1. Background
Reframe is a mobile program for adults who want to cut back on or quit drinking. It combines a daily curriculum grounded in cognitive behavioral therapy and neuroscience education with day-level drink tracking, in-the-moment craving tools, a moderated peer community, and progress feedback. Reframe is designed to support professional care, not substitute for it, and onboarding directs users showing signs of severe physical dependence toward appropriate medical care.
The platform's measured footprint at the time of analysis: 3.4 million registered accounts, 41.1 million tracked drinking days (24.4 million of them logged as alcohol-free), and 104.9 million completed program activities. This scale allows behavioral analyses on cohorts orders of magnitude larger than typical study samples, using logged behavior rather than retrospective recall.
2. Data and measures
Two data sources are used, both captured in the ordinary operation of the app:
- Onboarding self-report. During signup, users report their typical weekly drink count. This is the baseline measure.
- Daily drink log. Users log a drink count for individual days (one record per user per day). A logged count of zero is an explicit alcohol-free day; days with no record are treated as unobserved, never as alcohol-free.
Measure definitions and data quality guards| Weekly drink rate | (total drinks logged in a window ÷ days logged in that window) × 7 |
| Months 1 / 2 / 3 | Days 0–29, 30–59, and 60–89 after signup |
| Alcohol-free day | A day explicitly logged with zero drinks |
| Heavier drinking day | A logged day with four or more drinks |
| Data quality guards | Baselines outside 1–100 drinks/week excluded; per-day counts outside 0–50 excluded; future-dated records excluded |
3. Cohort
The base cohort is every user who created an account between September 1, 2025 and April 30, 2026 and reported a valid weekly drinking level at onboarding (n = 338,361). The signup window ends more than 90 days before the analysis date, so every user has a complete observation window. Two analysis groups are defined by tracking consistency:
Table 1. Cohort construction| Group | Definition | n | % of base |
|---|
| Base cohort | Signed up Sep 2025 to Apr 2026 with onboarding baseline | 338,361 | 100% |
| Logged ≥1 day, month 1 | Any drink log in days 0–29 | 86,417 | 25.5% |
| Consistent, month 1 | ≥10 logged days in days 0–29 | 40,756 | 12.0% |
| Consistent, month 2 | ≥10 logged days in days 30–59 | 20,319 | 6.0% |
| Month-3 cohort | ≥10 logged days in days 60–89; the primary analysis group | 14,709 | 4.3% |
| Longitudinal panel | ≥10 logged days in each of months 1, 2, and 3 | 12,475 | 3.7% |
The 10-day threshold requires a weekly rate estimate to rest on at least ten observed days per month. Section 4.5 shows results are not sensitive to this choice.
4. Results
4.1 Weekly drinking falls by half in the first month and holds
In the longitudinal panel (n = 12,475), median weekly drinks fall from 20 reported at onboarding to 9.3 tracked in month 1, then hold at 8.9 in month 2 and 8.8 in month 3.
4.2 Distribution of individual change
In the month-3 cohort (n = 14,709), the median per-user change in weekly drinks relative to onboarding is −49.4%. The distribution is broad and skewed toward large reductions:
Table 2. Per-user change in weekly drinks, month 3 vs. onboarding (n = 14,709)| Statistic | Value |
|---|
| 25th percentile (largest reductions) | −89.2% |
| Median | −49.4% |
| 75th percentile | −17.7% |
| Share drinking less than at onboarding | 85.6% |
| Share who cut by half or more | 49.8% |
One in four users in this cohort reduced by 89% or more, a level consistent with quitting or near-quitting. One in two at least halved their weekly drinking.
4.3 The heaviest drinkers reduce the most
Reductions appear across every starting level, and are largest, in both percentage and absolute terms, among users who started heaviest:
Table 3. Month-3 outcomes by onboarding drinking level (n = 14,709)| Onboarding level (drinks/week) | n | Median baseline | Median month 3 | Median change | Share reduced |
|---|
| 1–14 | 4,253 | 10 | 4.8 | −50.0% | 81.5% |
| 15–28 | 6,307 | 20 | 10.9 | −46.2% | 85.5% |
| 29+ | 4,149 | 38 | 17.0 | −54.9% | 89.7% |
Additionally, of the 10,456 users who started above 14 drinks per week, 55.6% were at or below 14 drinks per week by month 3.
4.4 Alcohol-free days and heavier drinking days
In the longitudinal panel, the median share of logged days that were fully alcohol-free is 60.0% in month 1, 59.3% in month 2, and 58.6% in month 3, a majority of tracked days in every month. The median share of logged days with four or more drinks is 10.3% in month 1 and 9.5% in month 3.
4.5 Robustness
Two checks address the main analytic choices:
Table 4. Sensitivity to the logging-consistency threshold (month 3 vs. onboarding)| Minimum logged days in month 3 | n | Median change | Share reduced |
|---|
| ≥5 | 17,353 | −48.8% | 84.5% |
| ≥10 (primary) | 14,709 | −49.4% | 85.6% |
| ≥15 | 12,732 | −51.2% | 86.4% |
Second, anchoring change to the first tracked week instead of the onboarding self-report yields a median change of −15.3% (users with ≥4 logged days in week 1; n = 12,951). Read together, the two anchors indicate that most of the reduction occurs rapidly, within the first weeks after joining, and is then sustained: month-1 tracked levels are already far below the onboarding report, and months 2 and 3 hold that level rather than regressing.
5. Limitations
- Observational design. There is no control group. These figures describe change among Reframe users, and causal attribution to the program cannot be made from this data alone.
- Engaged-user cohort. The primary cohort is users still tracking in their third month (4.3% of signups). Outcomes for users who stopped tracking are unobserved; results describe engaged users, not all signups.
- Instrument difference. The baseline is a one-time self-report; follow-up is day-level logging. Differences between recalled and logged consumption may contribute to the measured change; the week-1-anchored analysis in Section 4.5 bounds this from the other direction.
- Logging gaps. Users do not log every day (the threshold requires 10 of 30). Weekly rates are computed on logged days only, and unlogged days are treated as unobserved rather than alcohol-free.
Strengths run the other way: the sample is large, the follow-up measure is behavioral logging rather than end-of-study recall, the pattern is consistent across starting levels and analytic thresholds, and every figure is reproducible from the platform's records.
6. Conclusion
Among users who engage with Reframe's tracking through their third month, drinking falls substantially and quickly. Median weekly drinks drop from 20 reported at onboarding to about 9 tracked in month 1, and the reduction holds through month 3 rather than rebounding. Reductions are largest among the heaviest drinkers, a majority of tracked days are alcohol-free, and one in four engaged users reaches near-zero drinking. For clinicians, the practical read is that clients who adopt the app's tracking habit show large, sustained, measurable reductions that can complement work done in session.
Questions about methodology or requests for additional cuts of this analysis: clinicians@reframeapp.com