Editor's note: Joseph Curry is a vice president of Sawtooth Software, a company that writes and markets microcomputer software for marketing research. Since 1978, Curry has been involved in the development of microcomputer software systems for interactive interviewing and data analysis.
Pricing is a complex area of marketing. Much theory exists about pricing, but it often doesn't hold up when applied to "real-life" situations. So, marketers are often forced to resort to expedients such as cost-plus pricing or matching competitors' prices, and as a result end up leaving money on the table.
The situation is especially complex for products with separately priced features: cars, telephone services, PCs, condominiums, vacation packages, cable TV, and other products where the buyer can select from a menu of options. This article considers pricing product features.
Among the key determinants for pricing products or product features are prevailing market prices, production costs, desired margin, and the price sensitivity of the market. Market prices and desired margin are relatively easy to determine. Calculating production costs is more problematic: it depends on expected volume and learning curve effects. Estimating price sensitivities, more often than not, stops marketers in their tracks.
There are three ways to estimate price sensitivities. The first is to analyze actual sales as a function of price. Data may be available through company records, consumer panels, store audits, or retail scanning systems. Price sensitivity estimates can be derived by regressing sales volume or market share against price. This approach can also be used to develop in-store experiments for measuring price sensitivities.
Although estimates derived from sales data are the closest marketers come to measuring actual price sensitivities, this approach is often unworkable: it cannot be used for new products or for existing products unless detailed historical information exists.
Laboratory purchase experiments are a second way to estimate price sensitivities. Here, "buyers" are asked to participate in a simulated shopping trip and make purchases from an array of goods, including the product for which price sensitivity is being estimated. To arrive at an estimate, the price of the product is varied and changes in "demand" are measured.
Laboratory purchase experiments can be used for new products or for products where no historical data exist. They can also be used to control for buyer demographics or other variables that often contaminate price sensitivity measurements. Laboratory purchase experiments, however, are used only on a limited basis because of their high cost and the low number of concept alternatives that can be tested. For products with more than a few features to be priced, the latter becomes a problem.
A third way to estimate price sensitivity is through preference studies where "buyers" are asked to express their purchase likelihoods for a product at various price levels. Preference studies have a number of advantages: they are relatively inexpensive, variables are easily controlled, and any product can be studied. Their main drawback is that the price sensitivity estimates they produce can lack credibility because the1 circumstances under which they are measured are far removed from an actual purchase situation. According to Nagle, the best preference study methodology for estimating price sensitivities is conjoint analysis. In a conjoint study "buyers" are shown several alternative forms of a product concept. The concepts are profiled in terms of a set of attributes, including price, and buyers rate their preferences for each alternative. Preference and price sensitivities are then inferred from these ratings. What makes conjoint results more credible than those of other techniques is that buyers are forced to make tradeoffs among product features (including price) similar to those they make in actual purchase situations.
Conjoint analysis has been used more often to set product prices than to price product features. This is because conjoint traditionally uses just one pricing attribute and one attribute cannot be expected to cover a $15,000 automobile and its $300 radio. A new form of conjoint analysis (see ref. 2) allows the estimation of both feature prices and overall price.
Suppose a company that produces cameras is about to market a new automatic 35mm camera. To keep the example simple, suppose that this camera has just two options: lens type and flash. The lens type can be either auto focus or auto focus with zoom and the flash can either be standard or high-speed.
To measure the market's price sensitivities, a sample of camera buyers is asked to complete a questionnaire with conjoint questions where the prices of the options are varied in the range of established market values. Figure 1 shows two examples of the types of questions we might ask, presented in a pairwise format. Note that there are base prices, feature prices, and a total price for each concept. The buyer indicates which concept he prefers and the strength of his preference.
 |
By varying the prices for specific features, we can infer whether the buyer has more price sensitivity for one feature than others.
Analysis of the conjoint data results in a set of values, or "utilities," for each buyer which reflects that buyer's preferences - the higher a utility for a feature at a given price, the more the buyer values that feature. Figures 2a and 2b show one buyer's utilities as a function of price; that is, the price sensitivity for that buyer.
 |
Figures 2a and 2b reveal very different price sensitivities. In Figure 2a we see that the auto focus lens priced at $50 has about the same utility as the auto focus lens with zoom priced at $100. We would expect, therefore, this buyer to be indifferent between these two alternatives. The shallow slopes below $50 for the auto focus lens and below $100 for the lens with zoom suggest that this buyer is relatively insensitive to price changes in these ranges. On the other hand, the steep slopes above $50 and $100 indicate that the buyer is quite sensitive to price changes in these ranges. Note that this buyer shows a slight disutility for the zoom lens priced belong $100, perhaps indicating that quality is being inferred from price in this range.
Figure 2b shows that this buyer prefers the high-speed flash over the standard flash throughout the range of prices tested. The buyer is very sensitive to the price of the standard speed flash and less sensitive to the price of the high-speed flash. For this buyer, a high-speed flash appears to be a requirement.
This analysis can be extended easily to market segments or to the market as a whole, by aggregating results of individual buyers. From this information we could decide whether or not it would be worthwhile to develop specific feature pricing strategies for different market segments. For example, we might discover that there is one strategy for first-time buyers and another for the replacement market, or one for those who buy through camera shops and another for those who buy by mail order.
One reason conjoint analysis has become so popular is that its utilities can be used to construct computer-simulated market models. Buyer preferences are represented in the model by individual respondent utilities, which reflect the measured price sensitivities. Products are represented as combinations of features at specified prices. Market simulation models let marketers ask "what-if" questions in a context that approaches the complexities of the actual market. Figure 3 shows a price sensitivity curve for the entire market generated from a series of simulation runs where the price of the new camera's high-speed flash was raised in increments from $60 to $80.
 |
It is important to note that as compelling as the information in Figure 3 might be, it is only an estimate derived from a model. It should not be acted on alone; rather it should be used in the context of a more complete understanding of the buyers in a market. Nagle points out that managers must "first learn to know their buyers" by performing what he calls a managerial analysis of a market's price sensitivity. He states that "if managers try to use empirical estimates of price sensitivity as a substitute for knowledge of their customers' purchase motivations, attitudes, and incentives, the quality of their pricing decisions will suffer." On the other hand, if used properly "it can give managers new, objective information that can either increase their confidence in their prior judgments or indicate that perhaps they need to study their buyers further."
Pricing product features will continue to be a complex decision. Fortunately, decision aids are now available that are more equal to the task than ever before. Perhaps the time has come when we can stop leaving so much money on the table.
References
1. Thomas T. Nagle, The Strategy and Tactics of Pricing, Prentice Hall, Englewood Cliffs, NJ (1987).
2. Richard M. Johnson, "Conjoint Value Analysis, " technical paper, Sawtooth Software, Ketchum, ID (1990).