Customer satisfaction research: CSAT, NPS and CES programmes that lead to action

Questions this report answers

• How satisfied are customers overall, and has this changed since the last wave?
• What is our NPS, and how does it vary by segment, product, branch or channel?
• Which stages — onboarding, billing, delivery, support — drive recommendation most strongly?
• How much effort does it take customers to get a problem solved?
• Which dissatisfied customers are at risk of leaving, and why?
• How do we compare with competitors on the attributes customers value?
• Which improvements should come first, given their effect on loyalty?

When you need it

A satisfaction programme pays off when you have enough customers that you can no longer hear them all, and enough at stake that losing them hurts. Common triggers:

• Churn or complaints are rising and internal reports cannot explain why.
• Something changed in the operation: a new billing system, an app release, outsourced support or a merger of service teams.
• Management wants customer KPIs in scorecards or bonus schemes, which requires a method robust enough to survive scrutiny.
• You need a benchmark against competitors, which a survey of your own customers cannot provide on its own.

Two survey types serve different purposes. Relational surveys ask about the whole relationship once or twice a year; transactional surveys follow a specific interaction such as a delivery or a support call. Most organisations need both.

A satisfaction study is the wrong tool if you want to know why people never became customers; brand health tracking answers that. If the problem sits in one screen or form, UX research is faster. With only a handful of key accounts, structured interviews with each one beat a survey.

Sample table of contents

1. Executive summary
   headline scores, key drivers, priority actions
2. Methodology
   sample, response rate, fieldwork, weighting, margins of error
3. Headline KPIs
   CSAT, NPS and CES with confidence intervals and trend
4. Results by segment
   customer type, region, product, tenure and channel
5. Touchpoint performance
   scores for each stage of the relationship
6. Driver analysis
   importance against performance for every attribute
7. Voice of the customer
   coded open answers and sentiment by theme
8. Competitive benchmark
   where a market sample is included
9. Customers at risk
   profile of detractors and stated intention to leave
10. Action plan
    prioritised improvements with owners and target KPIs

Methodology and sources

Measures

CSAT uses a five-point satisfaction scale. Net Promoter Score asks how likely customers are to recommend you on a 0–10 scale; promoters answer 9–10, detractors 0–6, and NPS is the percentage of promoters minus the percentage of detractors, from −100 to +100. Customer effort score asks how easy it was to get an issue handled, typically on a seven-point scale. Attribute ratings per touchpoint and one or two open questions complete the survey.

Sampling

Your CRM is the sampling frame. Relational waves use a random or stratified sample; transactional surveys are triggered after an interaction. Invitations go by email or SMS, with CATI for customers without email and for B2B decision-makers. We compare respondents with non-respondents on CRM variables to check for non-response bias and weight where needed.

Analysis

Driver analysis, using regression or relative-importance methods, shows which attributes move overall satisfaction and NPS. Changes between waves are tested for significance before anyone celebrates. Open answers are coded with text analytics and checked by an analyst.

Closing the loop

The ICC/ESOMAR International Code requires research to be kept distinct from non-research activities, so respondents are only re-contacted about their feedback if they have explicitly agreed. That alert list is handled under GDPR.

What you receive

• Wave report and presentation with KPI trends and driver analysis
• Excel tables by segment, region, product and channel
• Coded verbatims with theme frequencies and examples
• Secure alert list of detractors who agreed to be contacted
• Dashboard refreshed after each wave
• Action-planning workshop with process owners

How the project runs

1. Objectives, KPI definitions and sample design — 1 week
2. Questionnaire, invitation texts and pilot — 1 week
3. Fieldwork for a relational wave — 1–2 weeks
4. Analysis and driver modelling — 1 week
5. Report and workshop — 3–5 days
6. Repeat waves quarterly or half-yearly, or run continuous transactional tracking

Common mistakes and how to avoid them

• Chasing the score instead of the reasons. NPS alone does not tell you what to fix; always pair it with drivers and open questions.
• Letting staff choose who is surveyed. Selection by front-line teams inflates scores; sample from the CRM automatically.
• Reading noise as change. A few points of NPS movement on a small sample may be random; report confidence intervals.
• Comparing scores across countries without context. Response styles differ between cultures, so compare trends within a market first.
• Tying bonuses to raw scores. It invites score-begging; combine KPIs with audits and verbatim reviews.

Who it suits

Updated: 5 Nov 2026