# Use Both Univariate and Multivariate Statistics To Make Decisions

Clients sponsor ongoing research programs and a variety of ad hoc projects with specific purposes in mind.�But at the end of the day, they all come back to essentially the same two questions: �what does this mean?� and �what do we do with it?��After collecting data from hundreds, even thousands of respondents, how does one make sense of all the data?

Both univariate and multivariate statistical methods give us the tools to look at how customers rate different aspects of products or services against their expectations as well as how important specific aspects of the product or service are in determining their level of satisfaction or willingness to purchase.

Most people are more familiar with univariate statistics. These include means, medians, standard error, and standard deviation, among others. Common uses of univariate methods include calculating the margin of error in a measure or significance testing between two groups.

For a practical example, if 70 percent of all customers surveyed answered that they were extremely satisfied with the courteousness of the customer representative, but only 60 percent of customers in the Southeast region gave the same rating, we can run tests of significance to determine if that difference is statistically significant at a desired confidence level (most often 95%).

To give another example, if 20 percent of customers surveyed rated the durability of a product as poor this year, but only 10 percent rated it as such last year, significance testing can determine if this difference represents a drop-off in the perceived durability of the product or if it could just be within the range of variability expected given the sample size.

Univariate methods such as significance testing are great tools for determining differences between two groups in how they rated a particular product or service attribute. What univariate statistics won�t tell you is how important a particular attribute is to customers.

Using the example above, you may know that customers rated the durability of the product worse this year than they did last.�What is not known is whether durability is important to customers in their overall assessment of the product or willingness to buy it.�If the product were an automobile, durability would likely be an important part of the customers� decision criteria, whereas if the product were a disposable ink pen, durability may not be of much concern at all.�As this example illustrates, univariate methods alone are insufficient for making decisions.�After all, why would one dedicate precious company resources to improve a product attribute that, while rated poorly, is of little concern to customers?

Multivariate statistical methods can be used to find the relative importance of specific product or service attributes.

Multiple regression analysis is often used for this purpose, and is commonly referred to as driver analysis. For a practical example, a key driver analysis may indicate that customers� ratings of product durability of an MP3 Player account for 20 percent of the variation in their willingness to purchase it, while ratings of portability account for 40 percent of the variation. Therefore, based on this analysis, portability is twice as important to customers as durability. Still, this analysis alone doesn�t offer much insight in terms of where to focus resources.�While portability may be of great importance to customers, the multivariate statistic alone doesn�t tell you how portable customers perceive the MP3 player to be.

In order to make sound business decisions regarding how to improve your product or service, savvy researchers must employ a combination of univariate and multivariate methods.�Multivariate statistical methods will derive what customers think is important. Univariate statistics will tell you how well the product or service performs along those key attributes.

To continue the MP3 Player example, if the portability of the MP3 player is very important to customers but is also the highest rated among competitors, it may not be the area of greatest opportunity.�Durability, while not as important as portability, may be a better place to invest in improvement if significance testing determined that the MP3 player significantly underperformed the competition.

The key to good decision making is to find product or service attributes that are of greatest importance to customers and also under-perform some standard such as past performance or competition.�Improvement in these areas will offer the biggest opportunity to make an impact in overall perceptions of the product or service and, as such, will help dictate where valuable resources should be allocated.