Measuring Customer Satisfaction: NPS, CSAT, and Customer Effort Compared

Which number shows whether customers stay?

Anyone who surveys customers soon faces a choice: which question becomes the metric that is compared every year? Consultants and software vendors recommend the Net Promoter Score one time, the classic satisfaction score another time, customer effort a third time. Each recommendation comes with a study to support it. For an owner who asks only once a year and receives a few dozen answers, this is confusing.

That raises one question: which customer satisfaction metric tells a small business most reliably something about future behavior, and how many of them does it need? To answer it, we first clarify the three metrics, then sort them by their time focus with a simple mental model, look at what studies show about their predictive power, weigh two perspectives, and finally arrive at a recommendation.

How a survey is structured overall, which topics belong in it, and how answers become actions is described in the guide B2B customer surveys. This article goes one step further and asks only about the one number that sums up the state of customer relationships.

A customer satisfaction metric is worth only as much as the behavior it predicts.

The three metrics: what NPS, CSAT, and CES measure

The Net Promoter Score (NPS) is based on a single question: “How likely is it that you would recommend us to a colleague?” on a scale from 0 to 10. Those who give 9 or 10 count as promoters, those who give 0 to 6 as detractors. The score is the share of promoters minus the share of detractors. The approach goes back to Frederick Reichheld (2003) and is widely used in companies.

The satisfaction score, or Customer Satisfaction Score (CSAT), asks for an overall verdict: “How satisfied are you overall with our work together?”, usually on a scale from 1 to 5. It is analyzed in two ways: as the average of all answers or as the share of top ratings, known as top-2-box. This is the share of customers who choose one of the two highest points on a five-point scale (Morgan and Rego 2006).

Customer effort, or Customer Effort Score (CES), does not ask for an overall verdict but about a single contact: how much effort did the customer have to make to get their request handled? The metric comes from Dixon, Freeman, and Toman (2010), who surveyed customers after a call to customer service or after using a self-service channel. It is therefore tied to an occasion, such as a complaint, a question about an invoice, or a change of appointment.

The mental model: looking back, at the contact, and ahead

De Haan, Verhoef, and Wiesel (2015) classify customer feedback metrics by their time focus: do they look at the past, at the present moment, or at the future? From this, a simple mental model can be derived, which I call the Three-Views Grid here. It does not ask which metric is best but which direction of view a business needs right now.

1. Looking back: the satisfaction score

   The customer judges the relationship so far. The number sums up many experiences and is therefore comparatively stable. It tells you how things have gone so far.

2. The contact: customer effort

   The customer judges a single transaction. The number is close to the process and shows where a workflow gets stuck. It says little about the relationship as a whole.

3. Looking ahead: the Net Promoter Score

   The customer states an intention for the future. Intentions are less certain than judgments about experience, because a lot can happen between intention and action.

The grid has two practical consequences. First, the three metrics answer different questions and therefore compete less than the debate suggests. Second, customer effort does not fit into an annual survey of all customers, because many customers had no reason to get in touch during the year. Wiesel, Verhoef, and de Haan (2012) point out explicitly that the metric applies only to customers who had contact with the company after the purchase.

What studies show about predictive power

By their own account, Morgan and Rego (2006) were the first to compare systematically which common metric best predicts later business performance. They used the American Customer Satisfaction Index, for which more than 50,000 US consumers rate the products of over 200 large companies each year. After excluding utilities and privately held firms, 80 companies with 569 observations from the years 1994 to 2000 remained. They linked these to six performance measures, including market value, operating cash flow, and total shareholder return.

The result was clear. Average satisfaction had the greatest predictive value, and the share of top ratings also had good predictive value. Repurchase intention and the share of customers complaining predicted something depending on the performance measure. Metrics based on recommendation intention, meaning the net promoter approach, and the average number of actual recommendations had little or no predictive value. The authors conclude that it is a mistake to focus feedback systems solely on recommendation.

De Haan, Verhoef, and Wiesel (2015) examined the question at the level of individual customers and with a different goal: does the metric predict whether a customer stays? Their data cover customers of 93 firms in 18 industries. The share of top ratings on the satisfaction score performed best, and metrics that focus on the extremes were better than those averaging the whole scale. All metrics except customer effort were significantly related to retention. At the same time, the best metric differed by industry, and a combination of several metrics improved the prediction further.

The opposing position comes from Dixon, Freeman, and Toman (2010). In a study of more than 75,000 people who had called a customer service center or used self-service channels, they found that special efforts to delight made little difference. Customers mainly wanted a simple, quick solution. According to the authors, customer effort predicted loyalty better than satisfaction or the Net Promoter Score.

Two perspectives: delight or effortlessness?

Behind the dispute over metrics are two ideas of how loyalty arises. The first says: customers stay when they are highly satisfied. Those who merely do decent work get middling scores and a customer who switches at the next opportunity. This is supported by the finding of de Haan and colleagues that the extremes reveal more than the average. It is also supported by the finding in the article increasing revenue per customer that share of wallet rises clearly only at very high satisfaction.

The second idea says: customers leave when things get tedious. Delight is rarely the reason to stay, annoyance often the reason to go. This is supported by the findings of Dixon and colleagues and by the study of lost business customers in the article why customers leave, in which neglect by the supplier was the most frequent reason for leaving.

I consider both ideas correct, but for different places in the business. The relationship as a whole is better measured by the satisfaction score, because it reaches all customers and predicted retention more reliably in the studies. The processes through which customers come into contact with the business are better measured by customer effort, because it is concrete and timely. According to the available findings, the Net Promoter Score is not a bad number, but not one that carries the weight on its own.

The recommendation for small businesses: a basic set of three questions

From the findings, a basic set can be derived that requires little effort and compensates for the weaknesses of the individual metrics. It consists of a lead metric, a supplement, and an occasion-based question.

1. Lead metric: share of top ratings

   In the annual survey, ask the satisfaction question on a five-point scale and track above all the share of answers with 4 or 5. It is easy to explain and, according to de Haan and colleagues, the best single indicator of retention.

2. Supplement: the recommendation question

   Ask the recommendation question as well and use it mainly for conversations: ask promoters for a referral, and ask detractors what you should do better.

3. Occasion question: customer effort

   After every closed complaint or major query, ask how easy it was to get it resolved. The answers belong to the review of processes, not to the annual metric.

4. Analysis by extremes

   Look at the edges: how many customers give top ratings, how many the worst? An average of 4.1 hides whether three important customers answer with 2.

The example shows a peculiarity of small businesses: the five customers with poor ratings matter more than the percentage. The conversation starts with them. Complaints, which often trigger such answers, are covered in the guide handling complaints.

What follows beyond the metric

The comparison holds a lesson that goes beyond the choice of question. None of the three metrics measures what a business actually wants to know, namely whether the customer stays and buys more. They are indicators that precede behavior. Their value only shows when you put them next to behavior: do customers with top ratings actually buy again the following year? Anyone who checks this once with their own numbers knows more about their market than any study.

Finally, the metric is only the beginning. Public reviews, complaints, and silent defection are further signals that may point in the same direction or may not. How online reviews are put into context is described in the guide online reviews. If you want to build the metric into a fixed monthly or quarterly review, you will find the structure in the guide a one-page KPI dashboard.

How to get started

• Decide which question becomes your lead metric and write down the exact wording and scale.
• Check whether you can analyze satisfaction from your last survey as a share of top ratings.
• Add the recommendation question as a supplement and decide what happens with promoters and detractors.
• Set up a single effort question after every closed complaint.
• Next year, compare whether customers with top ratings actually bought again.

Conclusion: which metric a small business needs

To answer the guiding question: for the customer relationship as a whole, according to the available studies, the satisfaction score tells you most reliably something about future behavior, especially as a share of top ratings. Morgan and Rego (2006) found hardly any predictive value for recommendation-based metrics, while de Haan and colleagues (2015) saw the share of highly satisfied customers in front, but with differences by industry. Customer effort is not a substitute for the annual metric but a tool for individual processes.

A small business therefore does not need the one best number but a fixed basic set: satisfaction as the lead metric, recommendation as a supplement for conversations, effort after contacts. The limits are clear, because the studies come mainly from consumer markets, customer service centers, and large companies. That makes it all the more important to check whether the number predicts the behavior of your own customers. A metric that does not do so is an opinion poll.