Diary Study in India: A Practical Design for Repeated Self-Reporting

The unit of analysis is the entry, not the person. Fifty respondents over fourteen days can produce seven hundred entries, which is enough to describe patterns within a respondent's week.

Entries can be time-based, event-based or interval-based. A time-based diary prompts at fixed moments, such as the end of each day. An event-based diary asks for an entry whenever a defined thing happens, such as a purchase or a service complaint.

Event-based diaries capture what matters but depend on the respondent noticing and remembering to log it. Time-based diaries are more complete but can miss the moment.

Pick one and say why in the methods section. Switching logic mid-study makes the entries incomparable.

Use a diary when the sequence matters: what happens before a skipped dose, a late payment or an abandoned cart. A survey can tell you that it happened. A diary can show what came just before.

Use a survey when you need a stable estimate for a population. Diary samples are small and self-selected, so they are poor at estimating prevalence.

Keep each entry under three minutes. A diary that takes ten minutes will be completed for two days and abandoned on the third.

Ask a fixed core every time, plus one rotating question. The fixed core makes trends visible. The rotating question keeps the task from becoming mechanical.

Write the prompt in the language the respondent uses on their phone, and keep the answer formats to taps and short lists. Long free text on a phone produces short, unhelpful entries.

Include a timestamp and a simple context field, such as where they were or who they were with. Context is what turns a log into an explanation.

Test the whole sequence on paper first, then on the phone, with five people who match the sample.

Recruit for compliance, not only for fit. The best respondent is the one who will still be answering on day ten.

Set expectations at entry: how many days, how long each entry takes, what happens if a day is missed, and how to stop. People drop out when the task turns out to be different from what they agreed to.

Send reminders at a consistent time, and stop reminding after the study ends. A reminder at the wrong hour becomes the reason someone quits.

Plan for attrition in the sample size. If you need forty completed diaries, recruit more than forty and track completion daily.

Keep a study log of who missed which day. Losing one heavy day can bias a week.

Pay for completion, not for signing up. A split incentive, part on entry and part on finishing, protects both the respondent and the study.

Explain what is collected and who sees it. Diary studies collect routines and locations, which is more sensitive than a one-off survey.

Ask for consent at entry and again if you add a new field mid-study. Being in a study on day one is not consent to a new question on day eight.

Give people a way to withdraw and to ask for their entries to be deleted. Keep the retention period short and tell respondents what it is.

Start with completion, not content. How many people started, how many finished, which days lost the most entries. Drop-off is a finding about burden as much as about the topic.

Then describe the sequence. Plot the fixed core across days for each respondent, and look for runs, reversals and steady states rather than single points.

Code the rotating question into a small set of categories, and keep one category for entries that do not fit. That category is usually where the insight is.

Compare entry and exit. A change in language or in what people notice is often more informative than the average of the whole period.

Do not report diary findings as population estimates. Report them as patterns within the studied group, with the number of respondents and entries.

The first failure is burden. Too many questions per entry, too many days, and a completion rate that makes the last week meaningless.

The second is reactivity. People change the behaviour they are recording, especially when the diary is about a habit they are trying to change. Say so, and look for the change in the entries.

The third is an entry window that is too wide. Asking people to recall the whole day at midnight produces a summary, not a diary.

The fourth is analysis that ignores time. If the report averages all entries, the study has thrown away the reason it existed.

Build the diary as a short survey and send it on a fixed daily schedule. Keep the core to a handful of taps and put the rotating question at the end.

Pilot with the free 30 responses, then check completion by day before you commit to a longer field period.