Health apps lose people in four days

A Stanford study kept its participants for an average of 4.1 days. That's the real reason Aneroid only asks for one tap.

The Stanford study MyHeart Counts kept participants in the app for an average of 4.1 days. Not 4.1 weeks. Four days and a bit. It's one of the better-funded health studies built around a smartphone app as its tool, and even that couldn't hold on to people any longer than that.

71% are gone within three months

This isn't a one-off mishap. A literature review in the Journal of Medical Internet Research estimates that around 71% of app users in general have dropped off within 90 days. Substantial drop-off isn't just an unpleasant figure — it weakens a study's power and threatens the representativeness of the sample, because those who remain are rarely a random selection of those who started.

Even the British pain study lost more than half

Cloudy with a Chance of Pain — the British study that got people to download an app and log their pain against the weather — is one of the best examples of how hard this is, even when the project is serious and well designed. The app was downloaded by 13,207 users over twelve months of recruitment, from all 124 British postcode areas — broadly based, not a small local trial. Participants were predominantly women (81%) with an average age of 49, and had a mix of diagnoses: arthritis, rheumatoid arthritis, fibromyalgia and other chronic pain conditions.

10,584 had complete baseline data and at least one entry. 6,850, or 65%, were still on board after the first week. After the first month it was 4,692 — 44%. In other words: more than half of those who had a usable start were gone again within four weeks.

There's even a study with the same title — Cloudy with a Chance of Pain — that specifically examines engagement and drop-off in daily data entry on weather, disease severity and physical activity in patients with rheumatoid arthritis. The fact that someone found it necessary to run a study purely about the drop-off is itself a testament to how big the problem is considered to be in this field.

The literature review also points to what tends to help: reminders in the form of push messages and texts, along with the option of contact with clinicians, are expected to increase retention. Aneroid has neither, in the traditional sense — only the questions the app itself raises after a pressure change, or three times a day for no reason at all. That's a deliberate trade-off against what the literature says works, and it's worth saying out loud, not hiding away.

The real threat to a project like this

It's tempting to think that the risk to an app like Aneroid is about whether the hypothesis holds — whether air pressure actually connects to how you feel. That isn't the real risk. The real risk is whether anyone bothers to stick around long enough for the question to be answered at all. A connection that only shows up after weeks or months can't be found in anyone who's gone after four days.

That's why the check-in in Aneroid is one tap, with no menus and no extra steps. That's why there's no diary, no triggers to enter, no fields to fill in. Every obstacle between a feeling and an answer is an extra reason to give up — and the dataset the app needs to say anything at all about your own connection only exists if you're still answering in three weeks' time.