The Keys to Key-Driver Analysis

Today I’m going to return to a favorite topic of mine: customer-satisfaction research. One of the most valuable ways to look at customer-satisfaction data is through key-driver analysis. In the world of customer-satisfaction research, key-driver analysis is used to tell you where you should get the most “bang from your buck” with respect to improving your customers’ attitudes and opinions.

Two pieces of background will be useful in reading this article: my column on customer-satisfaction surveys (if you’re not familiar with the standard format), and my column on correlation, which is the statistical tool underlying most forms of key-driver analysis. Or feel free to drop me an email.

What Is Key-Driver Analysis?

The purpose of key-driver analysis is to identify elements of a Web site experience (commonly called “critical attributes” in the market research world) that are most strongly related to your customers’ behavior or attitudes.

For example, you might be wondering which attribute — “ease of navigation,” “overall site appearance,” or “customer support” — is likely to have the strongest influence on your site visitors’ likelihood to return. Well, key-driver analysis can help answer these questions. Alternatively, you could use this type of analysis to measure the strength of the relationship between critical attributes and other attitudinal measures, such as “overall satisfaction.”

How Is Key-Driver Analysis Done?

The most straightforward method for carrying out key-driver analysis is to look at the correlation between critical-attribute satisfaction scores and the dependent variable that you’re interested in (the behavior or “other” attitude): The higher the correlation, the stronger the relationship between the attribute and the behavior or attitude. Ultimately, the key drivers are the attributes with the highest correlations.

For example, a commerce Web site might measure 15 site attributes. To find which are the key drivers, the company would measure the correlation between respondents’ satisfaction with the attributes and the dependent variable of interest — in this case, “likelihood to purchase” (see table, below).

The highest-scoring attributes, those that we’d be inclined to consider as key drivers, might look like this:

Site Attribute	Correlation With Likelihood to Purchase
Variety of Product Selection	0.83
Ease of Using Shopping Cart	0.80
Customer Support	0.75
Ease of Navigation	0.54

This hypothetical table of correlations suggests that “variety of product selection,” “ease of using shopping cart,” and “customer support” are key drivers of future purchasing intent. “Ease of navigation” might not be considered a key driver, because its correlation is substantially lower than the others.

In addition to simple correlations, other (more advanced) techniques can be used when computing a key-driver analysis. Instead of simply looking at the relationship between individual attributes and “likelihood to purchase,” we could try to isolate the effect of that attribute by controlling for all the others by using a regression analysis.

Whether this is a good idea depends on what you’re doing exactly. That’s a topic that would take us too far into the world of statistics for this column.

Derived Importance

Remember that in addition to rating critical site attributes on the dimension of satisfaction, most customer-satisfaction surveys will ask respondents to also rate the importance of the attributes to their visit experience. This is called “stated importance.” The key-driver analysis can be used to give a different perspective on importance. The stronger the relationship between respondents’ satisfaction with an attribute and their overall satisfaction score, the more important the attribute is assumed to be. This ranking of the attributes is called “derived importance.”

More often than not, the derived-importance rankings and the stated-importance rankings for attributes will be relatively similar to one another. When they’re not, further research may prove useful.

The Usual Warning

Remember the mantra of Statistics 101: Correlation is not causation. “Key driver” is standard terminology, but the word “driver” may be misleading. To get the most out of key-driver analysis (or almost any use of correlation in research), it’s important to keep in mind other explanations than the most obvious one of causality.

In sum, key-driver analysis is a powerful way to derive business value from customer satisfaction data. When it comes time to make spending decisions on your Web site, key-driver analysis can tell you which resource allocation will have the greatest impact. But, as with all research tools, you have to keep your brain turned on when you use it!