Table 1: Attributes and attribute levels for the conjoint experiment.
Offering the goods your customers want
by Bradley Jones
A well-made product that nobody buys is not profitable for the manufacturer. A website full of confusing and distracting ads and visuals is not an asset to the online marketer. And a well-stocked shelf of an unpopular item is a liability to the store owner. Why do businesses blindly introduce new products or services and simply hope that they sell? Surely it is safer to ask customers what they want before investing too much, especially in tough economic times.
Conjoint analysis is a technique for evaluating goods by considering their attributes jointly.
Products and services usually have several features that make them desirable. While an individual attribute of a product may be the primary feature, the decision to purchase comes from weighing all the attributes together. For example, the main purpose of cell phones is to allow people to talk regardless of their locations. But people buy cell phones for many additional reasons. Conjoint analysis is a technique for evaluating goods by considering their attributes jointly.
A conjoint analysis has three parts – a designed experiment, the statistical analysis of the resulting data, and the business decisions based on this analysis. Let us illustrate these steps using a recent project.
A local grocery store currently offers only domestic brands of beer. The store manager would like to expand the assortment with one or more imported beers. The manager wants to find out if imported beers are desirable, and determine the best packaging and pricing strategy. Conducting a conjoint analysis is a smart way to improve the store’s beer revenue.
The first step in designing a conjoint experiment is to choose the attributes of the product to study. Table 1 describes the packaging, pricing and other attributes the store manager considers important.
- First, is there a preference for bottles or cans?
- Because there is limited shelf space, it is useful to know whether customers prefer packages with 4, 6, 12 or 24 containers.
- The freshness indicator was popularized by domestic products and consists of a label on each container indicating either when the beer was packaged or by what date the beer should be consumed. The store manager knows that most people prefer the presence of such a label, but is curious about the relative importance of this attribute in the purchasing decision.
- The two categories of beer in the study are imported and domestic.
- Price per container of beer could be a deciding factor. However, brand identity may actually be the more important distinction.
Here is an example of a prospective product, called a profile: 24 bottles per package, no freshness indicator, domestic beer, priced at $.85 per bottle.
If a survey were constructed that offered the possibility of choosing any combination of the attributes in Table 1, a respondent would have to evaluate 128 possible combinations and make a single response. Instead each respondent usually compares two to four profiles and chooses one. An individual comparison is called a choice set. Each respondent usually evaluates several choice sets.
Suppose the business owner can interview 12 customers and expects each customer to fill out a choice survey sheet like the one in Table 2, which shows hypothetical results from a single respondent to eight choice sets.
- Each column in the survey identifies a beer attribute.
- Each line in the survey defines a beer profile, which is a collection of attributes.
- Each choice set consists of two attribute profiles.
Getting interpretable results with short surveys and few respondents requires a cleverly constructed designed experiment. When choosing choice design, you can specify any number of profiles per choice set -- this example uses only two. Additional options include the ability to generate multiple surveys, specify the number of respondents expected to complete each survey, specify the number of choice sets per survey, and limit the number of attributes that can change within a choice set.
In this study, 12 respondents completed surveys like the one in Table 2. The beer choice survey data was then analyzed by the Choice modeling platform.
The two most important factors shown by the analysis in Figure 1 are package size and style (domestic or imported). There is a clear preference for smaller package sizes and imported beer. Surprisingly, there is very little price sensitivity. This may be due to the fact that all the respondents were affluent males for whom the other attributes were more important.
The store manager now knows what his affluent male customers prefer, what beer to stock and how much he can charge. He knows to:
- Carry six-pack containers.
- Buy bottles, not cans.
- Make sure there are freshness indicators on each item.
- Stock imported beer.
- Price each bottle at $1.33.
Since there was little price sensitivity, he set the price at launch to $7.99 per six pack, which works out to $1.33 a bottle.
Knowing the relative value consumers place on various product features dramatically improves the odds of a successful launch of a new product or service. Conjoint analysis is a powerful tool for getting this information quickly and with little cost.