ONE GOOD IDEA
Decipher what the customer is telling you by using control charts
by Tony Gojanovic
More often than not, a company’s phone number and internet address appear on its product to provide consumers an outlet for expressing their opinions and experiences. Unfortunately, customer data obtained through these outlets are typically poor measures of performance.
Factors such as the price of the product, market, type and severity of a defect or consumer preferences and idiosyncrasies result in highly variable results. A torn package for a toy may not warrant a response from the vast majority of consumers and, hence, will go unreported. Some consumers will respond negatively to a product, even though there is no reason to do so, perhaps enticed by a discount coupon for future purchases.
Even though there are many shortcomings associated with using consumer data, with proper analysis, detective work and integration of other sources of information, consumer response data can be translated into a viable story of what the consumer is truly experiencing.
One of the greatest challenges with consumer data is to make sense of it. A company may get thousands upon thousands of phone calls or e-mails each year from consumers. Some companies have complex systems that classify different consumer contacts. A manufacturer of cartons for a consumer product might have categories related to defective packaging, with several subcategories.
Typical summarization schemes usually take the form of standalone statistics, such as a percentage increase or decrease in consumer complaint levels. For instance, a report might state that consumer dissatisfaction for torn handles has increased 50% over the previous month. Such summaries are largely useless because they have no context of variability or indication of what the underlying numbers are.
Chart a course
A particularly useful way to summarize consumer call-in information is the control chart, which can transform data into a time-series graph, along with an estimate of noise. Control charts, largely used in manufacturing, increasingly are finding their way into nontraditional areas as a way to make sense of data. When shown in relation to other pieces of solid information, control charts provide a formidable analysis tool for gauging consumer satisfaction.
A control chart used to summarize consumer complaint data for a packaging company can be found in Online Figure 1. A chart can be easily created in Minitab or Excel using simple control chart formulas found in many statistical textbooks.
In some businesses in which seasonality is a concern or monthly production numbers vary, indexing consumer response helps minimize the effect of seasonal variation. In the packaging company example, the index is the number of complaints per 100,000 packages for a consumer product. Because consumer data is count data, indexing is helpful in creating a more symmetrical data set, which is essential to an individual control chart, especially if accurate inferences are to be made.
In the example, data is plotted for 36 months. Looking at information in the context of history negates the effect of short-term analysis and thinking. Comparing this month’s number to last month’s number without the context of noise will lead to knee-jerk reactions, finger pointing and wasted effort.
The packaging company notices a drop in consumer complaints in the middle of the chart. Standard control chart analysis reveals this pattern as a process shift. Because correlation does not always imply causality, performing additional detective work and integrating additional pieces of information create a reliable story of what the consumer is experiencing.
An investigation of the shift reveals that it coincided with the introduction of a new supplier that provides paper cartons with enhanced performance characteristics. The company also notices the process shift hasn’t been sustained, which would require further analysis to determine the cause. A similar chart can be created for rare-event occurrences, such as safety incidents, based on time between events.
Consumer data can be an effective component in a continuous improvement process when properly applied. By integrating that data and other sources of evidence, a control chart can deliver instant benefits, form a credible picture of what the consumer is experiencing and help untangle many of the complexities associated with field data.
- Juran, Joseph M., Frank M. Gryna and R.S. Bingham, Quality Control Handbook, third edition, McGraw-Hill Book Company, 1974.
- Montgomery, Douglas C., Introduction to Statistical Quality Control, second edition, John Wiley & Sons, 1991.
- Wheeler, Donald J., "Shewhart’s Charts: Myths, Facts, and Competitors," ASQ Quality Congress Transactions, 1991.
- Wheeler, Donald J., Understanding Variation: The Key to Managing Chaos, SPC Press, 1993.
Tony Gojanovic is a statistician with MillerCoors in Golden, CO. He earned a master’s degree in mathematics and statistics from the University of Colorado in Denver. Gojanovic is a member of ASQ.