Loyalty, Satisfaction and Statistics

This topic has been in my head for a long time but I have not been willing to write about it. It concerns an issue about satisfaction and loyalty I have been struggling with even before Fred Reichheld came up with NPS.

In fact, the issue was one of the reasons I decided to write this blog. It relates to the supposed statistical relationship between satisfaction and loyalty.
If you have been following this blog you know I perceive a relationship between customer satisfaction and customer loyalty. I’ve discussed ways of building customer loyalty, tracking customer loyalty, five stages of customer bonding, the notion of treating employees as customers, keeping employee loyalty during downsizing and restructuring, customer partnerships, and the impact of customer corporate culture on loyalty.

I do believe there is a real relationship between customer satisfaction and customer loyalty. So what’s the problem?

I do not believe there is a simple statistical relationship between customer satisfaction and customer loyalty. I say this even though I have been guilty of acting as if there were.

Let me start at the beginning. In my consulting activities, I have helped customers develop satisfaction measurement systems. This activity has included assistance in questionnaire design and analysis of the responses from the customers. I have been asked more than once to include a question on the survey to measure customer loyalty.
As I have given the idea of a single question more and more thought, I have grave misgivings about the value of such a measurement.

The questions I have used on several surveys to provide these measurements are, “would you purchase from this company again,” and, would you recommend this company to an associate?”

The first question, “would you purchase from this company again,” is a yes or no question. On the surface this question would appear to indicate a loyalty level. The respondent is asked to make a decision regarding his next purchase which is, of course, one of the most direct definitions of loyalty.

Some of my concerns regarding this question are:
• Is the person responding to the questionnaire really the person who will make the purchase?
• Because the person says they will purchase again from the company now does not guarantee they will purchase again when the time comes, even if they are responsible for the purchase.
• To my knowledge, no one has ever gone back to the respondents and verified they did what they said they would do. The researchers at Sat Metrix have been doing some validation of this and are now publishing the results. I may soon be able to eliminate this concern.
• The person responding that he would purchase from the company again may not see any alternatives at the time the question is being asked, and so may have no other possible response.
• There is no allowance in the response to the question for equivocation. The respondent must either answer yes (as in absolutely yes) or no (as in absolutely no) or leave the question unanswered.

Since the question is being asked, of a current customer (who comes with a positive bias by virtue of being a customer), the limitation of the possible responses to yes, no or blank would appear to bias the response to the positive.

The second question, “would you recommend this company to an associate,” provides a little more granularity in the response since I have used the scale of “definitely yes”, “probably yes”, “probably no” and “definitely no.” From this scale I have used the response of “definitely yes” as a measure of a very satisfied and loyal customer.

I have used the logic that a person in business will not definitely recommend a company to an associate unless that person really believes in the company (by inference is loyal) since a bad recommendation to an associate would reflect back on the person making the recommendation. Some of the issues noted in the preceding paragraphs also apply to this question; e.g., is the person making the recommendation involved in purchasing product or services, thus implying loyalty by the company. Just because an individual is loyal to a particular vendor does not necessarily assure the company has the same level of loyalty.

The notion statistical relationships between two or more variables imply cause and effect is probably one of the most flagrant violations of statistical inference.
My favorite example, and one I use in my statistics class, is the statistical relationship between the number of drownings at the beach and the number of ice cream cones purchased. As the number of ice cream cones purchased increases at the beach, so do the number of drownings that occur.

There is a high statistical correlation between these two variables, and one can develop a regression line forecasting the number of drownings based on the number of ice cream cones purchased. If drownings are the result of ice cream cone sales at the beach, it would seem that the simple answer is to eliminate all ice cream cones sales so that there would be no more drownings. The fact is: ice cream cones do not cause drownings. The number of drownings increase as the number of people on the beach increase. That occurs on hot days which, by the way, leads to increased sales of ice cream cones.

I use the previous discussion of ice cream cones and drownings to point out statistical relationships do not guarantee cause and effect. In the case of the two variables satisfaction and loyalty, there is often a statistical relationship between the two. However, there is not a clear cause and effect relationship. If there was a strong cause-and-effect relationship between satisfaction and loyalty, customer loyalty could be assured with superior customer satisfaction programs.
Automobile dealers and manufacturers are finding high levels of customer satisfaction do not guarantee customer loyalty.

Satisfaction levels above 90 percent reported by J.D. Power for automobile manufacturers do not correspond to loyalty ratings of less than 40 percent. Hence, there must be other factors which have as much or more impact on loyalty.
In high technology businesses, loyalty can be as fleeting as the technology itself. A review of the PC industry would indicate customers are as sensitive to price and technology as they are to customer satisfaction (at least at the time of purchase). Often, customers of PCs later realize they may not have put enough emphasis on customer service at the time of purchase.

I recently purchased a new PC for my wife. I was most influenced in the decision by the price/performance relation ship of the PC (I bought the most PC I could get for my money). There was no loyalty in my decision even though I have been loyal to that specific manufacturer in the past. I was not dissatisfied with the performance of the old computer or the service I had received from that company. I just decided to buy on the basis of price-performance. I probably would have answered my questions positively (I would buy from the company again and I would definitely recommend them) but I was not loyal.

The objective of this blog has been to offer the suggestionand some logic that the measurement of customer loyalty is not simple. The simple one-question measure is fraught with ambiguity, especially with respect to the respondent (who the person is, what impact the person has on company loyalty, and how consistent that person will be when the time comes to make the purchase decision).

While there may be a statistical relationship between customer satisfaction and loyalty, that relationship does not guarantee cause and effect between them.
I will close with the conclusion based on my study of this topic from an academic perspective and my experience as a practicing service executive — customer loyalty cannot easily be acquired without a strong customer satisfaction program but customer satisfaction by itself will not assure customer loyalty.