Voice feedback: why spoken answers tell you more than typed ones

What voice feedback is

Voice feedback is feedback people give by speaking instead of typing: they record a spoken answer to your question, and you get the recording and a transcript. A voice survey is a survey that asks some or all of its questions this way. It works in the browser, on a phone or a laptop, with no app to install.

The term also has an older meaning. In customer service, a voice survey is the automated phone survey after a support call, where people press a number or answer a recorded voice. This page is about the other kind: asking your users, customers or prospects open questions and letting them answer out loud.

Typing edits. Talking tells.

When people type, they edit. When they talk, they tell you what they think. Anyone who has read a column of survey answers knows the typed version: “fine”, “too expensive”, “more integrations”. The person had more to say. Typing on a phone made it expensive to say it, so they wrote the headline and moved on.

Survey researchers have measured the difference. In a 2024 experiment, Timo Lenzner, Jan Karem Höhne and Konstantin Gavras randomly asked 1,001 smartphone users to answer the same open questions either by speaking or by typing. Spoken answers averaged 43.8 words. Typed answers averaged 14.2. An earlier study by Gavras, Höhne and colleagues described the difference in kind: spoken answers were intuitive and spontaneous, written ones intentional and conscious.

That spontaneous version is the one you want. A considered answer is the summary people think you want to hear. The first version holds the hesitation, the example that came to mind, and the words your customers actually use for their problem, which are the words your landing page should use too.

It is also faster for them. In a study at Stanford and the University of Washington, people entered text by speech at 161 words a minute, against 53 on a phone keyboard, with fewer errors. Höhne and colleagues found the same in surveys: voice answers were longer and took less time to give. You ask for more and take less.

Catch the reaction as it happens

Voice lets people react while they look, so you hear the first impression instead of the version they reconstruct afterwards. Show someone a new landing page and ask them to type what they think, and you get a verdict. Ask them to talk while they look, and you get the reaction as it forms: what they read first, where they hesitated, the moment they got it, and the moment they stopped trusting it.

Usability researchers have worked this way for decades. Jakob Nielsen calls thinking aloud the single most valuable usability method, because you hear people’s misconceptions as they happen. Voice answers bring the same idea to any question where people react to something you show them: a concept, a pricing page, a message, an onboarding screen.

Read the typed version of this reaction and you would get “Looks good, a bit pricey.” The spoken one found the real objection in the first ten seconds: fear of what the product says to a big client on your behalf. The price came up too, but later, and softer. Fix the fear first: let people approve each reminder, and say so on the page.

A survey that feels like a conversation

Answering out loud feels like talking to someone, not filling in a form, and people bring more of themselves to a conversation. Nobody enjoys typing a paragraph into a text box on a phone. Talking is how people have always given feedback: to a friend, to a colleague, to the shop owner. Voice brings that back, without anyone having to book a call. It even changes how long a survey feels: in the Lenzner study, people who answered by voice rated the survey as shorter than people who typed, while saying three times as much.

You also hear how something is said. A recording carries the pause before “it’s fine”, the laugh, the rush of words when someone describes the thing that annoys them most. In a 2024 study, Höhne and colleagues could predict how interested people were in a question from the pitch and loudness of their voice alone. Bob Moesta, who co-created the Jobs to be Done switch interview, says he listens for the energy in a story, because a decision that mattered sounds different from a habit. Text flattens all of that.

When to ask for voice

Ask for voice wherever the reason matters more than the count. Ask people to type when you need an exact answer, such as a name, a company or an email address, because a transcript can misspell it. Use a click where you want to count or compare. Most good studies use all three: a click to sort people, a few typed fields for the facts, and spoken answers for the why.

Voice works best for:

• Concept and message tests: the first reaction to an idea, a headline or a landing page, before people have had time to be polite. The concept testing guide has the questions.
• Customer discovery: the story of the last time the problem happened, which people tell far better than they type. See the customer discovery guide.
• Churn and lost deals: the reason behind “too expensive”, which is rarely only the price.
• The “why” after a score: the open question after a product-market fit or satisfaction question. See the product-market fit survey guide.

Not everyone will record, and that is fine. Some people are on a train or in an open office. In the survey research, forcing everyone to speak lost people, and giving them the choice kept them, as Melanie Revilla and Mick Couper found. So Feedbaq asks for voice first and keeps “Type instead” one tap away. You get the richer answer from everyone who talks, and an answer from everyone else.

How voice answers work in Feedbaq

Feedbaq is a tool for collecting feedback from your users. You write a few questions and share them as a link. People answer by voice, screen recording or text, and you get every answer with a transcript. Your AI agent can set it up and analyze the answers for you.

Voice is at the heart of it. Any open question can be a voice question, and this is how it works:

• A link, a tap, and they talk. Participants open your link on a phone or a computer, with no account and no download. Feedbaq asks for the microphone only at the first voice question, and “Type instead” is always one tap away.
• As long as it takes. An answer can run up to 10 minutes. You pay per response, not per minute, so a long, thoughtful answer costs the same as a short one.
• A follow-up while the thought is fresh. Feedbaq transcribes each answer in seconds. With AI follow-up questions turned on, the next question comes straight away, about what that person just said, up to three times per question. You set the questions, and the participant just talks.
• Every answer, word by word. You get each answer as a recording with its transcript, timed word by word. Click any word to hear the recording from that moment.
• Read by your agent. Your agent reads every transcript, finds the patterns and quotes people exactly, with a link to the second each quote was said. See your AI agent as a researcher.

Hear your users this week

1. Pick the question you most need a real answer to. The one where you have been guessing.
2. Make it a voice question, and turn on a follow-up. Or start from the concept test template, which puts a design in front of people and asks for their first reaction by voice.
3. Answer it yourself first. Record an answer, read the transcript and see the follow-up it asks.
4. Send the link to ten people. Customers, trial users, your waitlist. Listen to the first answers as they come in.

Building a study is free. You subscribe when you publish: $49 a month, with 100 responses and every feature included.

Questions people ask

Sources

1. Lenzner, Höhne and Gavras, spoken versus written answers to open-ended probes Journal of Survey Statistics and Methodology, 2024.
2. Gavras, Höhne, Blom and Schoen, “Innovating the collection of open-ended answers: the linguistic and content characteristics of written and oral answers” Journal of the Royal Statistical Society, Series A, 2022.
3. Höhne, Gavras and Claassen, “Typing or speaking? Comparing text and voice answers to open questions on sensitive topics in smartphone surveys” Social Science Computer Review, 2024.
4. Ruan, Wobbrock, Liou, Ng and Landay, “Comparing speech and keyboard text entry for short messages in two languages on touchscreen phones” Stanford University and University of Washington, 2016.
5. Höhne and colleagues, “The sound of respondents: predicting respondents’ level of interest in questions with voice data in smartphone surveys” Quality & Quantity, 2024.
6. Revilla and Couper, letting respondents choose between voice and text answers Survey Research Methods.
7. Jakob Nielsen, “Thinking aloud: the #1 usability tool” Nielsen Norman Group, 2012.
8. Bob Moesta on Jobs to be Done Intercom podcast.
9. Genesys, “What is a voice survey?” The call-center meaning of the term.