CustomerSat Insights (CI) interview with Professor Neil A. Morgan, Kelley School of Business, Indiana University.

CI: Professor Morgan, please tell us about your work, including your recent study published in the September-October issue of Marketing Science.

NM: For a long time I’ve been interested in how customer satisfaction and loyalty affect business performance. My colleague, Lopo Rego at the University of Iowa, and I looked at which business performance measures were best predicted by which customer measures, both attitudinal and behavioral. To do this, we used databases from the Center for Research in Security Prices (CRSP), the American Customer Satisfaction Index (ACSI), and CompuStat. We examined data on 80 different firms across dozens of business-to-consumer (B2C) industries over seven years.

CI: Why is it important to determine which customer measures best predict business performance?

NM: By understanding then improving the measures that best predict performance, companies get the most "bang for their buck." That makes these metrics the most important for management to track. Investors should also consider these measures when evaluating companies.

The business performance measures we examined were:

• Annual sales growth
• Total shareholder return
• Market share
• Cash flow
• Gross margin
• Tobin’s Q (a measure of a firm’s intangible-to-tangible assets)

The customer measures were a variety of attitudes and behaviors:

• Satisfaction
• Likelihood-of-repurchase
• Actual complaints
• Other word-of-mouth (WOM) measures

CI: What did you find?

NM: In a nutshell, we found that:

• Overall satisfaction is a very robust predictor of business performance – the best predictor among the customer measures we examined
• Likelihood-of-repurchase and number of complaints also have some predictive value
• Measures that exclude data have reduced predictability

CI: First tell us about the satisfaction measures. How were these defined?

NM: The best predictor of all was mean overall satisfaction, which we use to refer to the arithmetic mean of overall satisfaction, satisfaction relative to expectations, and satisfaction relative to ideals. Overall satisfaction is a robust predictor across all six measures of business performance. We were surprised that a single metric would be as predictive of so many facets of business performance as satisfaction is.

The other satisfaction-based measure we looked at, top-2 box percentage on overall satisfaction (5-point scale), is also a good predictor overall, though a poor predictor of total shareholder return. Top-2 box scores ignore all customers who rated satisfaction at a 1, 2 or 3. They also fail to differentiate between those who rate their satisfaction as a 4 or a 5. Discarding data like that reduces precision and predictive ability.

CI: After the satisfaction measures, what were the next-best predictors?

NM: Next was likelihood-of-repurchase. This attitudinal loyalty measure significantly predicts a firm’s future market share, gross margin, sales growth and Tobin’s Q performance, but not its total shareholder returns or net operating cash-flow performance.

After that, the next best predictor was the proportion of customers who reported having complained, either formally or informally. We were somewhat surprised that the number of complaints had a reasonably high (inverse) correlation with customer sat scores. However, complaints are not as good or stable a predictor as satisfaction. Complaints are predictive of annual sales growth and Tobin’s Q, but have little or no predictive value for market share. Neither likelihood-of-repurchase nor WOM measures are as robust or generalizable as satisfaction in predicting business performance.

CI: Why is that?

NM: Overall satisfaction directly depends on, and only on, how well the business is serving the customer. It is 100% knowable by the customer. It is the most accurate reflection of how a customer perceives the business at that point in time. In contrast, likelihood-of-repurchase, willingness-to-recommend, actual recommendation and complaint behavior are affected by issues outside of business’ and customers’ control.

For example, customers may be dissatisfied but will still recommend or believe themselves likely to repurchase, if no better alternative is available. But this kind of loyalty is short lived. It lasts only until better alternatives come along. Alternatively, for any number of reasons, a customer’s intention to recommend may not translate into an actual recommendation, and these recommendations may or may not be acted upon.

Most loyalty models assume that attitudinal loyalty (likelihood-of-repurchase, willingness-to-recommend) translates into behavioral loyalty (actual repurchase, actual recommendations) that translates into supplier’s business performance (sales, margins, stock value, etc.) But our results show that these linkages are weaker than satisfaction for predicting performance.

CI: So should a customer loyalty index (CLI) be based solely on satisfaction? Or can its predictive ability be improved by incorporating other measures, such as likelihood-of-repurchase or willingness-to-recommend?

NM: Design your CLI around the measures that best predict desired outcomes for your business. If you have data on various customer measures and performance outcomes, regression or similar modeling techniques will let you determine the combination of measures that will best predict outcomes. Avoid using satisfaction and loyalty measures simultaneously if they are highly correlated with each other as independent variables in the model. That can lead to statistical problems like multicollinearity. Despite those difficulties, multi-item measures – combinations of variables – are generally more reliable and valid, and offer more precision for a given sample size, than single measures alone.

CI: What about Net Promoter Score (NPS)?

NM: NPS is defined as top-2 box (9 or 10) percentage of willingness-to-recommend minus bottom-6 box (1 through 6) percentage on a 10-point scale (or bottom-7 box, 0 through 6, on an 11-point scale). As I mentioned earlier, it is a bad idea to discard data. NPS does not differentiate between scores as divergent as 0 or 1 and 6. It ignores 7 and 8, and doesn’t distinguish between 9 and 10. These factors reduce the effectiveness of NPS both as a predictor of business performance and as a trend indicator. For example, a company’s NPS can steadily rise while its overall performance of willingness-to-recommend, as measured by its mean score, declines. Such erratic behavior can happen when you discard data, as NPS does.

Because NPS discards data, trending it can be misleading. Here, over three quarters, NPS rises from 25% to 27%. Meanwhile, willingness-to-recommend mean score, which does not discard data, falls from 8.22 to 7.50.

CI: Any suggestions for using NPS?

NM: Ask willingness-to-recommend, calculate NPS – and keep and use valuable data. Track mean willingness-to-recommend and use it in conjunction with overall satisfaction and detailed attribute performance satisfaction. All of these are trendable and, together, actionable. Mean scores don’t discard responses.

CI: What are the time lags between changes in customer measures and their impact on business performance? How soon will they see results?

NM: We were a bit surprised with the speed at which this happens. In our data the strongest impacts of customer measures come within four quarters. There are steady effects from one to three quarters, depending upon the business performance metric. After three quarters, the effects level off and decline.

Averaged across the B2C industries we looked at, the effects on market share, gross margin, and sales growth appear after just one quarter. Effects on cash flow and Total Shareholders Returns (TSR) take around two quarters. The longest one – Tobin’s Q – takes around three quarters.

As I mentioned before, all the industries we looked at were business-to-consumer (B2C). We have no reason to believe the results would be different for business-to-business (B2B), but we have not yet studied B2B in detail. My sense is that the time lags will be somewhat longer for B2B products and services than for more frequently-purchased B2C products and services.

CI: You also looked at what makes customer feedback most valuable to businesses. What did you find?

NM: To answer this question, we conducted a multi-informant survey with managers in different functional areas of over a hundred utilities companies across the US. We found that disseminating the data throughout the organization, and making it actionable before disseminating it, are keys to making customer feedback valuable. Organizations need to ensure that recipients of customer feedback data:

• Believe the data is accurate
• Think it is relevant to what they do
• Can use the data to diagnose what they should do differently in future.

CI: Neil, thank you very much. How can interested readers get more information?

NM: See the recent articles, Neil A. Morgan and Lopo L. Rego (2006), "The Value of Different Customer Satisfaction and Loyalty Metrics in Predicting Business Performance," Marketing Science, 25 (5), 426-439; and, Neil A. Morgan, Eugene W. Anderson, and Vikas Mittal (2005), "Understanding Firms’ Customer Satisfaction Information Usage," Journal of Marketing, 69 (3), 131-151. Or send me an e-mail at namorgan@indiana.edu.