All work

Work — Interaction Craft

Fixing how the NHS App asks for notifications

Design LeadNHS App, Notification Strategy2025
The NHS App notification opt-in prompt, shown on a phone held in someone's hand

38M

users

470+

research participants

28%

turn notifications on

The ask: more opt-ins without losing people’s trust

Millions of people use the NHS App and never see the updates that matter, because notifications are off.

The brief was to increase opt-ins and cut the cost of letters and SMS. The harder question sat underneath it. How do you ask 38 million people for permission without damaging the reason they trust the app?

One prompt, first login, that was it

I started with why people had already said no.

The old prompt showed once, on first login, and never came back. Wrong moment, wrong wording, and you’d lost them. Someone who would have said yes a month later never got asked again.

The brief said cut costs. Research said don’t mention that.

We ran research with over 470 participants: moderated and unmoderated testing, surveys, in-person sessions. People weren’t against notifications. They just weren’t sold on them at that moment.

Then the finding that mattered. When the prompt was framed around what the NHS needed, people felt manipulated.

That put the brief and the research in direct conflict. Cost saving was the reason the work existed, and saying so was the fastest way to lose the opt-in. So the prompt had to make the case for the user, and let the savings come out of that.

Three versions, tested nationally

I designed three framings and ran an A/B/C test through Adobe Target instead of picking one and arguing for it.

The test ran nationally, so we knew which framing worked before anything shipped.

Ask again every 90 days, not sooner

People who decline now get a revised prompt every 90 days, with clearer wording on what they get out of it.

Ninety days was the deliberate bit. Long enough that it never turns into nagging, short enough to catch people whose situation has changed. The quickest way to push this number up would have been to ask more often. I didn’t build that version.

Coded and tested for accessibility

Every journey was built in HTML for validation and tested for accessibility, so it works across the full range of people using the app.

I simplified the notification settings page at the same time, so anyone who opts in can manage it without hunting.

Live, converting at 28%

More than one in four people who see the prompt turn notifications on.

More people getting health information they actually need.