Do self-improvement apps actually work?
Some do, for specific problems, by small amounts — and almost nobody uses them long enough to find out. The largest constraint is not effectiveness but abandonment: across mental-health apps with more than 10,000 installs, Baumel and colleagues (2019) found a median 30-day retention of 3.3%, meaning roughly 3 in 100 people who install one are still opening it a month later. Apps with narrow, structured targets tend to show measurable effects in trials, while broad "become a better person" apps mostly show engagement rather than outcomes.
The retention number is the headline
Baumel and colleagues measured actual usage through a panel rather than asking people, across mental-health apps in the Android store with over 10,000 installs. Median retention was 3.9% at 15 days and 3.3% at 30. Categories differed — peer support held about 8.9% at 30 days, mindfulness about 4.7%, and standalone breathing-exercise apps a median of 0.0%.
That reframes the question. An intervention that works in a trial where participants are supported and paid may still deliver nothing in the wild, because the wild version is uninstalled by week two. Any claim about an app that does not account for whether people keep using it is describing a laboratory, not a product.
What tends to work
- Narrow and structured. A programme aimed at one thing, with an order to go through, does better than an open library of content.
- Something to do rather than something to read. Interventions that ask for a behaviour outperform ones that deliver information.
- Feedback tied to your own data. A response that changes based on what you actually did is different in kind from a response that would have been the same for anyone.
- Human contact somewhere in the loop. Guided digital interventions consistently outperform fully self-guided ones, and the gap is not small.
What to be sceptical of
- Effect claims with no comparison group. "Users report feeling better" describes people who chose the app and stayed, which is the group most likely to report improvement whatever the app did.
- Engagement presented as outcome. Sessions, streaks, and minutes are inputs. They are reported because they are large.
- "Science-backed" as a bare phrase. The useful version names the study and lets you check what it measured and on whom.
How to decide for yourself
The general question does not have a useful answer, because the population effect of a category is not the effect on you. The answerable version is narrower: does this specific app change something specific in your life, measured before and after. That is a self-experiment, and it takes a few weeks.
Where Emerge fits
Emerge is one of these apps and the numbers above include its category. What it does differently is refuse the engagement layer — no streaks, no scores — and build in the measurement instead: Your Lab runs a baseline-then-change comparison so the app can tell you it did nothing for you.
Emerge on the App Store →Related questions
What percentage of people keep using a mental health app?
About 3%. Baumel et al. (2019) measured a median 30-day retention of 3.3% across mental-health apps with more than 10,000 installs, using panel data on real usage rather than self-report.
Are paid self-improvement apps better than free ones?
Price is not a quality signal on its own, but it changes the business model. A subscription app is paid whether or not you open it daily; a free ad-supported app earns more the longer it holds your attention. That difference shows up in the design.
Sources
Last updated 2026-09.