Research paper · RP02

The Attrition Cliff

Background. Digital wellbeing programmes reach millions of working adults. Most of those adults stop opening them within a fortnight. Baumel and colleagues reported, in 2019, a median 15-day retention of 3.9% and a 30-day retention of 3.3% across 93 mental-health apps. This review asks whether that floor still stands, whether trial dropout is the same construct, and which design features — reminders, human support, visible progress, channel, language — bend the curve for adults using apps, web tools or messaging programmes. Methods. PRISMA-2020 systematic review of PubMed, PsycINFO, Scopus, Web of Science, Google Scholar, CTRI, ClinicalTrials.gov and OSF preprints, 1 January 2019 to 22 September 2026. Eligible records described adult users of digital wellbeing or mental-wellbeing programmes and reported retention, engagement, attrition or dropout. Real-world open rates and trial assessment-dropout were pre-specified as separate constructs. A new random-effects meta-analysis of day-15 and day-30 same-day open rates was planned if five or more independent studies reported extractable events and denominators on that exact endpoint. They did not. Narrative synthesis was therefore used. Trial-side figures are taken from three published random-effects meta-analyses and are not re-pooled here, to avoid double-counting overlapping trials. Results. The only independent multi-app panel that reports same-day open rates remains Baumel 2019: median day-15 retention 3.9% (IQR 10.3%) and day-30 retention 3.3% (IQR 6.2%) across 93 apps. No second panel study from 2019–2026 overturns that floor. Rolling-retention cohorts sit higher because they count any later use, not an open on that calendar day. Workplace and commercially guided programmes report day-30 retention near 45–53%. Trial assessment-dropout is a different number: 26.2% (95% CI 18.1–36.3) in 18 depression-app trials; 24.1% short-term and 35.5% longer-term across 70 smartphone trials; 18.6% post-test and 28.4% at follow-up across 79 later trials. Reminders and human contact are associated with lower trial dropout. Visible progress counters do not have a verified pooled lift; gamification has been associated with higher trial attrition. No Indian consumer programme has published a day-15 or day-30 same-day open-rate. Conclusions. The day-15 floor of roughly 4% is still the best independent estimate for unguided store apps. It is not universal. A human hand-off and a workplace channel move the curve by an order of magnitude. Streaks do not. Registration. This review was not pre-registered. Search strings, eligibility and the fallback rule were specified in the commissioning protocol before extraction.