Activity ⚡ Medium

Experience Sampling Method

ESM prompts users to report their experience at random or contextually triggered moments — capturing in-the-moment data that retrospective methods systematically distort through memory reconstruction.

In-Context Research Habit Research Emotional Experience Mobile UX Longitudinal Research Real-World Behaviour

01 — TL;DR

Two sentences.

The Experience Sampling Method (ESM) is a research technique in which participants are prompted — at random intervals, at fixed intervals, or immediately following a defined event — to report their current thoughts, feelings, activities, and context via a brief structured questionnaire, producing a dataset of in-the-moment experience reports that is significantly more accurate than retrospective accounts because it captures experience before memory has had the opportunity to reconstruct, smooth, and rationalise it. The core value is temporal proximity: an experience report submitted thirty seconds after a frustrating interaction captures the actual emotional state of that moment, whereas an end-of-day interview about the same event captures a reconstruction influenced by what happened afterward, by the user's general disposition, and by the social dynamics of the interview context.

The method was developed by psychologist Mihaly Csikszentmihalyi and colleagues at the University of Chicago in the 1970s as a tool for studying daily experience — specifically to overcome the methodological limitations of retrospective self-report in psychological research. It was adopted into UX and product research as mobile technology made prompt delivery practical and as product teams recognised that the gap between what users said about their experience in interviews and what they actually did in product analytics was consistently large enough to require a different research approach.

02 — When to Use

Apply this when…

→ You need to understand the emotional context users are in when they open your product — whether they are stressed, distracted, rushed, or relaxed

→ A product feature is used sporadically and unpredictably enough that observation sessions cannot reliably capture it

→ You are researching the formation or disruption of product usage habits — ESM reveals whether a habit is establishing or fragile

→ You need to understand why a retention metric drops between day seven and day thirty — ESM captures the specific moments that precede disengagement

→ The product is used in private or inaccessible contexts — a meditation app, a mental health tool — where direct observation is inappropriate

→ You are studying the relationship between emotional state and product usage — whether anxious users use different features than calm ones

When NOT to apply it

Skip it when the research question does not require temporal or contextual precision — if the answer would be the same whether collected in a session or over two weeks, ESM is too expensive. Skip it when participant burden is a concern and the research timeline cannot accommodate two to four weeks. Skip it when the product usage is frequent and predictable enough that a diary study would capture it more cheaply. Skip it when the required sample size is not achievable — ESM studies with fewer than fifteen committed participants produce directional rather than reliable findings.

03 — How It Works

The mechanism

ESM works by interrupting the flow of daily life at structured or random moments and asking participants to describe their current state — what they are doing, where they are, how they are feeling, and what they were just thinking. This interruption is the method's defining feature: it does not wait for users to reflect on their experience later, when memory has processed and altered it. The prompt arrives while the experience is still live.

Memory reconstruction systematically distorts retrospective accounts

The cognitive science behind ESM's advantage draws on decades of research into reconstructive memory — particularly work by Elizabeth Loftus establishing that memory is not a recording but a reconstruction influenced by subsequent events, current emotional state, and social expectations. When a user is asked in a Friday interview to describe their experience on Tuesday, their account is influenced by everything that happened Wednesday and Thursday, by how they want to present themselves, and by their general attitude toward the product. ESM eliminates this distortion by collecting the report while Tuesday is still happening.

Three prompt types serve different research questions

Interval-contingent prompts are delivered at fixed time intervals regardless of what the participant is doing. Signal-contingent prompts are delivered at random intervals within defined windows — capturing the natural distribution of experience states. Event-contingent prompts are triggered immediately following a defined event — such as opening the product app or completing a specific action. For product research, event-contingent prompts are often the most actionable, because they capture the user's state immediately after specific product interactions.

Prompt burden is a research design constraint, not just a participant experience issue

The more significant design constraint is prompt frequency, not prompt length. A two-question prompt delivered twelve times a day is more burdensome than a five-question prompt delivered three times a day. Research by Kubiak and Krog established that prompt completion rates drop sharply when daily frequency exceeds six to eight for studies lasting more than a week. For most product research questions, four to six well-timed prompts per day over one to two weeks produces better data than more frequent prompts that participants begin skipping.

Compliance rate, context distribution, and within-person variability

Compliance rate — the proportion of prompts participants responded to — is the primary quality indicator: studies with compliance below 60% produce biased data. Context distribution indicates whether the study has sampled the full range of usage contexts. Within-person variability — how much an individual's reported state varies across prompts — reveals whether usage is stable across contexts (established habit) or variable and conditional (fragile behaviour).

Combine ESM with product analytics

ESM is a self-report method — it captures what users say about their experience, not what they actually did. Combining ESM data with product analytics produces a richer picture than either alone: ESM reveals the emotional and contextual state the user was in when they took a specific action, while analytics reveals what that action was. When ESM reports of high frustration correlate with analytics events showing abandonment, the combination identifies both the emotional precursor and the behavioural consequence.

04 — Real Example

Spotify's commute research and the context-dependent listening behaviour ESM revealed

When Spotify's research team investigated why users' playlist engagement varied so dramatically across the day — high in commute windows, much lower midday — they used ESM to capture the actual emotional and contextual state users were in during different listening sessions. Standard analytics could identify when engagement was high or low; it could not explain why. User interviews produced generalised accounts that did not capture the moment-to-moment variation the data showed.

ESM prompts delivered at random intervals throughout the day revealed a pattern invisible to analytics: the high engagement in commute windows was not primarily about transportation. It was about transition — users were using music to manage the psychological shift between home and work contexts, and the playlists they chose in these moments were chosen for emotional regulation, not entertainment. The ESM data captured users reporting elevated stress, anticipatory focus, and deliberate music selection that was contextually specific to the commute window and absent from midday sessions. This finding directly informed the design of Spotify's commute-aware playlist features.

Spotify · Commute Listening

In-the-moment ESM data reveals emotional context that analytics and interviews cannot

Commute listening was emotional regulation, not passive entertainment

05 — In the Wild

Test yourself & see real examples

No examples yet — be the first.

Spotted a product feature that clearly was designed with deep understanding of the emotional and contextual state users are in when they use it — or one that seems completely ignorant of the real-world conditions of its use? Submit what you observed.

✓ Reviewed before publishing ✓ Your name on every example you submit ✓ Violation or fix — both welcome

Seen Experience Sampling insights ignored in a real product? Help grow the evidence base.

06 — Common Mistakes

Where teams go wrong

✗ Using fixed daily prompts instead of event-contingent or random-interval prompts. A fixed 6pm daily prompt about morning product usage captures a reconstruction, not an experience report. For product research, event-contingent prompts triggered immediately after specific in-product interactions are significantly more accurate because they are delivered while the experience is still current.

✗ Designing prompts that are too long to complete in natural contexts. ESM participants complete prompts on a train, between meetings, while waiting for coffee. A prompt that takes more than two minutes will be skipped whenever the participant is in a context that does not accommodate focused attention — meaning high-activity, high-stress contexts are systematically underrepresented. Three to five questions, each answerable in under twenty seconds, is the practical ceiling.

✗ Failing to account for non-response bias in analysis. Participants skip prompts when they are busy, distracted, or disengaged — meaning collected data is biased toward moments when participants had time to respond. Teams that analyse without accounting for this bias over-represent low-activity contexts and under-represent the high-stakes moments the research was designed to capture.

✗ Using ESM when a simpler method would answer the question. ESM is expensive in researcher time, participant burden, and analysis complexity. If the answer would be the same regardless of when in the day, in what emotional state, or in what context the data is collected, a survey or interview is appropriate at a fraction of the cost. Reserve ESM for questions where timing and context are themselves the research subject.

07 — Variations & Related Principles

Connected ideas

ESM sits within the ecological momentary assessment family of methods — approaches that collect data in the moment of experience rather than through retrospective self-report. Understanding its position relative to adjacent methods helps teams choose the right tool for the temporal and contextual precision their research question requires.

The most important pairing is ESM with product analytics. ESM captures the subjective, contextual, and emotional dimension of product interaction. Product analytics captures the objective behavioural dimension — what users actually did. Neither source alone produces a complete picture: analytics without ESM reveals patterns without context; ESM without analytics reveals context without behavioural consequence. Together they produce the most complete available picture of real product experience.

08 — 10-Min Exercise

Run it right now

⏱ 10 minutes · Solo · No prep

The Context Audit

Identify one key product metric that varies across time of day or day of week — engagement that peaks in the morning, feature usage that drops on weekends, completion rates that differ between weekday and weekend.

1. Write down the metric and the pattern you observe in your analytics.

2. For each time period where the metric is notably higher or lower, write down your current hypothesis about why. Be specific — "users are busier in the afternoon" is a hypothesis worth testing; "users use it more in the morning" is a description, not a hypothesis.

3. For each hypothesis, write the ESM prompt that would test it. An effective prompt should be answerable in under twenty seconds and should capture the specific contextual variable your hypothesis depends on.

4. Note whether your hypotheses are testable with ESM or whether you have been answering them with assumptions. If more than two hypotheses are untested assumptions driving product decisions, you have a contextual research gap worth prioritising.

10 minutes