Measure satisfaction: NPS and SERVQUAL

“How are we doing?” is the most common question in marketing research. Two widely used measures answer it from different angles: the Net Promoter Score summarises customer loyalty in one number, and SERVQUAL breaks service quality into dimensions and compares what customers expected with what they got.

The Net Promoter Score

NPS is based on one question: “How likely are you to recommend [company] to a friend or colleague?”, answered from 0 (not at all likely) to 10 (extremely likely). Respondents are grouped into promoters (9–10), passives (7–8) and detractors (0–6), and NPS = % promoters − % detractors, from −100 to +100.

In the example, 36.6% are promoters and 32.2% detractors, so the overall NPS is 4.4. The tool adds what most NPS dashboards leave out: a 95% confidence interval, here [−2.0, 10.8]. With 640 responses the margin of error is ±6.4 points, so a change from 4.4 to 8 next quarter would be within the noise. Always report the interval with the score.

Comparing groups

By channel, Store has NPS = 20.4, Web −1.4 and App −5.0. Are these differences real? The pairwise comparisons test each difference with a z-test and correct the p-values for multiple comparisons with Holm’s method. Store beats App by 25.4 points (p_Holm = .004) and Web by 21.8 points (p_Holm = .011), while Web and App do not differ (p = .655). The overall score is also significantly below the benchmark of 30 (p < .001). The managerial story: the app experience is the weak spot, which is exactly what the redesign should address.

What NPS can and cannot do

NPS is simple, widely benchmarked and easy to track. But it is a single item, it throws away information by collapsing 11 points into three groups, and it says nothing about why customers would or would not recommend you. Always add an open question (“What is the main reason for your score?”) and a driver analysis (Lesson 3). Benchmarks differ enormously by industry, so compare yourself with your own past scores and direct competitors.

SERVQUAL: the gap model of service quality

SERVQUAL (Parasuraman, Zeithaml and Berry, 1988) measures service quality on five dimensions with 22 items, each asked twice: what customers expect from an excellent company, and what they perceive from yours.

• Tangibles: physical facilities, equipment, appearance of staff.
• Reliability: performing the promised service dependably and accurately.
• Responsiveness: willingness to help and prompt service.
• Assurance: knowledge and courtesy of staff, ability to inspire trust.
• Empathy: caring, individualised attention.

The gap score is perception minus expectation. Negative gaps mean the service falls short.

All five dimensions show significant shortfalls in the example of 210 customers. The largest gap is Reliability (−1.26, d = −1.52), followed by Responsiveness (−0.88), Empathy (−0.72), Assurance (−0.53) and Tangibles (−0.39). The weighted SERVQUAL score (−0.86) uses the importance weights customers assigned to the dimensions, of which Reliability received the most (32%). The conclusion is unambiguous: fix reliability first. Cronbach’s alpha for each dimension (.77 to .83 for perceptions) confirms that the items within each dimension measure the same thing.

SERVQUAL in practice

Asking 44 items is long. Common variations are SERVPERF (perceptions only, which often predicts overall quality as well as gaps), shortened versions with two or three items per dimension, and industry adaptations (e-service quality, hotel or hospital versions). If you adapt the items, check reliability and, ideally, the factor structure (Course 1 and Course 2 show how).

Reporting

The Net Promoter Score was 4.4 (95% CI [−2.0, 10.8], n = 640), significantly below the industry benchmark of 30 (p < .001). Store customers (NPS = 20.4) were more likely to recommend the company than app (−5.0) and web customers (−1.4; Holm-corrected p ≤ .011). SERVQUAL gap scores were negative on all dimensions, largest for reliability (−1.26) and responsiveness (−0.88).