Customer Experience

Voice of Customer: how to build a programme in 6 steps

How to build a Voice of Customer programme that drives action: relationship vs transactional NPS, CSAT, CES, AI analysis of open comments and closing the loop.

Short answer: a Voice of Customer (VoC) programme continuously collects customer feedback, analyses it and turns it into action. It combines surveys (relationship NPS, transactional NPS, CSAT, CES), analysis of open comments and unsolicited sources. A good programme is not judged by its score, but by the number of improvements it triggers.

Many companies measure their NPS. Far fewer actually use it. According to Forrester’s 2023 study of VoC practices, almost all programmes (96%) rely on surveys, but few make real use of other sources of customer feedback (Forrester). The study’s title sums up the problem: Many surveys, small impact.

I have led VoC programmes in public transport, telecoms and insurance: life and non-life, claims as well as sales. These industries are very different, but the same mistakes keep coming back. Here is the 6-step method I use, with the pitfalls I have learned to avoid.

First things first: what is NPS?

The Net Promoter Score rests on a single question: “On a scale from 0 to 10, how likely are you to recommend [company] to a friend or colleague?”

Respondents fall into three groups:

  • Promoters (9-10): loyal customers who recommend you.
  • Passives (7-8): satisfied but not engaged, and open to competitors.
  • Detractors (0-6): disappointed customers, likely to leave and to criticise you.

NPS = % promoters − % detractors. Passives count in the total number of responses but not in the calculation. The score ranges from −100 to +100.

Example: out of 200 respondents, 90 promoters (45%), 60 passives (30%) and 50 detractors (25%). NPS = 45 − 25 = +20.

But the same question can be asked in two very different contexts, and this is where many companies get lost.

R-NPS and T-NPS: two metrics, two uses

R-NPS (relationship) T-NPS (transactional)
What it measures The overall relationship with the brand The impact of a specific interaction
Typical question “Would you recommend [brand]?” “Following your claim, would you recommend [brand]?”
Who is surveyed A representative sample of the whole customer base, including customers with no recent contact Customers who have just had an interaction
When Once or twice a year, on fixed dates Continuously, within 24 hours of the interaction
What drives it Price, image, product, competition, history… The journey, the channel, the agent, the lead time
Used by Leadership, strategy, benchmarking Operational teams, coaching, closing the loop
Comparable to the market? Yes, with care (same industry, same method) No: it depends too much on the journey measured

The golden rule: never compare an R-NPS with a T-NPS. T-NPS is almost always higher, because a customer who has just been served well judges the interaction more than the brand. A T-NPS of +45 and an R-NPS of +5 do not contradict each other: they tell you that your teams are good, but that the overall relationship suffers from something else (price, product, complexity).

Used together, they complement each other. R-NPS tells you where you stand; T-NPS tells you where it is won or lost. The most useful analysis links the two: which moments in the journey pull the overall relationship down the most?

Step 1 – Start from decisions, not questionnaires

The first question is not “what should we ask customers?” but “which decisions do we want to make better?”. For example:

  • Where should we focus the customer experience improvement budget?
  • Why do customers cancel their contract?
  • Is the new digital journey simpler than the old one?

For each decision, name a sponsor in leadership and an operational owner. Without a sponsor, a VoC programme becomes a monthly report that nobody reads.

Deliverable: a one-page document linking 3 to 5 business objectives to what the programme must measure.

Step 2 – Map the moments that matter

Customers do not experience “your company”; they experience moments: taking out a policy, filing a claim, calling customer service, receiving a bill, a disrupted journey.

Identify on your customer journey the 5 to 8 moments that weigh most on satisfaction and loyalty. These are the ones that will receive a transactional survey. They vary by industry:

  • In insurance, claims handling often weighs more than the rest of the relationship combined.
  • In telecoms, it is the first bill, an outage or a change of plan.
  • In public transport, it is disruptions and the information given during the incident, more than the daily commute.

Deliverable: a customer journey map showing the key moments to listen to first.

Step 3 – Choose the right metrics and channels

Metric The question it answers When to use it
R-NPS Are our customers attached to the brand? Annual or half-yearly barometer
T-NPS Does this moment strengthen or damage the relationship? After high-stakes moments (claim, new contract, complaint)
CSAT Is the customer satisfied with this interaction? After a contact, a delivery, a purchase
CES Was it easy? After a digital journey, a request, a procedure

My advice: an R-NPS for the big picture. Then, for each key moment, one main metric: a T-NPS for moments that shape the relationship, a CES for tasks where effort is the issue, a CSAT for simple interactions. Avoid stacking all three in the same questionnaire.

Always add an open question: “What is the main reason for your score?”. That is where the real information lies.

Do not forget unsolicited sources: Google reviews, complaints, customer service calls, chats, social media. They complement surveys, whose response rates are falling, and they capture customers who never answer questionnaires.

Deliverable: a listening plan (which metric, at which moment, through which channel, how often).

Step 4 – Collect cleanly: data quality first

A VoC programme lives or dies by the quality of its data:

  • Short questionnaires: 3 to 5 questions for a transactional survey.
  • The right timing: the T-NPS or CSAT survey goes out within 24 hours of the interaction.
  • Contact rules: no more than one transactional survey per customer every 30 to 90 days, and no transactional survey in the weeks before the R-NPS barometer. If a customer goes through several moments, the most important one wins (a claim takes priority over an information call).
  • Linked data: every response must be tied to the product, the channel, the branch or the team. Without that link, you cannot act.
  • Sufficient volumes: below 100 responses, an NPS can move by more than 10 points by chance alone. Show team scores over a rolling period, and do not react to a 3-point change. Our sample size and margin of error guide explains why.
  • GDPR: a clear legal basis, customer information and a defined retention period.
  • Two languages: in Belgium, French and Dutch at a minimum, with truly equivalent questionnaires.

Platforms such as Qualtrics or Medallia automate sending and reporting, but the tool does not replace the method.

Deliverable: an automated collection set-up, with a monthly quality check.

Step 5 – Use AI analysis to find the real levers

The score tells you “how much”. The comments tell you “why”. The problem: nobody can read 5,000 comments a month.

Automated comment analysis (text analytics and sentiment analysis) lets you:

  • classify each comment by theme (lead time, price, welcome, clarity…);
  • measure positive or negative sentiment by theme;
  • identify the drivers, the themes that weigh most on the score;
  • spot new problems as soon as they appear.

This is also the step where you link T-NPS to R-NPS: which moments, and which themes within them, best explain the overall relationship? That is how you prioritise investment.

At one insurer, automating comment analysis took us from manual reading of samples to a complete view of friction points, journey by journey: new contracts, claims, customer service.

Deliverable: a dashboard with scores, themes and action priorities (in Power BI, for example).

Step 6 – Close the loop and drive action

This is the step most companies neglect. It works at two levels.

The inner loop: the customer. It relies mainly on T-NPS. Every detractor from a critical moment is contacted within 24 working hours, others within 48 hours. The goal is not only to fix the problem, but also to understand what happened.

The outer loop: the organisation. It draws on R-NPS and recurring themes. Recurring problems go to a monthly committee that decides on improvement projects, appoints an owner for each one and tracks results.

The two loops feed each other. Bain describes a European health insurer: by calling back unhappy customers, its claims unit discovered that they had to call repeatedly to find out the status of their reimbursement. The process was redesigned to inform them automatically by phone or text message. The result: an NPS up by more than 10 points and a clear rise in renewals (Bain & Company).

Finally, share feedback with frontline teams. Nothing motivates an employee more than a happy customer’s comment, and nothing helps them improve more than a specific piece of criticism.

Deliverable: a follow-up process, an improvement committee and action tracking.

KPIs to tell whether your programme works

Do not judge your programme on NPS alone. Also track:

  • Response rate: it shows whether customers feel heard.
  • Closed-loop rate: the share of detractors actually contacted.
  • Time to follow-up: the time between the feedback and the first contact.
  • Number of actions completed per quarter, and how the themes addressed evolve.
  • Business impact: retention, complaints and cost to serve.

That last point is crucial, and rarely addressed. According to CustomerGauge’s 2026 report on B2B programmes, only 4.5% have quantified the financial impact of their customer experience, and 74% cannot state their impact on revenue at all (CustomerGauge). Yet that figure is what secures leadership support over time.

The 6 most common mistakes

  1. Measuring for the sake of measuring: surveys with no decision behind them.
  2. Comparing a T-NPS with an R-NPS, or a T-NPS with a market benchmark.
  3. Too many questions: response rates collapse and quality drops.
  4. Ignoring comments: looking only at the score throws away half the information.
  5. Not calling back unhappy customers: they quickly learn that their opinion is useless.
  6. Making NPS a bonus target: teams end up optimising the score (or even steering answers), not the experience.

What about SMEs?

A VoC programme is not only for large groups. An SME can start within a few weeks:

  • an R-NPS sent to its customer base, with an open question;
  • a T-NPS or CES on the most critical moment of the journey;
  • AI analysis of the comments;
  • a simple dashboard;
  • a workshop to agree on 3 priority actions.

Then repeat the R-NPS every 6 months and measure progress.

Review how you listen to your customers

Already measuring NPS, but not sure what to do with it? We review your current set-up free of charge: your metrics (R-NPS, T-NPS, CSAT, CES), your collection rules and your closed loop. We then suggest 3 concrete areas for improvement.

Request my review or discover our AI Voice of Customer solution.

Frequently asked questions

What is a Voice of Customer programme?

It is an ongoing system that listens to customers (surveys, reviews, complaints, interactions), analyses their feedback and turns it into improvement actions.

How is NPS calculated?

Subtract the percentage of detractors (scores 0 to 6) from the percentage of promoters (scores 9 and 10). The score ranges from −100 to +100.

What is the difference between R-NPS and T-NPS?

R-NPS measures the overall relationship with the brand, on a sample of the whole customer base, once or twice a year. T-NPS measures the impact of a specific interaction, right after it happens. The first steers strategy; the second improves journeys and triggers follow-up with unhappy customers.

What is a good NPS?

It depends on the industry and on the type of NPS. A positive R-NPS is already decent in insurance, telecoms or public transport, while a T-NPS is usually much higher. The most useful comparison is with your own score over time, and with direct competitors measured the same way.

What is the difference between NPS, CSAT and CES?

NPS measures loyalty and willingness to recommend, CSAT measures satisfaction with one interaction, and CES measures how easy a task was.

How long does it take to set up a VoC programme?

Allow 6 to 8 weeks for a first working set-up in an SME, and 3 to 6 months for a multichannel programme in a large organisation.

Do I need a specialist platform such as Qualtrics or Medallia?

Not to get started. A simple survey tool and a dashboard are enough. Specialist platforms become useful as volumes and the number of touchpoints grow.

How do you measure the ROI of a Voice of Customer programme?

By linking scores and actions to business indicators: retention rate, complaints, cost to serve or additional sales.

Let’s turn your data into your next opportunity.