, a fractional factorial design is created where я full factorial design consists of all possible arrangements of the levels of all the attributes. Sometimes a full factorial is too large for any one survey respondent to handle which is why a fraction is used. The fraction is selected to give an optimal design that allows the best estimation of part- worth utilities. Part-worth utilities are weights placed on each level of each attribute of a product. See Paczkowski [2018] for design principles for conjoint studies.

An OLS regression model is estimated using the preference data with the attributes and their levels, appropriately coded, as the independent variables. A more complicated model is estimated for a discrete choice problem and this estimation will be discussed in Chapter 4. A conjoint model for a product with four attributes for illustrative purposes might be:

where

P'₀ = grand mean preference /Г. = part-worth utility for level j, attribute a ()'_hk = part-worth utility for level k, attribute b p^r(t = part-worth utility for level /, attribute c pj = part-worth utility for level m, attribute d

and Y' is the total utility or total worth to customer r of a product defined by a combination of the attributes in a set r. In this example, r consists of an arrangement of the four attributes, an arrangement derived from an experimental design. So г is the product customers are asked to evaluate. The coefficients in (3.1), called part- worths, are the contributions of the attributes to the total worth. See Paczkowski [2018] for a thorough discussion of this model. These part-worths show the importance of each attribute and, as I explain shortly, the importance of each level of each attribute in defining a product. It is knowledge of these part-worths that enables product designers to formulate the right combination of attributes and their levels, the combination being contained in the new product. Basically, the attributes and their levels are design parameters and conjoint part-worths tell designers how to set these parameters.

• [1] reliance on attributes with discrete, mutually exclusive, and completelyexhaustive levels;
• [2] an experimental design to arrange or combine the levels into choice alternatives that are interpreted as products;
• [3] 2