2025 Annual Report

QuitMate is a community app for people working to recover from addiction. It spans 12 categories: gambling, alcohol, overeating, porn, tobacco, drugs, social media, shopping, gaming, caffeine, codependency, and custom goals.

This report aggregates behavioral data observed in the app over the 12 months from January 1 to December 31, 2025, to show what is happening in the QuitMate community. It proceeds in order: who showed up (the population), what happens during an attempt, when attempts break, how far people get, how that compares with the outside world, and how things differ by category. The community is based in Japan, so these findings reflect Japanese users.

Period

Jan 1, 2025 00:00 – Dec 31, 2025 23:59 (JST)

App

QuitMate (iOS / Android). 99%+ of posts are in Japanese; users are predominantly based in Japan.

Data source

In-app behavioral logs (posts, cheers, comments, start/end of attempts, AI-derived mood and craving scores, and more)

Scope

4,858 users active at least once in 2025, and 4,040 users who newly registered in 2025

Streak cohort

Of the 1,248 users active in December 2025, the 1,143 who started at least one attempt within 2025. Limited to those whose streaks can be observed.

Exclusions

Developer test accounts, official/admin accounts, obvious spam

"Longest streak"

The longest abstinence streak a user has ever achieved. Each streak runs from start to end (for ongoing attempts, capped at the user's last login or 2025-12-31, whichever is earlier). Self-reported start dates before service launch (Aug 29, 2023) are clamped to the launch date.

This report is observational. It does not establish that using QuitMate causes recovery from addiction. See Limitations at the end for all caveats.

Nearly doubled in a year

QuitMate's user base expanded clearly in 2025. New sign-ups rose +62%, from 2,487 the prior year to 4,040, and cumulative registrations reached 6,824 by year end.

What people are trying to quit

In 2025, 3,860 new "habits to quit" were registered. The breakdown by addiction category:

Gambling stands out at roughly 60%. QuitMate is designed to span all 12 categories, but in 2025 the user base skewed heavily toward people quitting gambling. Overeating, alcohol, porn, social media, and tobacco follow, broadly matching the categories most often raised at addiction support services in Japan.

How many stay

This report is about what happens to these people next. As context, here is how well 2025 sign-ups stuck around (registered Jan–Sep 2025, n=3,093; share still active at each point).

About half (47.8%) are still active 30 days after sign-up, and one in three (35.2%) at 90 days. The cohort used to compute "longest streak" and other figures below consists of these retained users; it does not represent people who left soon after signing up, who are excluded from the outset (Limitations 1).

Mood rises, cravings fall

While an attempt continues, what happens to people? All 49,398 posts in 2025 carry an AI-derived mood (1–10) and craving (0–10) score. Aggregated by the poster's days into their attempt, a clear trajectory appears.

On day 0, the average is mood 3.4 and craving 6.0; by 3–6 months in, they shift to mood 4.9 and craving 4.9, a change of about +1.5 in mood and -1.1 in craving. The improvement is clearest in the first 3–6 months. Beyond six months the number of users still posting drops into the low hundreds and the trend becomes unstable, so this chart is capped at six months. Per-bucket user counts are shown below the chart.

This data cannot separate whether continuing improved mood, or whether only those who improved continued. It is an observed association, not causation (Limitations 4).

People speak up most in the first week

Splitting 2025's 49,398 posts by days into the attempt, the first week dominates.

The most posts land in "Days 1–7," with 13,986 for the year. Adding day 0 through day 30, 67.5% of all posts are written within the first month. During the period of highest relapse risk, people are not silent. Each post draws an average of 16.0 cheers (788,615 for the year), so a post is not something that scrolls by unseen.

9 PM, Sundays, and the 15th

In 2025, 11,181 resets (attempt endings) were recorded. Looking at when they happen, patterns emerge on three axes.

Both posts and resets peak at 9 PM

Both posts and resets peak at 9 PM (21:00), followed by 10 PM and 8 PM. The hours when cravings hit hardest, when the community is most active, and when attempts most often end all land at the same time.

Hold out through Friday, break on the weekend

Sunday is highest at 1,882, then Monday at 1,726. Friday is lowest at 1,414. The pattern: hold out through Friday, slip over the weekend, and record the reset on Sunday or Monday. Weekends bring more exposure to the addiction (gatherings, solitary downtime, release from stress).

A small spike on the 15th

The 15th sees 424 resets, above the monthly average (~360). Many companies in Japan pay salaries on the 15th, which may coincide with a window where spending on the addiction (gambling, shopping, drinking) becomes easier (causation cannot be established).

The warning signs show up in the data

Beyond "when," we can also see the "lead-up." For attempts that were reset in 2025, here is the average craving / mood of posts in the days before the reset.

Eight or more days before a reset, craving sits at 5.18 and mood at 5.12; just before the reset, craving reaches 5.99 and mood drops to 3.96. Craving climbs and mood sinks as the breaking point nears. The state just before a reset is almost identical to day 0 of an attempt (craving 5.98, mood 3.67): right back to where it started.

Higher craving in posts just before a reset is partly expected (people write when they are wavering). This is shown as an observed warning sign, not causation (Limitations 6).

Median longest streak: 66 days

So, after cycles of attempt and restart, how far do people get? Narrowing to the 1,143 people who were active in December 2025 and started an attempt within 2025, we aggregate each person's longest streak ever. Observed streaks are capped at each user's last login.

Median

66d

75th pct

163d

90th pct

358d

Longest observed

856d

More than 1 in 10 (9.8%) have, in some past attempt, sustained a year or more.

About 8 attempts per person

How many attempts does it take to reach that record? Per user, the median is 2 attempts and the mean is 8.1 (a few heavy users pull the average up; the maximum is 417). Typically, after several resets, the longest attempt happened to last about two months. Relapse is not failure; it is recorded as the starting point of the next attempt.

Streaks running right now

"Longest streak" is a past peak and may not be ongoing. Because QuitMate starts a new attempt the moment one is reset, every active user always has a "current streak." For the 1,143, here is what is ongoing as of 2025-12-31.

1+ month ongoing

604ppl

3+ months ongoing

351ppl

6+ months ongoing

198ppl

1+ year ongoing

89ppl

Currently sustaining a year or more: 89 people (7.8%); six months or more: 198 (17.3%); one month or more: 604 (52.8%). More than half the cohort is currently holding a streak of at least a month.

On par with self-help and treatment

Where does this sit relative to the outside world? Below, the share of users "currently sustaining 1+ year" by category (n≥30) is placed next to reference figures reported in the literature for self-help and treatment programs.

Reference column: AA/TSF from Project MATCH (1997; 24% continuous abstinence over one year in the outpatient Twelve-Step Facilitation arm; the Cochrane Review, Kelly et al. 2020, also supports AA/TSF raising abstinence, RR 1.21); NRT from smoking-cessation trials (12-month continuous abstinence, roughly 13–19%); GA from Stewart & Brown (1988; ~8% abstinent at one year). Study populations, follow-up windows, and definitions of "abstinence" differ, so these are not a like-for-like comparison with QuitMate's figures (share currently sustaining 1+ year as of December 2025).

For alcohol, tobacco, and gambling, the continuation rates observed among QuitMate users sit in the same range as reported outcomes for self-help groups and standard treatment programs. That an anonymous online community shows comparable figures suggests QuitMate may play a complementary role for people who find it hard to access in-person support (no causal claim; see Limitations).

How "quittable" a habit is shapes the odds

Finally, how does the picture change by type of addiction? Below are longest streaks by category for users active in December 2025.

* Categories with fewer than 20 users have wide confidence intervals; treat as indicative.

The 1+ year achievement rate is higher for addictions you can cut out of daily life, and lower for those with unavoidable daily exposure. Alcohol 20.4% and tobacco 18.6% versus overeating 4.5% and SNS 5.6%: roughly a fourfold gap. How feasible it is to "quit" at all shows up directly in the distribution of longest streaks.

Still, every category except caffeine (with just 7 users) has at least one user who reached a year or more. The achievement rate differs, but long-term streaks appear in every category.

In 2025, 4,040 people newly joined QuitMate and exchanged 49,398 posts and 788,615 cheers. Half of all attempts reset within two weeks, yet 1 in 10 people still going has reached a year or more in some past attempt. Things break most easily at 9 PM and on Sundays, and in the days before a reset, craving rises and mood falls as a warning sign. Relapse, across a mean of about 8 attempts, gets woven into the next try.

Recovery from addiction is not linear. Through cycles of attempt and restart, a longest record accumulates. This report is one year of that log.

Limitations of this report

1. Self-selection bias. QuitMate's users skew toward people who voluntarily downloaded the app and kept using it. In particular, figures like "longest streak" are limited to retained users (cohort of 1,143) and do not represent the full population of people attempting recovery.
2. Self-reported data. Start/reset of attempts, days elapsed, and category are all user inputs. There is no objective verification such as substance testing.
3. Language and region. 99%+ of 2025 posts are in Japanese, and nearly all users are estimated to be in Japan. These findings cannot be generalized to other languages or cultures.
4. No causal claim. This report does not claim QuitMate use causally contributes to recovery. It describes usage patterns only.
5. Multiple addictions / category counts. When a user has several addictions, each is registered as a separate "habit to quit," so per-category counts reflect registrations, not people.
6. Interpreting mood / craving and warning signs. The Section 02 by-elapsed-days mood/craving figures are per-user averages (the mean across users in each bucket), not within-user before/after comparisons; beyond six months the number of users drops into the low hundreds and the trend is unstable, so the chart is capped at six months. The Section 03 by-time-until-reset figures are per-post averages. In all cases we cannot separate whether continuation improved outcomes, only improvers continued, or a "people post when wavering" bias. These are observed associations and warning signs, not causation.
7. Cutoff in "longest streak." Counting attempts of users who stopped using the app without recording a reset would make the records of people who left grow indefinitely. To avoid this, each streak's end is capped at the user's last login or 2025-12-31, whichever is earlier.
8. Self-reported start dates. Some users report a "quit start date" before service launch (Aug 29, 2023). Since these are self-reported past dates rather than in-app observations, the longest-streak calculation clamps the start to the launch date.
9. "Current streak" and app design. "Currently sustaining 1+ year" and similar in Section 04 use all active users (1,143) as the denominator, based on QuitMate automatically starting a new attempt when one is reset. "People with an ongoing attempt" could be defined other ways.
10. Methodological differences in literature comparison. The self-help / standard-treatment continuation rates referenced in Section 05 differ by study in population, follow-up window, and definition of "abstinence" (complete abstinence vs. harm reduction). Placing them alongside QuitMate shows figures "in the same range," not a strict equivalence.
11. Effect of operator interventions. Push notifications, media exposure, and UX changes during 2025 may have affected figures on specific dates or categories.