App design

Most people drop a mental health app within two weeks. Texts and rewards kept more smokers going.

Of new users of 59 popular Android mental health apps, a median 3.9% opened one again on day 15. In smoking trials, automated text messages raised quit rates, while apps on their own showed no clear effect. Putting money down worked for the few who agreed to it, but plain rewards reached far more people and helped more smokers overall.

Most people stop within days

In November 2018 Baumel and colleagues searched Google Play for apps on anxiety, depression and emotional wellbeing with 10,000 or more installs¹. They included 93, and 59 of them had data on whether users came back¹. The data came from a panel of phones whose owners shared their app use, from August 2017 to July 2018¹.

The day after people first opened one of these apps, a median 69.4% opened it again¹. By day 15 it was 3.9%¹. By day 30 it was 3.3%¹. Nearly all of the fall came in the first two weeks, and after that there was little left to lose. On a typical day, a median 4.0% of the people who had an app installed opened it¹. Having an app on the phone and using it turned out to be very different things.

Median share of new users who opened the app on each day after first opening it. 59 Android mental health apps with 10,000 or more installs, August 2017 to July 2018.

Baumel, Muench, Edan and Kane, Journal of Medical Internet Research, 2019. Usage from a panel of phones whose owners shared app data.

The panel is anonymous, so the paper cannot say who these users were or why they stopped¹. It covers Android only and nothing past day 30¹. Opening an app on a given day also says nothing about whether it helped¹. These were apps for anxiety, depression and wellbeing, not for porn use. Nothing in the data suggests an app for porn use would hold people any better.

Trials make it look better than it is

A 2018 review by Fleming and colleagues came at the same problem from another side². It looked at digital self-help for depression, low mood and anxiety once it left the trial and reached real users². Only 2.8% of teenagers using MoodGYM finished all 5 modules². 19.5% of the 267 people who registered for CBTPsych, a program for social anxiety in people who stutter, finished it². In controlled trials of online programs, an earlier review it cites had found completion of 43% to 99%².

That gap is worth keeping in mind whenever someone cites a trial of an app. A trial result says what a program can do for the people who stay with it. The real-world numbers say how few do¹². A program nobody finishes cannot help much, however good its last module is.

So if most people stop within days, what keeps them going? The best tested answers come from smoking, where quitting can be checked with a breath or saliva test³. Everything from here on is about smokers, not porn.

A text that arrives on its own

In 2019 a Cochrane review by Whittaker and colleagues included 26 trials of phone-based help for quitting smoking, found up to October 2018³. Every trial followed smokers for at least 6 months³. Where a trial confirmed quitting with a breath, saliva or urine test, the review used that over what people said. Anyone who dropped out counted as still smoking³.

Automated text messages, against minimal support, raised the share who had quit when checked 6 to 12 months later from 6 in 100 to 9 in 100³. That result came from 13 trials and 14,133 people³. Texts added to other help did about as well, 12 in 100 against 8 in 100, across 4 trials and 997 people³. Three extra quitters per 100 sounds small³. Spread over thousands of smokers, it is not.

Smokers who had quit when checked 6 to 12 months later, per 100, in three comparisons from the Cochrane review. The per 100 figures are the review's own.

With phone support
Comparison

Whittaker, McRobbie, Bullen, Rodgers, Gu and Dobson, Cochrane Database of Systematic Reviews, 2019. Trials found up to October 2018.

One trial in the review, txt2stop, shows what that looked like in practice⁴. It randomised 5,800 UK smokers, and the intervention group got five texts a day for 5 weeks, then three a week for 26 weeks⁴. At 6 months, 10.7% were confirmed quit by saliva test, against 4.9% of those sent texts unrelated to quitting⁴.

The saliva test mattered. 150 of the 542 people in txt2stop who said they had quit, 28%, were still smoking⁴. A study that took people at their word would have counted every one of them as a success. Self-help tools that report what their users say about themselves deserve the same doubt.

Apps on their own

Smartphone apps did not show the same effect. In 5 trials with 3,079 people, 8 in 100 quit with an app and 8 in 100 with less intensive support³. The review rated that evidence very low certainty. Its confidence interval runs from a clinically significant harm to a clinically significant benefit, so it cannot rule an effect in or out³.

Thin evidence does not prove that apps fail. But set it next to the drop-off numbers and a pattern shows. A text arrives whether or not the person opens anything. An app waits to be opened, and by day 15 almost nobody opens it¹³. That is our reading of two separate studies, one on smokers and one on mental health apps. Neither one tested it.

The text trials had problems of their own. They disagreed with each other more than chance explains³. Most ran in high-income countries, and the review's search ended in October 2018³.

Putting money on the line

Messages are one way to keep people going. Money is another. In a US trial published in 2015, Halpern and colleagues assigned 2,538 smokers at random to usual care or to one of four programs⁵. They were employees of one US company and their friends and family, in 2012⁵. Two programs paid about $800 in rewards for staying off cigarettes⁵. Two asked for a $150 deposit up front, refunded with $650 in rewards for those who quit⁵. Saliva tests checked quitting at 14 days, 30 days and 6 months, and anyone who sent no sample counted as smoking⁵.

The deposit looked like the winner at first. Among those who paid it, 52.3% were off cigarettes at 6 months, against 17.1% of those who took up a reward⁵. But only 13.7% of the people offered the deposit paid it, while 90% took up the reward⁵.

Deposit and reward programs for quitting smoking, each worth about $800. 2,538 smokers in the US, quitting checked by saliva test at 6 months.

Deposit
Reward

Halpern, French, Small and others, New England Journal of Medicine, 2015. Employees of one US company and their friends and family, 2012.

Count everyone assigned, whether they accepted or not, and the order flips. 15.7% quit in the reward programs and 10.2% in the deposit programs, against 6.0% with usual care, which included information, plus free aids to quit for the 41% with company health benefits⁵. The deposit worked well for the few who chose it. The reward reached far more people, so it helped more smokers overall⁵.

Those few may not be typical. People who chose to pay a deposit may have been readier to quit, so the 52% cannot all be credited to the deposit⁵. The authors' adjusted estimate, which allows for that, still put deposits 13.2 percentage points ahead among people who would accept either kind of program⁵. The gains also faded. At 12 months, about half of those who had quit at 6 months still tested clean, and only the reward programs still beat usual care⁵. Fewer people sent samples at 12 months, so the authors think true relapse may have been lower⁵.

An earlier trial ran into the same low take-up. Giné and colleagues had a bank in the Philippines offer smokers a savings account that went to charity if a urine test at 6 months found nicotine⁶. 83 of 781 offered it, 11%, signed up⁶. The offer raised the share who passed by 3.3 to 5.8 percentage points, and the gain held at a surprise test at 12 months⁶. The money came mostly bundled with weekly visits from the bank's deposit collector, and the authors could not tell which of the two helped⁶.

Two trials, two countries, and the same shape. A stake on the line works for the people who agree to it, and only 11% to 14% of those offered one did⁵⁶. Every number in this section comes from smoking, which a saliva or urine test can check. Nothing checks porn use that way, so a stake that depends on proof of staying off has nothing to pay out against.

How they found it

Baumel and colleagues searched Google Play for apps on anxiety, depression or emotional wellbeing with 10,000 or more installs and included 93. 59 of them had data on whether users came back.
A 2018 review of real-world use found that 2.8% of teenagers using MoodGYM finished all 5 modules, and 19.5% of the 267 registered users of CBTPsych, a program for people who stutter, finished it. An earlier review of controlled trials, which it cites, found completion of 43% to 99%.
The Cochrane review included randomised trials that followed smokers for at least 6 months. Where a trial confirmed quitting with a breath, saliva or urine test, the review used that over self-report. People who dropped out were counted as still smoking.
txt2stop randomised 5,800 UK smokers and sent the intervention group five texts a day for 5 weeks, then three a week for 26 weeks. 10.7% were confirmed quit by saliva test at 6 months, against 4.9% of those sent texts unrelated to quitting. 150 of the 542 who said they had quit, 28%, were still smoking by the saliva test.
Halpern and colleagues assigned smokers at random to usual care or to one of four programs. Two paid about $800 in rewards. Two asked for a $150 deposit up front, refunded with $650 in rewards for those who quit. Saliva tests at 14 days, 30 days and 6 months, and anyone who sent no sample counted as smoking.
A trial run with a bank in the Philippines in 2006 and 2007, published in 2010, offered smokers a savings account that went to charity if a urine test at 6 months found nicotine. 83 of 781 offered it, 11%, signed up. The offer raised the share who passed by 3.3 to 5.8 percentage points.

What it cannot tell you

The app usage panel is anonymous, so the paper cannot say who the users were or why they stopped. Android only, and nothing past day 30.
Opening an app on a given day says nothing about whether it helped.
The text message trials disagreed with each other more than chance explains. The app result rests on 5 trials, and its confidence interval runs from a clinically significant harm to a clinically significant benefit.
Most phone trials ran in high-income countries, and the review's search ended in October 2018.
People who chose to pay a deposit may have been readier to quit. At 12 months, about half of those who had quit at 6 months still tested clean, and only the reward programs still beat usual care. Fewer people sent samples at 12 months.
In the Philippines trial the savings account came mostly bundled with weekly visits from the bank's deposit collector, and the authors could not tell which of the two helped.
Mental health apps and smoking, not pornography. A saliva or urine test can check smoking. Nothing checks porn use that way.

In P-Minus

For its first 7 days, P-Minus shows one thing at a time, in a single card on Home. It sends at most one message a day, and none names the habit, because anyone can read a lock screen.

Sources

Sources checked October 2026.