By CustomerSat Research & Consulting Services

How can you best put NPS, the popular loyalty index, to work in your enterprise? What’s the best way to roll up NPS results across touchpoints, departments and divisions? What are pitfalls to watch out for? In this article, members of CustomerSat Research & Consulting Services discuss how to use NPS to help improve business processes and drive customer-centric action in your organization.

What is NPS?

NPS is traditionally based on your customers' answers to the question, “How likely are you to recommend us to others?” For 0-10 or 1-10 scale, customers giving you a score of 0–6 or 1–6 are labeled Detractors, 7–8 are Passively Satisfied, and 9–10 are Promoters. To calculate NPS, simply subtract the percentage of Detractors from the percentage of Promoters.

% Promoters – % Detractors = NPS

Putting it another way, NPS is a top-2 box percentage minus a bottom-6 box percentage (on a 1-10 scale) or bottom-7 box percentage (on a 0-10 scale). As we will see below, the same calculation can be done for other rating questions as well, to help put NPS in context.

Does Scale Matter?

NPS was originally based on a 0–10 scale, but the difference in NPS between 0-10 and 1-10 scales is small and usually negligible, since a small percentage of responses typically fall into the 0 and 1 boxes.

For organizations using 5-point scales, we generally recommend calculating NPS as top-1 box % minus bottom-3 box %. On a 5-point scale, ratings of 4 are considered Passively Satisfied, and are thus excluded from the NPS calculation, as are ratings of 7 and 8 on 0-10 and 1-10 scales.

For companies using a 4-point verbal scale (Definitely Will Not, Probably Will Not, Probably Will and Definitely Will), NPS is usually defined as percentage of Definitely Will minus the total percentage of Definitely Will Not and Probably Will Not.

There are many variations on NPS. One of our clients defines it as top-2 box minus bottom-2 box. Another refers to NPS as a Net Customer Allegiance Score (NCAS).

At what levels in the organization is it meaningful to measure and track NPS?

• At global levels—worldwide enterprises and major business units—NPS can be used by senior management to rally the entire organization toward greater customer-centricity. Its simple math can be understood by everyone. This is where NPS has the greatest value, in our view.
• At intermediate levels—divisions, departments, and touchpoints—NPS should be coupled with other metrics to ensure actionability and accountability, as discussed below.
• At individual levels — such as individual products and customer service reps — we recommend using overall and detailed satisfaction metrics, over which the individual product manager or CSR has more direct influence and consequently can be better held accountable, rather than NPS based on likelihood to recommend. Of course, every loyalty index requires an adequate sample, an issue which deserves particular attention at individual levels.

How can my company make NPS most actionable?

For greatest actionability, NPS should be used in conjunction with any or all of the following:

1. Correlations between detailed performance attributes and overall business outcomes. These enable you to derive the importance of attributes, and thus better prioritize quality improvements by attribute.
2. Attribute scores relative to performance benchmarks, such as those provided by the CustomerSat Benchmark Program.
3. Cross-tabs that let you gauge attribute performance by business segment, such as product, geography, and vertical market.
4. Open-ended comments, in the customers’ own words, offering suggestions for improvement.

Some advocate using NPS with the fourth item alone. We recommend using all four.

Can likelihood to recommend be used as the overall business outcome in a correlation analysis to derive the importance of performance attributes?

We don’t recommend it. The three most-commonly used overall business outcomes are:

• Likelihood to recommend
• Likelihood to repurchase
• Overall satisfaction (OSAT)

Of those, overall satisfaction is the one most directly dependent on the quality of a client’s product and service delivery alone. The customer behaviors we are ultimately most interested in, likelihood to recommend and likelihood to repurchase, introduce factors unrelated to product and service delivery. These factors may be completely out of your control. Consider these scenarios:

• Scenario 1 - If no better alternatives are available, a customer might recommend your company - even if they’re dissatisfied with the quality of your products and services.
• Scenario 2 - Maybe it’s not their personal style to recommend, or their company’s policies prohibit them from making a recommendation. In these cases, customers may not be willing to recommend even if very satisfied.

These factors unrelated to product and service performance introduce “noise” in the correlations and in the calculation of the importance of the attributes, and can lead to sub-optimal prioritization of performance improvements, as the table below demonstrates. The first column lists detailed customer service performance attributes; the second shows each attribute’s correlation with overall satisfaction; and the third shows each attribute’s correlation with likelihood to recommend. The correlations with OSAT are consistently stronger than those with likelihood to recommend. In fact, the weakest correlation with OSAT is stronger than the strongest correlation with likelihood to recommend (other than OSAT itself).

Using Correlation to Derive Importance of Performance Attributes
Performance Attributes	Correlation with Overall Satisfaction (OSAT)	Correlation with Likelihood to Recommend
Overall satisfaction	1	0.753
Escalation	0.773	0.492
Support rep overall	0.765	0.467
Technical ability	0.728	0.436
Answer quality	0.720	0.433
Answer timeliness	0.698	0.443
Follow-up/Closure	0.689	0.427
Keeping informed/status	0.682	0.466
Initial response timeliness	0.617	0.418
Courtesy-Attitude	0.545	0.378
Accessibility	0.516	0.444
Average correlation coefficient	0.614	0.403

Note that the order of importance of the attributes when based on correlations with likelihood to recommend is different from the order when based on correlations with OSAT. That’s not surprising, given the “noisier” measurement of importance using likelihood to recommend as compared to using OSAT. Prioritizing attributes based on noisy correlations can lead to less-than-optimal results compared to stronger correlations. Consequently, we recommend:

1. Using OSAT as the business outcome (dependent variable) for deriving importance.
2. Always asking OSAT in surveys in addition to likelihood to recommend.

How long should a loyalty questionnaire be? What about the 2-question survey?

Some NPS proponents insist that surveys need only one or two questions: likelihood to recommend and an open-ended, follow-up question asking, "Why?"

Surveys with only two questions usually enjoy high response and completion rates. Unfortunately, the responses convey little actionable intelligence. To gain an understanding of the respondents’ true feelings and experiences requires costly follow-up by phone in virtually every case. We strongly recommend including multiple diagnostic questions – ratings of detailed performance attributes – in the survey. These additional responses convey far more useful, actionable intelligence, vastly reducing the number of follow-up calls required and saving time and expense.

Of the few of our clients who initially ask only two questions, most subsequently modify their surveys to add supplemental questions. In contrast, the majority of our clients start out trying to incorporate too many questions into the survey, which we recommend reducing. An organization that enjoys strong affinity with its customers can take the liberty of a slightly longer questionnaire. CustomerSat Research & Consulting Services can suggest many best-practices to help build response rates for a survey over several quarters or years (such as rigorous communication back to customers of findings, planned actions, and improvements made as a result of their feedback). In short, we help clients design the optimal NPS/feedback solution for their businesses, weighing a variety of factors they may not have even considered.

How Can We Compare NPS with Other Performance Metrics?

To provide context for NPS, do the NPS calculation for multiple performance metrics, not just willingness to recommend. A mean score for a particular performance metric is meaningless without knowing how it compares to other mean scores. The same is true for NPS. So consider calculating NPS for several or all of the rating questions in your survey. When NPS calculations (top 2-box percentage minus bottom-6 box percentage) are performed on various rating questions, we call them Net Percentage Scores (rather than Promoter). Two CustomerSat Enterprise analytics – Configurable Rating Scores and Comparative Statistics – let you do this easily.

Below, we compare Net Percentage Scores for three questions: OSAT, likelihood of re-purchase, and likelihood to recommend. In Scenario 1, “NPS_Recommend” is higher than "NPS_OSAT" and "NPS_Re-purchase", indicating that customers are more willing to recommend than they are satisfied (there may be no attractive alternatives). Scenario 2 is the opposite, indicating that customers are more satisfied than likely to recommend (corporate policy may prohibit B2B customers from recommending). In Scenario 3, the three scores are the same. In all three scenarios, comparing Net Percentage Scores helps put “NPS_Recommend” into context.

	Net Percentage Scores (Top 2 box % minus Bottom 6 box %)
Question	Scenario 1	Scenario 2	Scenario 3
Overall satisfaction (OSAT)	15%	25%	20%
Likelihood of re-purchase	20%	20%	20%
Likelihood to recommend	25%	15%	20%

Is there a simple way to graph Net Percentage Scores?

In CustomerSat Enterprise, graphing all flavors of NPS is easy. You can calculate and/or graph NPS — and a wide variety of other statistics — using the Range % and Range % Gap features. You can also trend NPS over weeks, months, or quarters for different segments of your business. The graph below, for example, shows NPS by business unit/division.

Does NPS contain any potential pitfalls?

Definitely. As the example below indicates, it is quite possible for NPS to move in one direction while the mean score of the underlying performance metric moves in the other direction. Such erratic behavior can arise because NPS discards data – it combines scores as disparate as 0 and 6 and ignores scores of 7 and 8. The CustomerSat team can advise you on how best to anticipate and manage such pitfalls.

In Conclusion...

NPS based on likelihood to recommend can be a valuable tool for rallying an organization to greater customer-centricity. It uses simple math that everyone can understand, enhancing its appeal. But as with any single statistic, it provides a very limited view of your business. For comprehensive and more actionable views, always use NPS in conjunction with other performance metrics, in particular detailed and overall measures of satisfaction. We call this NPS Plus™.

For details, and to discuss further, please contact your CustomerSat representative, email us at expert@CustomerSat.com, or call us at 650.237.3300.

NPS Plus is a trademark of CustomerSat, Inc.