- Customer satisfaction is measured on concrete experiences and on attributes of the service, not only through a general question about being satisfied.
- NPS is the difference between the percentage of promoters (9-10) and detractors (0-6) on the recommendation question; it is a simple, comparable indicator, but explains nothing on its own.
- Driver analysis shows which attributes influence satisfaction and loyalty most, and therefore where investment pays.
- The sample is drawn at random from the customer base, stratified by segment, not from customers who volunteer to answer a link.
- A satisfaction study has value only when repeated at equal intervals, with the same questionnaire, and tied to concrete actions after each wave.
1. The problem: customers who leave without saying why
Most dissatisfied customers do not complain. They leave. The company notices the churn rate, the fall in repeat orders or the rise in complaints on one channel, and discovers the problem after it has become expensive. The customer satisfaction study is the instrument that moves detection earlier: it measures systematically how customers perceive the service, identifies the points where the experience breaks down and shows which improvements would have the greatest effect on loyalty.
Two errors make most satisfaction studies useless. First: only a global score is measured, "how satisfied are you overall", which says that something is wrong but not what. Second: only customers who volunteer to answer a link sent to everyone are asked, which brings to the surface the very satisfied and the very dissatisfied, and the average between them describes nobody. A proper study is quantitative research with a random sample, questions about concrete experiences and an analysis of the factors.
2. What is measured: satisfaction, loyalty, effort
A complete study measures three distinct things, because they move differently.
- Satisfaction: the evaluation of the experience, overall and by attribute. Usually on a scale from 1 to 5 or 1 to 10, for the service as a whole and for each relevant component: waiting time, staff courtesy, clarity of information, problem resolution, value for money.
- Loyalty: the intention to continue, to buy again and to recommend. A customer can be satisfied and still leave for a better offer; satisfaction is necessary but not sufficient.
- Effort: how easy it was for the customer to get what they needed. The effort indicator, measured on a scale from "very easy" to "very difficult", predicts churn well in services: customers do not forgive complications.
Each indicator relates to a concrete, recent experience: the last visit, the last interaction with support, the last delivery. Questions about "the service in general" produce answers about the company's image; questions about "the last time you…" produce answers about the service.
3. NPS: what it is, how it is calculated, what it does not say
NPS (Net Promoter Score) starts from a single question: "How likely are you to recommend [the company] to a friend or colleague?", on a scale from 0 to 10. Respondents who give 9 or 10 are promoters, those giving 7 or 8 are passives, those giving 0-6 are detractors. The score is the percentage of promoters minus the percentage of detractors and can range from -100 to +100. A company with 40% promoters, 35% passives and 25% detractors has an NPS of 15.
The advantage of NPS is simplicity: one question, one figure, comparable across companies, branches, periods and, with caution, sectors. The disadvantages are equally real. The score compresses 11 points into 3 categories and loses information: a move from 6 to 8 does not change the score, although it changes the customer. It varies with culture and language: in some countries respondents rarely give 9 or 10 however satisfied they are, and international comparisons are misleading. It is sensitive to the data collection method: the same customer gives different scores by telephone and online. And, most importantly, it explains nothing: an NPS of 15 does not say why detractors exist or what would turn them into promoters.
4. Driver analysis: where investment pays
The most valuable part of a satisfaction study is not the score but the driver analysis: which attributes of the service influence overall satisfaction and loyalty the most. The link between the rating of each attribute and the global indicator is calculated statistically, through correlation or regression, and the result is presented in a matrix with two axes: the derived importance of the attribute and its current performance.
| Quadrant | Importance | Performance | What it means |
|---|---|---|---|
| Improvement priority | High | Low | The attributes dragging the score down; the first money is spent here |
| Strengths to maintain | High | High | What works and must not be broken by cost cutting |
| Secondary priority | Low | Low | Real problems, but with little effect on loyalty |
| Possible overkill | Low | High | Resources invested in what customers do not reward |
Statistically derived importance often differs from stated importance. Customers say price matters most; the analysis shows that resolving the problem at first contact explains three times more of loyalty. Decisions are taken on derived importance.
5. The sample: random, from the customer base
The company holds the complete list of its customers, so the sample can be drawn at random, with equal probabilities, possibly stratified by the segments to be reported: product type, branch, tenure, customer value. This is the most favourable situation in quantitative research, and wasting it is the most frequent error: the link sent to everyone, with a 3% response rate, produces a self-selected sample in which motivated complainers and fans are over-represented.
The sample size is set by the reporting segments: every branch or segment you want to compare needs at least 80-100 respondents, and detecting a 5-point NPS change between waves requires several hundred respondents per wave. The standard method is the telephone survey, with good response rates on customer bases and control over who answers; for customers with e-mail addresses, the online survey with individual invitations and reminders is a cheaper alternative, provided the response rate is reported and the structure of respondents compared with that of the base.
6. The questionnaire: short, concrete, with one open question
A satisfaction questionnaire lasts 8-12 minutes and has a fixed structure:
- Confirmation of the recent experience: which service, when, on which channel.
- Overall satisfaction with that experience.
- Attribute ratings, 8-15 of them, on the same scale, in rotated order.
- NPS and the open question "what is the main reason for your score?".
- Effort: how easy it was.
- Loyalty: the intention to continue, to buy again; for contract services, the probability of renewal.
- Problems: "have you had any problem in the last 6 months? was it resolved? how quickly?" Customers whose problem was resolved well are often more loyal than those with no problems; those with an unresolved problem are the main detractors.
- Segmentation data not available in the customer base.
The open question is the most important in the questionnaire. Coded thematically, it turns NPS from a score into a list of causes, with the frequency of each among promoters and detractors.
7. The rhythm: transactional or relational
There are two types of satisfaction study, with different purposes. The relational study measures the overall relationship with the company, on a sample of all customers, usually half-yearly or annually, and serves strategic decisions. The transactional study measures a specific interaction immediately after it took place, continuously, on a sample of customers who had that interaction that week, and serves operations: an alert when a branch slips, feedback on each channel. The two complement each other and should not be confused: a transactional NPS after a successful delivery is systematically higher than the relational NPS of the same customer.
Whatever the type, the questionnaire is kept identical between waves, and each wave ends with a list of actions, with an owner and a deadline. A satisfaction study repeated without actions between waves predictably produces the same results and, worse, customers who stop answering, because they already answered and nothing changed.
8. From results to actions
The report of a satisfaction study has four levels. The global scores, with the change from the previous wave and significant differences marked. The scores by segment and by unit: branches, channels, customer types, identifying those below average. The driver analysis, with the priority matrix. And the causes, from the open question, with representative quotes for each theme. From these four levels follows the list of actions: what changes in process, in training, in communication, where, by when.
Some findings call for an additional method. If scores differ widely between branches on staff behaviour attributes, the cause is checked through mystery shopping, which observes directly what happens at the counter. If detractors cite an unclear reason, a few in-depth interviews with them clarify more than a new survey wave.
9. What to ask the agency
- Random extraction of the sample from the customer base, stratified by the reporting segments.
- The data collection method and the estimated response rate, with a comparison between respondents and the base.
- The attribute list, built from the customer's real experience, possibly after a few exploratory interviews.
- NPS calculated correctly, reported with the distribution across the three categories and the margin of error.
- Driver analysis and the importance-performance matrix.
- Thematic coding of the open question.
- A repeatable report format and a recommendations section with priorities.
- Data protection: how customers' personal data passed to the agency are handled.
Measured properly, customer satisfaction becomes a management indicator as concrete as sales and earlier than them. Measured badly, it becomes a score presented quarterly that changes nothing. The difference between the two is almost entirely methodological, and it is decided before the first question.
See also: Quantitative research Mystery Shopping Decision support Telephone surveys (CATI) Online surveys (CAWI) Mystery shopping: measuring customer experience
Do you have a decision to take?
Tell us the problem, not the method. We propose the right research design and explain what it can and cannot deliver.