Voice of Customer Survey: Questions and Voice Answers

What a voice of customer survey is (and is not)

A voice of customer (VoC) survey is a questionnaire sent to customers to learn what they expect, what they experienced and what they would change. It is one input to a VoC program, not the program itself: the program is the habit of collecting feedback from several sources — surveys, support tickets, reviews, calls — and acting on it. If you are weighing platforms for the whole program, that is a different decision from the one below. This page is about the survey, and specifically about getting better answers out of it.

The rating questions in a VoC survey are easy to score and hard to act on. A 6 out of 10 tells you something is wrong, not what. The open-ended question that follows the rating is where the usable material lives, and it is also the question most respondents answer in four words or skip.

When to send one

The timing decides what you can learn, so decide the trigger before you write a single question.

• Relationship surveys go out on a calendar — quarterly or twice a year — to the whole base or a sample. They measure overall loyalty and priorities (typically an NPS or satisfaction question plus a few drivers).
• Transactional surveys fire after an event: a purchase, a support case closed, an onboarding step. They measure that one interaction while it is fresh.
• Lifecycle surveys target a moment of change: a renewal, a cancellation, the first 30 days. These are the ones where the “why” matters most, because the customer is about to stay or leave.

Keep them separate. Mixing a relationship question into a post-support survey muddies both readings.

The questions that carry the survey

You do not need 60 questions. Most useful VoC surveys are five to eight, built from a few jobs:

Two rules save most surveys. Ask the rating first and the reason second, so the open end is anchored to a number the respondent just gave. And ask one thing per question: “How was our pricing and support?” cannot be answered or analyzed.

The open-ended problem

Every VoC team hits the same wall. The “why” question is the most valuable one and gets the shortest answers, because typing a thoughtful reply on a phone is work. The result is a column of “good service” and “too expensive” — accurate, and not enough to decide anything from.

Speaking is a lower-effort way to answer the same question. People explain more when they can just talk, and the answer arrives with the context that a one-line typed reply drops. That is the case for adding a spoken option to that one question, not to the whole survey.

Adding voice answers to the survey you already run

Voice Capture is a widget you paste into the survey you already have. It does not replace your survey platform or your VoC process. The respondent taps a button, speaks, and the answer is transcribed automatically (36 languages) and written back into the same field as the typed answers, so it comes out in your normal export, next to the rating it explains. It works on platforms that let you run your own JavaScript — Alchemer, QuestionPro, FormAssembly and Sawtooth Discover, or any web form that accepts custom code. Platforms that strip custom JavaScript, such as SurveyMonkey and Typeform, cannot host it; the full list is in Which survey platforms support voice responses.

Three practical points for a VoC setting:

• Keep the text box. Voice is an option. Some customers are in a meeting, on a train or simply prefer to type, and forcing speech costs you their whole response.
• Mind the length limit. The maximum answer is 90 seconds on the free trial, 180 on Essential and 300 on Professional. For a “what is the main reason for your score?” question that is plenty; word the prompt to fit (“in a sentence or two” or “about a minute”).
• Transcript, not recording. Voice Capture transcribes the audio and discards it; only the text is stored, in EU-based infrastructure. That suits VoC analysis, which codes and quotes text. If your program needs to keep recordings, for tone analysis or compliance, it is not the right tool.

For an NPS-style survey specifically, the setup is described on NPS with voice responses.

What changes in the data

On our own production data (7 studies, 7,169 responses), spoken answers averaged 23.1 words against 10.1 for typed answers in the same studies — roughly twice as long, not the order-of-magnitude gap some vendors advertise. The sample and the caveats are on /methodology. Longer answers usually carry two or three codable ideas instead of one, so a code frame built for typed one-liners may need more codes. Read a sample of transcripts first: brand and product names are the words most often mis-heard, and a consistent error is a find-and-replace.

Turning answers into something you can act on

1. Read a sample before coding. Fifty verbatims will show you the themes and the transcription errors.
2. Code the open ends into a frame of themes (price, reliability, support speed…) and count them against the rating, so the finding is “detractors mention onboarding 3× more often”, not a pile of quotes.
3. Close the loop. Tell the customer, or at least the owning team, what changed. A VoC survey that never visibly leads to anything lowers the next response rate.

Frequently asked questions

What is a voice of customer survey?

A questionnaire that collects customers’ expectations, experiences and suggestions, usually combining rating scales with at least one open-ended question. It feeds a wider VoC program that also draws on support tickets, reviews and call data.

How many questions should it have?

Five to eight is enough for most purposes: an overall rating, one or two driver questions, the reason in the customer’s own words, and a gap question. Longer surveys lower completion without adding much you can act on.

Does the survey have to be spoken to count as voice of customer?

No. “Voice of customer” is about customers’ feedback in general. Letting people answer an open question by speaking is one way to get richer answers, not a requirement.

Can I add voice answers without changing survey platform?

If your platform lets you run custom JavaScript on the respondent page, yes: you paste a script once and point it at the question that should accept a spoken answer. The setup guides cover each supported platform.

How much does it cost?

One credit is one transcribed response and credits do not expire. The free trial includes 250 credits; the Essential pack is $99 for 1,000. Current packs are on the pricing page.

Where to start

Pick the one open-ended question in your current VoC survey that you most wish had longer answers, and test voice on that question alone with a tenth of the sample. The free trial is enough to do it. Start free.