How to run a diary study with voice notes

The textbook diary study assumes a participant who, fourteen evenings in a row, opens a form on their laptop and writes three paragraphs about something that happened that day. Roughly half of them quit by day four. The other half write less and less each night until the entries are one-line summaries that say "the same as yesterday". The data thins exactly as the study goes long, which is the opposite of what a longitudinal method is supposed to do.

This is a working guide on how to run a diary study with voice notes: what the method is, when it beats a one-off interview, the steps from recruit to synthesis, and the operational details (cadence, completion rate, prompt length) that nobody writes down because they only matter once you have ten participants halfway through a fourteen-day window.

What a diary study is

A diary study is a qualitative user research method in which the same participants record short entries on a recurring cadence across a longer window, usually two to four weeks. Each entry answers one or two prompts about an experience as it happens, in context, on the participant's own device. The output is a longitudinal series of voice notes or text entries per participant, synthesized after the window closes.

It is not a survey repeated daily. A survey asks the same closed-ended question and aggregates a number. A diary study collects a thread per participant, where the value comes from reading the thread end to end, not averaging across them. The structure is async and longitudinal. The unit of analysis is a story across days, not a count.

Bolger, Davis and Rafaeli's 2003 review in the Annual Review of Psychology is still the cleanest summary of why the method matters: diaries capture the particulars of experience close to when it happens, in a way retrospective interviews cannot. The cost of that fidelity is that participants have to keep showing up for two weeks, and most diary studies fail because the entry burden is too high.

When a diary study beats a one-off interview

Three cases where a diary study is the right method:

• The behavior is rhythmic, not eventful. A morning routine, a weekly review, a Monday standup. A single interview asks the participant to summarize a pattern they don't actually remember in detail. A two-week diary catches the pattern at the moment it repeats.
• The experience changes across the window. Onboarding into a new tool, recovering from an injury, learning a language. The story on day one is not the story on day twelve. An interview compresses that arc into a paragraph. A diary keeps the arc.
• The context is private. Bedtime, parenting, money decisions, anything participants will not say in front of a stranger on a Zoom call. A voice note recorded alone at 10pm is the only setting where some of these answers exist at all.

Two cases where a diary study is the wrong method:

• You need depth on one specific decision. A 45-minute interview chases the thread several turns deep. A diary study collects breadth across days but never goes back for follow-up in real time.
• The behavior is rare. Diary studies need the experience to happen on something close to the cadence you're sampling. If the action happens once every three weeks, a fourteen-day diary captures it from one or two participants and the rest produce empty entries.

This isn't an either/or. We routinely run a diary study to find the moments worth interviewing about, then a single 1:1 conversation to chase the entry that did not make sense. The companion piece on async user research methodology covers the broader point: async is not a lesser interview, it is a different shape, and diary work is where the shape pays off most.

Why voice notes change the math

The reason diary studies have a reputation for high dropout is not that participants get bored. It is that typing three paragraphs every evening on a phone keyboard, alone, into a form that nags about word counts, is genuinely unpleasant. Twenty seconds of speech contains roughly the same information as ninety seconds of typed text and asks much less of the participant. The medium decides the response rate.

A voice diary entry sounds like the participant. Hesitations, corrections, the "actually, hold on, I want to redo this part" moments stay in the recording, and those are the moments that carry most of the qualitative signal. A typed entry strips them out before the researcher ever sees them. We cover the longer case on what voice catches that text loses in the essay on voice versus text; for diary work the practical version is shorter: voice keeps the participant in for week two.

"Right, so today is the day I gave up on the import feature. I tried it last week and it kind of worked, but today there was this thing where the columns came in shifted by one and I just... I'll do it manually next time. I'm tired."

That entry is twenty-two seconds of voice. Typed, it would have been three sentences and the "I'm tired" would not have been written down.

How to run a diary study with voice notes, step by step

Six steps, in order. They mirror the async methodology playbook but are tuned for the longitudinal shape.

01 · Frame the question that needs days, not minutes

Write one sentence: what would a daily entry across two weeks tell me that a 45-minute interview cannot? If the answer is "it would be more thorough", you don't need a diary study, you need a better interview. If the answer is "I would see the change between day one and day fourteen", or "I would catch the moments that participants don't remember a week later", run the diary.

The single most common failure is running a diary study because the team thinks longer equals better. Diary studies are not better. They are different. They catch arc and rhythm. A one-off conversation catches depth on a specific topic. Pick deliberately.

02 · Pick the cadence: event, interval, or signal

The diary-study literature names three sampling cadences and they are the only choice that matters for the rest of the design.

• Event-based. Participants log an entry every time a specific event happens (every time they use the feature, every time they decide between two options). Best for capturing rare-but-important moments. Caveat: participants miss events when they are in flow, so the data underrepresents the busy days.
• Interval-based. Participants log on a fixed schedule, usually once a day or once a week. Best for tracking change over time. Caveat: some entries will be empty because nothing happened on that day. That's information too. Don't punish empty entries.
• Signal-based (also called experience sampling, ESM). The tool pings the participant at random or scheduled times. Best for capturing in-the-moment state on consumer behavior. Caveat: requires a tool that can send the prompt at the right time, and a participant willing to be interrupted.

For most product-team diary studies, interval-based with a daily prompt is the right default. Event-based is a good second choice when the behavior is well-defined and bounded. Signal-based ESM belongs in academic and consumer research; it is overkill for a feature evaluation.

03 · Recruit for completion, not just fit

In a synchronous study you recruit for a 45-minute slot. In a diary study you recruit for participation across a window of two to four weeks, which is a meaningfully larger ask. The recruitment screener has to filter for two things, in order: relevance (right user segment, right context) and stamina (willing to log fourteen entries on their phone in their evenings).

The practical implication: recruit roughly 1.5 to 2× your target. If you want ten finishing participants, send the link to fifteen or twenty. Diary completion is not flat across the window. Roughly 60 to 75% of participants log on day one. Day-one logs are not the predictor of completion; day-three logs are. The participants who post on day three almost always finish. The ones who skip day three almost never come back.

A short pilot with two friendly participants, three days long, is the cheapest way to find out whether your prompts are clear. Most diary studies that fail do so because the prompt is ambiguous on day one, the participants get an unhelpful first entry, and the researcher tries to fix the prompt mid-study. By then the data is mixed. Pilot first.

04 · Write prompts that survive being read on a phone at 9pm

The craft is covered in how to write user research questions that open people up; the diary-specific version has three additional rules.

• Every prompt stands alone. It cannot reference yesterday's entry, because the participant cannot see yesterday's entry on the entry screen. If you want them to compare to yesterday, the prompt has to remind them what yesterday was about.
• One question per day, not three. Five prompts for fourteen days is the prompt design that kills diary studies. One open prompt with one optional follow-up is what people finish.
• Anchor to the day, not to the study. "What was the most frustrating part of using [tool] today?" works. "Reflecting on your overall experience..." doesn't, because participants on day three have no overall experience yet.

Most product teams overdesign the prompts. The shortest version of the question that still captures the answer is almost always the right version.

05 · Keep the entry burden under sixty seconds

The completion curve on diary studies is mostly a function of how long an entry takes to record. Sixty seconds of speech is roughly 150 words, which is enough to cover one experience with one follow-up thought. Two minutes is too much. Five minutes guarantees a dropout by day six.

Build the form so the maximum recording length is one minute and the participant sees that limit before they tap record. Friction in the form (multi-step uploads, confirmation screens, "are you sure" dialogs) is what kills the second week. The participant should be able to open the link, tap record, speak, and close the tab in under ninety seconds total.

06 · Synthesize after the window closes, not during

The biggest operational mistake in diary work is reading the entries as they land. You will anchor on whichever participant logged most consistently in week one, and every subsequent entry from a quieter participant will be read as "less interesting". Let the window close. Then read each participant's thread end to end, before you start clustering across participants.

The order of operations is the inverse of how most tools default. Read participant by participant first. The story is in the arc, not in the cross-section. Once you have a sense of every participant's story, then code across participants for themes. Our companion piece on analyzing user interview transcripts covers the coding pass; for diary work the addition is that thematic saturation lands faster because the per-participant signal is denser. Six to ten finishing participants, well-recruited, is usually enough.

The operational details nobody writes down

Three things that only matter when you are halfway through a diary study and starting to worry about the data.

Reminders should be infrequent and human. One short message at 48 hours if the participant has not started, one mid-study check-in at the halfway point, and nothing else. Daily nag emails are a faster way to lose participants than to win them back. The companion ask, mentioned in our mobile-first research notes, is to send the reminder by SMS or by the channel the participant uses, not by email. Email goes unread; the channel they are already in does not.

Empty days are data. A participant who logs "nothing happened today" on day six is telling you something real about the behavior you are studying. Don't filter empty entries out of the synthesis. If three of ten participants logged empty on the same day, the question is what they had in common, not how to cull them.

Voice transcripts lie unless you keep the audio. A diary entry transcribed-only looks like a slightly messier survey. The "I'm tired" at the end of the example above does not survive the transcription pass cleanly; the tone does not transfer. If your tool gives you transcripts and not clips, you have rebuilt a daily form in a more expensive way. Tools that keep both, like Talkful, let you read the thread fast and listen to the entries that matter.

FAQ

What is a diary study?

A diary study is a qualitative user research method in which the same participants record short entries on a recurring cadence across a longer window, usually two to four weeks. Each entry answers one or two prompts about an experience as it happens. The output is a longitudinal thread per participant, synthesized after the window closes. It is best for capturing change over time, rhythmic behaviors, and contexts a one-off interview cannot reach.

How long should a diary study run?

For most product-team work, fourteen days is the practical default. Shorter than seven days is too short to catch a pattern; longer than four weeks is too long to keep most participants engaged on a voice or text diary. Studies tied to a specific event (onboarding into a new tool, the first week of a behavior change) can run shorter. Studies tracking habit formation often run longer, with a planned drop in cadence in the second half.

How many participants do you need for a diary study?

Six to ten finishing participants per homogeneous group is usually enough for thematic saturation, following the same logic Guest, Bunce and Johnson found for interview research. Recruit roughly 1.5 to 2× your target to absorb the diary completion tail. Sending the link to fifteen to twenty people to close ten finishers is a reasonable planning ratio.

What is the difference between a diary study and a user interview?

A user interview is a single conversation, usually 30 to 60 minutes, that goes deep on one set of questions live. A diary study is a series of short entries from the same participants across two to four weeks, async, that captures change across days. Interviews give you depth on one decision; diary studies give you arc across many days. Most teams use them together: a diary study to find the moments worth interviewing about, then one interview to chase a specific entry.

What completion rate should I expect from a diary study?

On voice-first diary studies we run on Talkful, finishing rates typically land between 65 and 80% of recruited participants completing at least 70% of entries across a fourteen-day window. Text-only diary studies complete lower and produce thinner entries. Completion improves when prompts are short, the entry burden is under sixty seconds, reminders are infrequent, and the medium matches the shape of the answer. Voice helps on all four.

Are diary studies the same as ESM (experience sampling)?

No. Experience sampling method (ESM) is a specific signal-based diary cadence in which the tool pings the participant at random or pre-scheduled times during the day to log a short entry. ESM is one of three diary-study cadences (alongside event-based and interval-based), used heavily in academic and consumer research. Most product-team diary studies use interval-based daily prompts, which are simpler to run and require less of the participant. The wider taxonomy is in Bolger, Davis and Rafaeli's review and NN/G's overview of diary studies.

---

A diary study is not a longer interview, and it is not a survey on a schedule. It is a different shape: short entries, repeated, by the same people, across a window long enough to catch change. The reason most diary studies fail is that the entry burden is set for the medium of the form field, not the medium of the participant's evening. Voice notes pull the burden down to where the data starts to behave. If you have been thinking about a diary study and stopping at the dropout question, run one with voice prompts and check the day-three logs. That is the only test that counts. Talkful has a free plan that's enough for a first study, and the end-to-end voice user research guide covers what to do once the entries start landing.