Objective User Engagement With Mental Health AppsSystematic Search and Panel-Based Usage Analysis

Open your phone and you'll find a universe of mental health apps—breathing timers, mood journals, meditation guides, and peer communities. The promise is obvious: help at your fingertips, anytime. And the scale is staggering. As Baumel, Muench, Edan, and Kane remind us, most adults now carry a mobile device, and a World Health Organization scan from the mid-2010s counted tens of thousands of health apps, with a significant portion aimed at mental health. But here's the rub. In the wild, where there are no reminder calls from a researcher and no gift cards for filling out surveys, people often stop using these tools quickly. We've seen this story before. MoodGym, one of the classic web programs, had completion rates under one percent. Across web and app programs for depression and anxiety, a broad review found only a minority of users—somewhere between roughly a tenth and under half—stuck with them enough to count as moderate use. Even for respected offerings like PTSD Coach, about four in ten were still around after a month and only about two in ten after six months. On another platform, Happify, just a few percent completed a six-week assessment. These aren't flukes; they're a pattern. So the question becomes less "Do these approaches help under ideal conditions?" and more "What do people actually do when they download these apps on their own?" That's the lane Baumel and colleagues chose. Rather than rely on developer dashboards or small, incentivized trials, they went to an independent panel that tracks anonymized, aggregated app usage at scale. They started with a systematic search of Google Play in late 2018, using broad terms—depression, anxiety, stress, mood, and happiness—and cross-checking against curated lists like PsyberGuide. To make sure they were looking at apps with a real user base, they set a threshold: at least ten thousand installs. They kept it unguided—no therapist in the loop—and focused on English-language apps that targeted self-management of mental health or emotional well-being, not just general fitness. Out of that sweep came ninety-three apps. If you picture the install counts as a stack, the one in the middle had about one hundred thousand installs, which gives you a feel for the scale without getting distracted by the mega-hits at the top. Two trained raters coded what each app primarily did—mindfulness or meditation, tracking and journaling, psychoeducation, peer support, or stand-alone breathing exercises—and also tagged the mental health focus, like anxiety problems, depression, emotional difficulties, or a general happiness tilt. Agreement between raters was very strong, with a reliability coefficient near perfect, which is a fancy way of saying they were almost always seeing the same thing. The usage data came from SimilarWeb's panel, which pulls anonymized signals from a lot of places to estimate engagement. The window was from August 2017 through July 2018, and the platform was Android. Three simple metrics did most of the work. First, the daily open rate: on a given day, out of everyone with the app installed, what percent opened it? Second, activity among those who did: how many sessions did the average active user have, and how many minutes did they spend that day? Third, retention: of those who opened the app on the day they installed it, what percent came back on day one, day fifteen, and day thirty? A subset of larger apps also revealed when during the day people used them. Before getting to results, a quick note on the math. App engagement is lumpy—some apps have rabid fans, while others sleepwalk—so the distributions aren't nice bell curves. The team leaned on nonparametric tests that don't assume normality: Kruskal-Wallis to see if groups differed, then Mann-Whitney to zero in on which pairs diverged. They also checked whether the massive, most-installed titles were skewing things and found only a limited effect. To sanity-check the panel itself, they cite validation work showing strong correlations between SimilarWeb's measures and independent usage records, as well as with the install brackets published on Google Play. So what does real-world behavior look like across those ninety-three apps? At baseline, it's modest. The median daily open rate was four percent. That means on a typical day, ninety-six out of a hundred people who had installed a given app didn't open it. Among those who did, the average active user had around three sessions and spent about thirteen minutes total that day. Those are ordinary users going about their lives, not people being nudged by a study team. It's a grounded starting point and it immediately suggests the action isn't just, "Can we get people to download?" It's, "Can we give them a reason to come back and spend meaningful time?" Peel back the averages and you start to see why some apps hook people better than others. Of the ninety-three, about two-thirds targeted mental health problems broadly—things like anxiety-related issues and depression—while the rest leaned into general well-being or were framed around mindfulness and meditation. In terms of what they asked users to do, mindfulness and meditation were the single most common primary technique. Trackers, including diaries and mood journals, were close behind. Breathing exercises and psychoeducation were a smaller slice, and pure peer support—apps that primarily connected people to one another—were rare, with only two in the set. Some apps blended techniques, but the team focused on a primary category to keep comparisons coherent. Time on task diverged sharply by technique. Mindfulness and meditation apps stood out: their median active user time was about twenty-one and a half minutes a day, roughly double the overall median. That fits the use case. A guided session often lasts ten to twenty minutes, so when people do it, they stay. Peer support apps, while just two in number, came in even higher on minutes, in the mid-thirties, which makes sense if you imagine scrolling, posting, and responding. Trackers and stand-alone breathing tools clustered lower, under ten minutes. This isn't a verdict on effectiveness; it's a snapshot of how people actually spend time when they open the app. Open rates told a complementary story. Trackers and peer support apps were opened by a bigger share of their installed base—trackers around six percent on a given day, and peer support closer to seventeen—while breathing exercise apps lagged around one to two percent. Mindfulness sat near the overall median, around four percent, which is interesting: when people open those, they tend to stay longer, but they aren't necessarily opening them more often than other categories. The statistical contrasts were strongest when comparing everything to breathing-only apps. Trackers and peer support beat breathing exercises handily in daily opens, and mindfulness and peer support beat them on minutes. Now for the staying power. If you take everyone who opened the app on the day they installed it—call that day zero—and watch how many open it on each subsequent day, you see a steep early slide and then a long tail. About seven in ten were back on day one. By day fifteen, it was just under four percent. By day thirty, around three percent. That's sobering but not shocking in consumer tech, and it helps ground our expectations. What mattered more was how that curve differed by what the app actually did. When the team lined up thirty-day retention by technique, there were real gaps. Peer support had the highest median thirty-day return, just under nine percent. Trackers hovered around six percent, and mindfulness around five. Breathing exercises, by contrast, were at zero in the median case. Statistically, those differences across techniques were unlikely to be due to noise, whereas splitting the apps by mental health focus—depression versus anxiety versus general well-being—didn't yield meaningful retention differences. The when of use also varied. In the subset of apps with hourly data, tools aimed at mental health problems peaked in the evenings, around eight p.m.—that moment when the day's stress catches up. Mindfulness apps had a dual rhythm, with a morning bump between seven and nine and a late-night bump closer to midnight, mirroring how people slot meditation into their routines. Weekly patterns were quirky too; mindfulness use ticked up on Thursdays, which you can read a hundred ways or chalk up to habit before the weekend. One potential confound in any app analysis is the blockbuster effect. Perhaps the few giants have marketing muscle and that's where the engagement lives. Baumel's team checked that by comparing the top five most-installed apps to the rest. The big ones did have higher daily open rates—about nine percent versus four for the rest—but they didn't otherwise dominate on sessions or minutes. And those top five weren't all the same kind of app; one leaned on peer support and several were trackers, which helps explain why those categories popped in the broader analysis. It's important to be clear about what this study can and cannot tell us. It's Android-only, and while Android dominates globally, its share in the United States is lower, so we need to be careful generalizing to iPhone-heavy populations. The retention lens stops at thirty days; it doesn't tell you what happens in month three. The data are aggregated and anonymized, which is a strength for privacy but means we can't peer inside a given app to see which screens or prompts did the work. Peer support's eye-catching numbers come from just two apps, which urges caution. And techniques were coded at the primary level, so a mindfulness app that also offers psychoeducation still sits in the mindfulness bucket. Even with those caveats, the patterns are instructive. They suggest that the "what" of an app—its core activity—shapes not just how long people stay when they open it but whether they open it at all after the honeymoon period. Meditation and community invite longer sessions. Trackers and diaries seem to build daily engagement, nudging people to check in. And stand-alone breathing tools, which often perform a simple task very well, may be treated more like a fire extinguisher than a daily practice. That's not a failure; it's a design choice. But if your goal is sustained, month-long engagement without human guidance, the technique matters. For developers and clinicians, this kind of real-world benchmark is valuable. It tells you what "normal" looks like across a crowded marketplace, so you can ask better questions. If your meditation app gets seven minutes a day on average, you're not broken; you're below a peer median that's closer to twenty-one. If your tracker's daily open rate lags at two percent, you might borrow tactics from the stronger performers—richer prompts, easier logging, streaks that actually motivate rather than shame. And if you're evaluating impact, you now have a sense of the denominator: how many people you can expect to see by day thirty given the technique you've chosen. There's one more reason to care about this. Trials are essential for testing whether an intervention works under controlled conditions, but public health lives in the messy middle—people download, they dabble, and life intervenes. Baumel and colleagues built a bridge into that world with independent, convergently validated usage data, and they set simple, interpretable yardsticks: daily opens, minutes among active users, and thirty-day return. That clarity allows different teams, working on different problems, to compare themselves honestly without waiting for a randomized trial to finish. Where does this go next? Carefully, and with humility. The study points to techniques that, on average, retain people longer, but it doesn't tell you what will hook a specific person. One person's perfect morning meditation is another's guilt machine. The obvious next step is better measurement across platforms and contexts, with privacy at the center, so we can track not just whether someone opened an app, but whether the cadence matched their needs. And yes, smarter matching—steering people toward the techniques that have higher real-world retention for people like them—seems within reach if the data infrastructure is there. For now, the takeaway is straightforward. Unguided mental health apps do get used, but sparingly, and their staying power hinges on the core activity they invite. Mindfulness and community stretch time spent per visit. Trackers encourage people to check in more days. Breathing tools serve as quick relief, not long-term practice. If we want digital care to scale beyond the install, we have to design for the rhythms of real life—and measure those rhythms in the wild.