How Diary Studies Made A Mobile Workflow Actually Fit
A story of using surveys, diary studies, and unmoderated tests to choose a merchant mobile workflow and ship a pilot that fit real days.

Interviews told us what merchants said they did. They did not tell us when or where. Our mobile flow kept missing their day. Here is how we fixed that with a short survey, a tight diary study, and unmoderated tests built from what we saw. We ended with a pilot that fit real life, not a lab.
- Start with a short survey to find who actually does the job.
- Run a focused diary on the smallest, riskiest task to learn cadence and context.
- Test unmoderated prototypes built from diary triggers, not from interview words.
- Cross-check with telemetry before you set scope.
- Ship a pilot that mirrors diary contexts. Watch it in the wild.
The false start
We were redesigning merchant workflows for phones. Interviews sounded clear. Owners said they handled issue refunds and stock updates on the go. We made a clean flow. In tests, it did well. In the field, it gathered dust.
The gap was cadence and context. When did the job happen? What kicked it off? What got in the way? Interviews froze one moment in time. They missed the mess.
So we started again. We would not guess who did the job or how often. We would measure first.
Step 1: a fast, targeted survey
We wrote a seven-question survey. Job, frequency, where it happens, device, and who else helps. We sent it to active merchants. We tagged those who did task X at least twice a week on mobile. That became our pool. If you need a quick way to do this, PikoPM Survey can help.
We picked the highest-risk slice. One task where failure would hurt: confirm fulfillment while moving stock. Not the whole order flow. Just the fragile step.
Step 2: a five-day diary, built on triggers
From the validated pool, we asked for five days of entries. Each entry asked: time, place, what kicked it off, what device, what got in the way, and a photo or screen recording. We prompted at fixed times to catch misses too.
Patterns showed fast. The trigger was not a menu label. It was a ping from a courier, a customer text, or a teammate’s note. Interruptions were constant. Hands were full. One thumb was the norm.
Diary entries made it obvious: triggers drive action, not our IA. If we ignored the trigger, we would miss the moment.
Step 3: unmoderated tests from diary scenarios
We turned diary triggers into tasks. “You get a courier delay. You have one minute before the next customer.” We built light prototypes of only those flows. Participants recorded their screens at their own shop, with noise and interruptions. We measured completion, backtracks, and bail-outs.
Good lab scores had fooled us before. These tasks did not. When a hand was wet or gloved, swipe targets were wrong. When a phone was on a counter, reach zones changed. We fixed the flows to match those moments.
Step 4: check against telemetry
Before we set scope, we pulled event-level usage. How often did the task fire on mobile now? Where did sessions drop? Did diary cadence match? In our case, the peaks lined up with delivery windows and store open/close. That gave us confidence on timing and entry points.
Step 5: ship a pilot that mirrors diaries
We did not ship the polished flow. We shipped the messy one. Quick entry from a notification. Big tap targets in the bottom zone. A save that worked offline for 10 minutes. We limited the pilot to merchants whose diaries matched the pattern. Support knew the contexts. We watched task time, retries, and next-day edits.
It worked because it bent to the day we saw, not to the day we wished.
A two-week plan you can run
Worked example
- Day 1: Send a short survey to segment merchants who do task X at least twice per week on mobile. Recruit from those who say yes.
- Days 2–9: Run a five-day diary. Ask for time, place, trigger, device, workaround, and a photo/screen recording. Prompt at two fixed times daily.
- Days 10–12: Build unmoderated tests from the most common diary-triggered scenarios. Participants record screens in their real setting.
- Day 13: Compare findings with event-level telemetry. Write a one-page decision: launch a constrained pilot, iterate, or drop.
- Day 14: If pilot, set success metrics tied to diary contexts. Choose a small, matched cohort.
Methods at a glance
| Method | Best for | Risk if used alone |
|---|---|---|
| Interviews | What and why in one moment | Miss cadence and context; overfit words |
| Survey | Who does the job and how often | Thin on depth; may hide edge cases |
| Diary study | Cadence, triggers, workarounds | Anecdotes can mislead on prevalence |
| Unmoderated test | Flow fit in real settings | Can miss intent without setup |
| Telemetry | What actually happens at scale | Explains what, not why or when |
Common mistakes
What it taught me
The point: if you want a feature to fit a day, design from cadence and triggers, not from interviews alone. Use a short survey to find the doers, a tight diary on the riskiest task to capture context, unmoderated tests from those moments to shape the flow, and telemetry to size the bet. Then pilot in the same contexts you saw.


