Designing behavioural prompts that work
Most digital trials include reminders. They prompt participants to complete diaries, attend visits, or take supplements. But not all reminders are equal. Some quietly encourage behaviour. Others irritate or confuse. And a few do the opposite of what they were meant to do.
The difference usually comes down to two things: how the prompt is written, and whether it functions as a notification or a nudge. These are not the same. A notification is a message. It provides information, neutrally and directly: "Your diary entry is now available." "Reminder: complete your Day 10 questionnaire." These are necessary, but they rely entirely on the participant to act without added motivation. A nudge is a design choice. It makes the desired action easier, more appealing, or more automatic. An analysis of how Apple Watch notifications function found that the device turned what could easily have been distractions into genuinely positive wellness tools by pairing alerts with personal context: goals the user had actually set, and nudges triggered by a relevant moment rather than a generic schedule. The most effective nudges feel like part of the process, not an interruption, and that distinction between generic and contextual is doing most of the work.
The difference in practice
Here is the same situation handled two ways. Notification: "You haven't completed your last three diary entries." Nudge: "Thanks for what you've submitted so far. Your next diary entry is ready when you are." The second avoids shame. It invites the participant back in without scolding them. In long studies especially, tone matters more than most teams realise.
Beyond messaging, a nudge can live in interface design too: a progress bar showing how close the participant is to finishing their weekly tasks, a pre-filled form using yesterday's value with the option to change it, a calendar that marks completed days and builds a habit loop without words, or a task that links directly to the right screen, skipping the home screen entirely.
What the evidence says about personalisation and timing
A factorial randomised trial run within the Framingham Heart Study's digital cohort tested exactly these variables directly, rather than relying on intuition. Across 655 participants, personalised notifications, ones that used the participant's name and referenced whether their data had actually come through the previous week, produced a meaningfully higher rate of continued data transmission than standard, generic ones: 38.9% versus 28.8% at three months, widening to 36.8% versus 22.4% at six months. Personalisation wasn't a nice-to-have. It was the single strongest lever the trial tested.
What the same trial did not find is just as useful to know: sending notifications on a weekend versus a weekday, or in the morning versus the evening, made no measurable difference to whether participants responded. That's a genuinely useful correction to the common assumption that exact send-timing is the main lever worth optimising. It's worth testing timing for your own population if you have the data to do it, but don't spend more design effort on timing than on making the message itself feel like it's actually about the person receiving it.
The right frequency
Too few prompts and participants drift. Too many and they switch off. Some studies benefit from a brief escalation path:
- Send the initial reminder at the usual time
- If there is no response after 24 hours, send a gentler, personalised follow-up
- If still no response, try a different channel, such as SMS if the app push was ignored, or vice versa
The goal is not to maximise message volume. It is to reduce the number of things standing between the participant and the next completed task, using the lever the evidence actually supports: making the message feel personal, not just well timed.