Skip to main content
Playbook

Twenty user interviews, no calendar required.

Send one well-framed question to 20 people. They record a 60-second voice note when it suits them. You read the AI summaries over morning coffee. Same qualitative insight, none of the scheduling overhead.

Try it free
No login required for your prospects

The short answer

Write one open-ended question anchored on a real past moment. Create a Magic Link, send it to 20 people via email or Slack, and wait 24 hours. Each respondent records a 60-second voice note in their browser, no login needed. Read the AI summaries in one sitting the next morning, then book live calls only with the two or three answers that surprised you.
20 interviews
completable in 48 hours, no scheduling required
60 sec
average response, long enough for a real answer, short enough for anyone to record
No scheduling
calendar stays clear, respondents record whenever works for them

The playbook

Five steps. Do them in order. Step one is where most people skip ahead and pay for it later.

  1. 1

    Pick one question: the hardest part

    Most user interviews fail before anyone says a word, because the question is too broad. "Any feedback?" gives you nothing. "What do you think of the product?" gives you "it's good". You need a question specific enough to invite a real story. Ask about a moment, not an opinion. "What almost stopped you from signing up?" works. "Walk me through the last time you tried to [the job to be done]" works. If you cannot answer your own question in 60 seconds, sharpen it before sending.

  2. 2

    Create a Magic Link in under 2 minutes

    Log in to HeySpeak, paste your question, copy the link. That is it. You can add your calendar as an optional fallback so respondents who want to talk live still have that path. One link. One question. Works on any phone or laptop without a login screen on the receiver's end.

  3. 3

    Send it to 10–30 people

    The link works in an email, a Slack message, a WhatsApp, or a LinkedIn DM. Write one short, personal message: "Hey, I'm doing quick research on X, would love your honest take, takes 60 seconds." Send to 10–15 people to start. If responses are coming in, send the next batch. Warm contacts reply at 40–60%. Cold contacts are worth trying once you have refined the question.

  4. 4

    Wait. They record when it suits them.

    Recipients click the link and record a 60-second voice note in any browser, no app download, no account creation. Most people respond within a day. Some within the hour. You do not chase calendar slots. You do not send Calendly links and wait for a slot two weeks out. The research runs while you work on other things.

  5. 5

    Read the AI summaries in one sitting

    Each voice note gets transcribed and summarised automatically. Open your dashboard the next morning. Scan all 20 summaries in order. You are looking for the patterns: the repeated phrases, the same hesitation appearing in four different answers, the problem framed the same way by strangers who never talked to each other. Mark the two or three responses that break the pattern. Listen to the audio on those. Tone carries things text does not.

What makes a good question for voice research?

The question does most of the work. A bad question gets you "it's great" or "I don't know" in 15 seconds. A good question gets you a story. The difference is almost always specificity. Broad questions ("What do you think of the product?") activate the polite part of the brain. Specific questions about a real past moment activate memory.

Voice works best when the answer has texture, when you want to hear hesitation, a change in pace, the phrase someone reaches for before they land on the right word. For that to happen, the question needs to invite a story, not a rating. "What almost stopped you from signing up?" is better than "How satisfied are you?" "If you had to explain this to a colleague in one sentence, what would you say?" reveals how users actually frame your product in their own words. More useful than any copy test.

Two things to avoid. First: questions with yes/no answers. If the honest response is a single word, use a form instead. Voice is wasted on it. Second: future-intent questions ("Would you pay for X?"). People consistently overestimate their own intent. Ask what they did last week, not what they plan to do next month.

Question examples

  • "What almost stopped you from signing up?" Surfaces real friction, invites honesty, opens with a moment not an evaluation.
  • "If you had to explain this to a colleague in one sentence, what would you say?" Reveals how users actually think about your product, not how you describe it.
  • "Walk me through the last time you tried to [job to be done]. What happened?" Gets you a story with a beginning, middle, and the place where things went wrong.
  • "What is the thing in your week that takes longer than it should?" Broader opener for early discovery, before you have a specific hypothesis to test.

Keep going

Two more pages on the same problem from different angles.

For PMs

HeySpeak for product managers

How PMs use async voice for feature validation, retros, and stakeholder alignment.

Comparison

Voice feedback vs. text forms

When async voice gives you more signal than a Typeform survey, and when it does not.

Common questions

How many people should I send it to?
Start with 15–20. That is usually enough to see a pattern emerge: you will notice the same phrase coming up from people who do not know each other. If you want sharper segmentation (by role, company size, use case), run separate rounds of 15–20 for each segment. Async makes running multiple rounds fast enough that this is practical.
What is the best question to ask in a user interview?
Anchor on a specific past moment, not a hypothetical future. "What almost stopped you from signing up?" surfaces real friction. "If you had to explain this to a colleague in one sentence, what would you say?" reveals how people actually think about your product. "Tell me about the last time you tried to [job to be done], what happened?" gets you a story. Avoid "Would you use X", as people overestimate their own intent.
How do I share the link?
Copy and paste it anywhere. Email, Slack, WhatsApp, a LinkedIn DM, a Twitter reply: the link works in all of them. It opens in the recipient's browser with no login or download required. A short, personal message with the link outperforms a templated bulk send every time.
What do I do with the summaries?
Read them all in one sitting, ideally the morning after you sent the batch. Write down any phrase that appears more than twice. Those repeated phrases are your users' words for the problem: use them in your copy, your positioning, your next sales conversation. Book a live follow-up call only with the respondents whose answers surprised you. That is usually two or three people, and you arrive at those calls already knowing the baseline.
When does async voice not work for user research?
Two cases. First: questions that require back-and-forth. If you need to follow up on every answer with a clarifying question, a live conversation is faster. Second: yes/no or rating questions. If you are running a satisfaction survey or NPS, use a form. Voice is overkill. Async voice earns its place when tone and storytelling matter and when the question is open-ended enough that each person has something genuinely different to say.

Run your first round this week.

Five free responses to start. Setup takes under 2 minutes.

Create your Magic Link