---
title: "Client feedback survey or voice interview: how to choose"
description: "An online form measures what you already know how to ask. A voice interview finds what you did not know to ask. Here is how to choose, then run both together."
url: "https://raconte.ai/en/blog/questionnaire-ou-interview-vocale"
---

# Survey or voice interview: how to choose the method for your customer listening

An online form measures what you already know how to ask. A voice interview finds what you did not know to ask. Here is how to choose, then run both together.

October 28, 2026

Godefroy

Key takeaways

*   A customer feedback survey measures what you already know how to ask: it confirms or disproves your hypotheses. A voice interview finds what you did not know to ask, because someone asks the follow-up question.
*   The open-ended question is the least filled-in field of a survey: about 18% non-response, against 1 to 2% for a closed question. The median response rate of an online survey sits around 10% of your base.
*   Answered out loud, the same open question brings three times more words than typed: 44 spoken against 14 typed. The interviewer's follow-up turns a bare “too expensive” into something a team can act on.
*   The survey stays the right choice to count a precise behaviour, reach a very large volume, meet an accessibility constraint, or respect an audience that refuses recording.
*   The cadence that works combines both: a short survey to the whole base every quarter to measure, targeted interviews every month on the low scores to understand.

A customer feedback survey measures what you already know how to ask. A voice interview finds what you did not know to ask: someone listens to the first answer, asks the next question, and a score that said nothing starts talking. The choice of method follows the decision you need to make: track a trend across your whole customer base, or understand why it moved. And in most teams the two methods work together rather than compete.

The question usually comes up while preparing the quarterly survey. A questionnaire costs almost nothing to send, an interview costs time, and the trade-off looks settled in advance. It is less settled than it looks: the very cheap survey is very bad at measuring the thing you actually lack, the why behind the scores, while hand-run interviews do not scale. This article compares the two methods on what they bring back and what they cost, says plainly when the survey remains the right choice, and lays out a cadence that runs both together.

## What a survey does well

An online survey is still the fastest way to ask a whole customer base anything. Five thousand people receive the link the same day, each answers whenever it suits them. Ticking a box takes a second, so processing the results fits in a dashboard.

Volume makes up for attrition. Survicate, a survey tool vendor, analysed [4,332 surveys run in 2025 by 460 companies](https://survicate.com/reports/survey-response-rate-benchmarks/): the median response rate reaches 10%, and 8% for B2B surveys. Ten percent of five thousand customers still makes five hundred answers, and no team runs five hundred interviews.

Comparability over time is the second strength. Asked with exactly the same wording every quarter, the same question produces a series you can read quarter over quarter, segment by segment. That is what makes a feedback survey valuable: a single score says almost nothing, a run of scores shows a trend and flags an incident on a specific segment.

Accessibility and cost complete the picture. A survey can be read, reread and filled in at any hour, with no recording, by people who cannot or will not talk. And one more respondent costs a few cents: no other method comes close to that ratio. The questions to ask in this format, with what each one is for, are ready to copy in [our commented customer satisfaction survey example](/en/blog/customer-satisfaction-survey), and the short forms tied to a precise moment of the customer relationship, like a support ticket or a cancellation, are in the [four feedback form templates](/en/blog/customer-feedback-form).

## What a written survey misses

A closed list gives back the items it contains, and nothing else. The reason your customers leave is not on the list, because you do not know it yet. A survey confirms or disproves hypotheses and produces none of them: that is its job, and it does it well. The trouble starts when you ask it for an explanation.

The reflex for that explanation is the free-comment field at the end. It is the question respondents skip: the Pew Research Center measured [18% non-response on open-ended questions, against 1 to 2% on closed ones](https://www.pewresearch.org/decoded/2021/10/14/why-do-some-open-ended-survey-questions-result-in-higher-item-nonresponse-rates-than-others/), across 92 questions spread over more than 40 surveys of its American panel. The answers that do arrive often come in three words, “too expensive”, “nothing to report”. None of them says too expensive compared to what.

The result is a score with no why attached. Three customers give an 8: the product is excellent but the price too high, usage is light and the opinion barely formed, an incident got handled well by support. The same score covers three situations that each call for a different decision. Thomas Jones, a company executive, and Earl Sasser, a professor at Harvard Business School, showed this in the Harvard Business Review in 1995, in [Why Satisfied Customers Defect](https://hbr.org/1995/11/why-satisfied-customers-defect): between 65% and 85% of customers who switched suppliers described themselves as satisfied or very satisfied just before leaving. The score can be good and the customer gone anyway.

A survey stretched too long makes all of this worse. Every extra question costs respondents. The promised duration has to hold too: a form announced at two minutes and actually taking seven loses its respondents halfway through and keeps only the most patient ones. These blind spots of the closed format are detailed question by question in [the commented survey example](/en/blog/customer-satisfaction-survey), along with the share that voice recovers.

## What voice brings

A conversation picks up what a text field loses: the hesitation before an answer, the tone that rises on one subject, the word dropped and then taken back. These signals do not turn into decisions on their own, but they point to the passage worth rereading in a transcript. Written answers carry none of them.

The follow-up question is the main contribution. “Too expensive” is where a form stops. In a conversation, the next question comes straight away: too expensive compared to what, since when, at which point of the journey. The answer to that second question is the one a product team can act on, and a form never asks it.

People also speak at far greater length than they type. Timo Lenzner, a survey methodologist at GESIS, the Leibniz Institute for the Social Sciences, and his co-authors asked the same open questions of 1,001 people in Germany, out loud for some and on a keyboard for the others. Their comparison, published in 2024, found [44 words on average for a spoken answer against 14 for a typed one](https://academic.oup.com/jssam/article/12/5/1295/7726361). The study calls for care: 39% of the people invited to speak answered nothing, against 8% of those invited to write. The protocol tested a microphone button under a question, with nobody on the other side. An interview, where someone asks the next question, is a different setup: the conversation is exactly what a bare recording lacks.

A spoken verbatim is usable as it is. The word-for-word transcript goes to the product review, the summary and the per-answer sentiment show which passage to read first, and the quote goes to the steering committee verbatim. The “too expensive” from a comment field serves none of those three purposes.

## When the survey remains the right choice

Four situations call for a survey. Recognising them keeps you from running all your customer listening on interviews.

*   Counting a precise behaviour: the share of customers who contacted support, finished setup, renewed. One closed question is enough, and an interview adds nothing to a tally.
*   Reaching a very large volume. Past a few thousand customers, interviewing becomes a program in itself, so the survey stays the only tool that reaches everyone at a sane cost.
*   Meeting an accessibility constraint. Some respondents prefer to read, or cannot talk when they answer: the written form is the one that includes them.
*   Respecting an audience that refuses recording. Some customers will never speak to a machine that listens, and a written form is then the only method they accept.

A fifth case joins them as soon as an industry benchmark imposes its own scale and questions: the survey copies them, because external comparison comes first.

Let us also state plainly where the human interviewer stays ahead: a customer in a dispute, a strategic account cancelling, a conversation where the relationship matters as much as the information. Asynchronous voice interviews replace neither the survey nor the human interviewer. They hold the middle ground: the depth of a conversation, with the volume of a bulk send.

## A cadence for running both together

The two methods do not answer the same question, so the setup that works puts them on the same team. The most common cadence fits on two lines: a survey measures, interviews explain.

Every quarter, the survey goes to the whole base with the same wording: one overall score, one or two indicators, two segmentation questions. That is the measurement. [The available indicators, CSAT, NPS and CES, are chosen by the question you ask](/en/blog/customer-satisfaction-survey), and the [short forms tied to a precise moment](/en/blog/customer-feedback-form) cover the in-the-moment measures, after a purchase or a support ticket.

Every month, the interviews go to the cases the measurement flags without explaining them: the low scores first, by [following up with detractors within 24 hours](/en/blog/net-promoter-score), then the surprising cases, like the very satisfied customer who mentions a problem in the comment field. That is the explanation. Twenty interviews are enough to surface nearly all the needs of a segment, and the transcripts feed the monthly review. This two-speed setup is the core of a [voice of the customer program](/en/blog/voice-of-the-customer), which adds support tickets, public reviews and the reply to the customer.

```mermaid
graph TD
  A["Every quarter, the survey goes to the whole base"] --> B["The measurement flags low scores and surprising cases"]
  B --> C["Every month, an interview link goes to the customers concerned"]
  C --> D["The transcripts and summaries feed the monthly review"]
  D --> E["The fix enters the next survey wave"]
  E --> A
```

We built [Raconte](/en) for the interview half of that cadence. You write the interview once as a prompt, for instance “this customer rated their last support exchange 3 out of 5, find out what cost them the most effort”. Each customer concerned receives a link, opens it in a browser whenever it suits them and talks for a few minutes: the AI listens, asks a follow-up as soon as an answer stays vague, and stops once it has asked everything. Interviews run in French and English. A few seconds after the conversation you get the transcript, the audio, a summary and the sentiment of each answer, interview by interview. Raconte does not aggregate multiple interviews into a single report: to compare twenty interviews, you go through the API or the MCP server, where an AI assistant reads the transcripts and fills in your table.

The first 60 interview minutes each month are free, with no card, then each minute costs $0.12. Twenty ten-minute interviews make 200 minutes, so the 140 minutes beyond the free tier cost $16.80. The details are on [the pricing page](/en/pricing).

## The comparison table for the three methods

Online survey

Human-run interview

Asynchronous voice interview

Response rate

Median around 10% of the base (Survicate, 2025)

The slots you manage to book

The customers who accept the follow-up

Depth of the answer

A score, sometimes three words in the comment field

High, the interviewer adapts to each answer

High, the AI probes the vague answers

Effort on the respondent

Two to three minutes of reading and clicking

A slot in the calendar, often half an hour

A few minutes of conversation, whenever it suits them

Effort on the team

Send the link, read the numbers

Half an hour per interview, plus note-taking

Write the prompt, read the summaries and transcripts

Usable verbatims

Rare and short in the comment field

Rich, but inside the interviewer’s notes

The word-for-word transcript, per-answer sentiment

Comparison over time

Yes, same wording, same scale

Hard, every interview differs

Yes, same prompt from one wave to the next

![A survey measures what you already know how to ask, a voice interview finds what you did not know to ask: the key figures of both methods](/.netlify/images?url=_astro%2Ftwo-methods-en.BWxcYIo1.jpg&w=1376&h=768&dpl=6b031bf5959953076155d961)

## Where to start

Take the survey you are preparing and sort its questions into two columns: the ones that measure, scores and behaviours to count, and the ones that explain, reasons behind a score, friction points, motives for leaving. The measuring questions go into the survey, sent to the whole base every quarter. The explaining questions become the prompt of a voice interview, sent each month to the customers the measurement flags.

Then answer every customer you interviewed, if only briefly. Someone who spent five minutes explaining their score deserves to know it was read, and that reply is what will get the next survey filled in.

The core of a [voice of the customer program](/en/blog/voice-of-the-customer) lives in that alternation. To launch those interviews, [create a free Raconte account](/en): the first 60 minutes each month are free, and [per-minute billing](/en/pricing) then follows the real volume of your interview waves.

## Frequently asked questions

### How do you run a customer feedback survey?

Start from the decision the measurement must inform, then sort your questions into two columns: the ones that measure and the ones that explain. The measuring questions go into a short survey sent to the whole base every quarter, with exactly the same wording from one wave to the next. The explaining questions become interviews, sent each month to the customers whose score calls for one. The survey template to copy is in our commented example.

### What are the 3 indicators for measuring customer satisfaction?

CSAT measures satisfaction after a specific interaction, on a 1-to-5 scale. NPS measures the intent to recommend, from 0 to 10. CES measures the effort a journey demanded, from 1 to 7. All three track a trend over time, and none of them explains a score: the why gets collected in conversation, not in the indicator.

### What are the 5 levels of satisfaction?

Very dissatisfied, dissatisfied, neutral, satisfied and very satisfied. Keep the last two levels separate in your reports: Thomas Jones and Earl Sasser measured at Xerox that very satisfied customers repurchased six times more than merely satisfied ones, so adding them together erases the one distinction that predicts a behaviour.

### Does a voice interview replace a survey?

No, the two methods do not answer the same question. The survey counts and tracks a trend across the whole base, the interview explains the scores the survey collects without commentary. A serious customer listening setup alternates both: a quarterly survey to measure, targeted interviews every month to understand the low scores and the surprising cases.

### How long does an asynchronous voice interview take?

A few minutes of conversation, with no slot to book: the customer opens the link whenever it suits them and talks as long as they need, the AI probes the vague answers and then stops. Billing is per minute of active interview, and the first 60 minutes each month are free.

Ready to launch your first interview?

Describe your interview, share the link: an AI agent runs it by voice with each participant. You get back the transcript and analysis.

[Try for free](/login)

## Continue reading

[![Customer satisfaction survey: an example and its limits](/.netlify/images?url=_astro%2Fthumbnail.BVaPkpjh.jpg&w=1376&h=768&dpl=6b031bf5959953076155d961)

August 19, 2026

## Customer satisfaction survey: an example and its limits

A customer satisfaction survey template you can copy, what every question tells you, and the four blind spots of a score out of 10.



](/en/blog/customer-satisfaction-survey)[![Customer feedback form: 4 templates by moment](/.netlify/images?url=_astro%2Fthumbnail.BylD6fd7.jpg&w=1200&h=630&dpl=6b031bf5959953076155d961)

September 27, 2026

## Customer feedback form: 4 templates by moment

Four customer feedback form templates to copy, for after a purchase, a support ticket, onboarding and a cancellation, with the reason for every question.



](/en/blog/customer-feedback-form)[![What is a good NPS score, and how do you calculate NPS?](/.netlify/images?url=_astro%2Fthumbnail.D0ClII0G.jpg&w=1376&h=768&dpl=6b031bf5959953076155d961)

October 28, 2026

## What is a good NPS score, and how do you calculate NPS?

To calculate NPS, subtract the share of detractors (0 to 6) from the share of promoters (9 and 10). Then compare the score with your industry average for 2026.



](/en/blog/net-promoter-score)[![Voice of the customer (VoC): definition, examples, template](/.netlify/images?url=_astro%2Fthumbnail.BE9tw1Sw.jpg&w=1376&h=768&dpl=6b031bf5959953076155d961)

October 13, 2026

## Voice of the customer (VoC): definition, examples, template

What VoC means, examples from tickets, reviews, surveys and interviews, and a lightweight program template to copy, with no enterprise platform.



](/en/blog/voice-of-the-customer)
