Editor's note: Timothy R. Kula is director of customer satisfaction for the Gilmore Research Group, Seattle.
In today's global marketplace, it is not enough to just measure customer satisfaction. To be competitive, businesses need to continually make improvements that will raise customer satisfaction so they can increase what ultimately determines their financial health - profitability.
While measuring customer satisfaction is the all-important first step, unless measurement is linked to improvements, its ability to bolster the bottom line is minimal. This article suggests a new technique that allows you to do both simultaneously: measure customer satisfaction and have a positive impact on profits.
For some companies, measuring and raising customer satisfaction is a centralized process. Often these companies have only a few locations that interact with customers, in which case customer sampling and centralized reporting of customer satisfaction results makes perfect sense.
However, many companies have wide distribution networks (which are generally independently owned) that interact with customers. Customer satisfaction data gathering for these companies needs to be either a complete census or at least a statistically significant sampling at an individual outlet level.
Detailed and customized analysis of customer satisfaction data by individual outlet has not traditionally been done because the task is labor-intensive and time-consuming - and because reporting is not easily understood by non-analytic users. Expert system analysis overcomes the problem by enabling rapid, detailed analysis and truly actionable reporting of customer satisfaction data to be performed for every outlet.
By taking an individual outlet's data and running it through an expert system, a prioritized list of targeted improvement areas is tailored for each outlet. Importantly, the report is designed so that it can be easily understood and used by company field and outlet personnel.
Traditional reporting from customer satisfaction programs can be divided into two basic categories: data and graphics. Unless you're an analyst (and the people who can actually make improvements usually aren't), the reporting can prove frustrating.
Reports with lots of data - for instance, a multi-page report with hundreds of numbers - can be confounding to a territory manager, outlet owner, salesperson or customer service representative. They elicit responses like What do the numbers mean? Which number is most important? What do I need to be working on?
Give me graphics, they say, at least I can understand them. But wait, the graphics are oversimplified and don't provide the detail that's required to identify and fix real customer problems. If only an analyst could be everywhere to explain the numbers, what's most important, and what needs to be done to raise customer satisfaction. Just think what could happen to sales and profitability.
With the implementation of an expert system, this becomes a reality.
A brief history
What is an expert system? It is a computer capability that captures the knowledge of a domain expert to solve a set of problems. Think of it this way: You're an expert at market research analysis; the computer is proficient at high-speed data processing. Combine the two and expert analysis can be performed thousands of times in a matter of seconds.
Expert systems gained prominence in the early 1970s and have been used in many diverse fields. Expert systems have diagnosed and prescribed cures for diseases; configured complex minicomputer systems; provided aircraft mechanics with advice on how to repair helicopters; and evaluated information that led to the discovery of geologic deposits worth hundreds of millions of dollars. An expert system even helped a race car driver break the speed record at the Indianapolis 500. (There are several good sources on expert systems. One is "The Prentice Hall Guide to Expert Systems.")
However, despite their potential for solving problems, expert systems are still little-used in many areas of business, including market research.
My own experience includes designing and implementing an expert system for analyzing customer satisfaction data as a consultant to large-scale international companies. Recently, as the director of customer satisfaction for the Gilmore Research Group in Seattle, I have developed the core expert system capabilities that can be applied across any customer satisfaction data set.
Expert systems hold tremendous potential for rapidly analyzing vast streams of customer satisfaction data and generating truly actionable reporting. Before I tell you more on this score, let's first take a look at what characterizes an expert system.
Expert system methodology
The purpose of the expert system is to model an expert's problem-solving strategies. The domain expert, by definition, is a knowledgeable person with a reputation for effective solutions in a particular field. This expert status includes the ability to arrive at solutions efficiently. The key to an expert system is the accumulation and codification of the expert's knowledge.
The expert's knowledge is translated into a series of if-then-else statements for the computer to process. To demonstrate, a simple weather forecasting model can be built. Let's say the simplest way to predict the weather for tomorrow is by saying it will be the same as today's. Our if-then-else statement would be:
- IF the weather is sunny today, THEN it will be sunny tomorrow, or ELSE it will be rainy.
This statement alone does not lead to very accurate weather forecasting. Thus, we need to add further rules: - IF the TV weather person says it will be sunny, THEN it will be sunny, or ELSE it will be rainy.
- IF the season is summer, THEN it will be sunny, or ELSE it will be rainy.
Or maybe . . .
- IF a picnic is planned, THEN it will be rainy, or ELSE it will be sunny.
Many other variables such as barometric pressure, wind direction, weather in adjoining states and views from satellites could also be incorporated. As more knowledge is added, our weather-forecasting model becomes progressively more accurate.
Of course, this is a simplified example of a much more complex task. It does, however, illustrate the basic building block of the expert system: the if-then-else statement, also known as a decision rule. At their core, expert systems are nothing more than an expert's knowledge formalized in a computer as a series of decision rules.
Now let's see how an expert system can greatly improve data analysis and reporting in customer satisfaction programs.
Applying an expert system to customer satisfaction data
Just as in our weather forecasting model, the first step in applying an expert system to customer satisfaction data is to establish the decision rules - only this time you and the research sponsor are the experts.
What are the appropriate analytic comparisons? This will dictate the decision rules. Let's say you are a large company that sells business-to-business products through 1,000 national outlets. There are a number of performance comparisons that could be made. For example, you could compare:
- an outlet to other outlets in their territory, region or nationally;
- an outlet to same-sized outlets;
- outlets to a standard of performance; or
- outlets to pre-determined improvement goals.
While on the surface this may appear to be ordinary analysis and reporting - the comparison of composite factors at an overall satisfaction level - what differentiates the expert system is its ability to also make these comparisons or any combination of comparisons at a detailed question level for each outlet. These more detailed comparisons are especially significant when customers' importance rankings of questions are applied.
There are other analysis features an expert system can employ which are typically not performed with traditional customer satisfaction analysis because the amount of data is overwhelming. Perhaps it would be beneficial for your sponsor to know changes and trends. The expert system can easily perform these types of analyses and generate clear, easy-to-use reports tailored by outlet.
Once you've established the decision rules, you're ready to analyze data. The chart below illustrates an expert system applied to customer satisfaction data. In our example, you've collected 50,000 surveys from customers representing your 1,000 outlets. All have purchased and used various products for up to a year. The survey assesses their satisfaction with product and service features.
Statistical analysis (regression) is applied in order to rank importance factors by product or service feature, each relating to a specific question on the survey. The importance factors are fed into the expert system, along with the decision rules, and then applied to outlet-specific data. The output is a prioritized list of targeted improvement areas by outlet.
Prioritize improvements
Remember the old reports? Too much data - hard to use. Oversimplified graphics - don't tell enough. Reports generated by the expert system solve this dilemma with one simple page.
The new reports prioritize the areas that need improvement based on the importance factors established through statistical analysis of customer responses. No longer are there multiple pages of numeric data, just one page with the prioritized targeted improvement areas.
In our example, the list would be by specific product or service feature and would reference the exact question or questions on the survey. Reporting can include other valuable information for the outlet such as whether listed items reoccur over time and what percentage of customers would buy again.
What is most important? What needs to be worked on? There is no mistake. The expert system enables you to rapidly analyze the customer satisfaction data of each individual outlet and generate a report that addresses each individual outlet's customer satisfaction issues.
Linked to improvements
If the goal of customer satisfaction programs is to raise customer satisfaction in order to increase profits, measurement needs to be linked to driving improvements. An expert system can help you achieve this goal by overcoming the limitations inherent in traditional methods for analyzing, reporting and effectively using customer satisfaction data.