DAU/MAU is a compact stickiness signal that summarizes how often monthly users return on a given day; this note gives a practical measurement approach, directional benchmarks, and a realistic 1- to 6-week playbook to test lifts. Expect directional signals in 1-2 weeks for onboarding fixes and 4-6 weeks for habit changes, but timelines depend on instrumentation quality, segment sizes, and team bandwidth. Tradeoffs include engineering time, campaign spend, ARPU vs retention impacts, and privacy/consent constraints that may limit targeting or measurement.
Early proof block
| Band | DAU/MAU range | Minimum sample guidance | Quick experiment |
|---|---|---|---|
| Baseline | 5% - 15% | ~1k MAU | 7-day onboarding funnel audit (time-to-first-value) |
| Momentum | 15% - 30% | 2k - 4k MAU per segment | Segmented lifecycle campaign with 14-day holdout |
| Habit | 30% - 60%+ | 1k+ DAU for fatigue signals | 2-week notification frequency A/B test measuring churn and ARPU |
What this means: DAU/MAU compresses retention into a quick operational metric you can act on. Use the band to prioritize diagnostics and run one targeted experiment before committing more resources. Business impact: onboarding fixes are typically fastest and lowest cost; scaling lifecycle campaigns requires ROI discipline; habit products need fatigue controls to protect monetization.
What do DAU/MAU bands indicate and how should you act?
Category: Early proof
Statistic: 5 - 15%
Label: Baseline DAU/MAU band
Context: Use only if ≥1k MAU to reduce noise
Category: Growth
Statistic: 15 - 30%
Label: Momentum DAU/MAU band
Context: Segment and iterate across 2 - 4k MAU cohorts
Category: Retention
Statistic: 30 - 60%+
Label: Habit DAU/MAU band
Context: Run fatigue + ARPU tests as usage intensifies
Use the band to decide whether to fix onboarding, optimize lifecycle, or protect monetization; each band has one prioritized experiment you can run immediately. Below are concise actions and interpretation.
Benchmark bands (directional)
Baseline (~5 - 15% DAU/MAU)
Action: run a 7-day funnel audit for time-to-first-value and D1/D7 leaks. Minimum stable sample ~1k MAU; below that expect high noise and longer test windows.
Momentum (~15 - 30% DAU/MAU)
Action: run segmented lifecycle campaigns with randomized holdouts and a 7-14 day evaluation. Require 2k - 4k MAU per segment for directional confidence.
Habit (~30 - 60%+ DAU/MAU)
Action: treat as daily habit and run a 2-week randomized frequency test measuring churn and ARPU per active user. Monitor fatigue; benefits often taper with higher touch.
Interpretation: Baseline calls for onboarding fixes; Momentum is about scaling with ROI guardrails; Habit requires protecting ARPU and customer experience. One practical caution: many early lifts decay by day 14, so include persistence checks before rollout.
How to act on the bands this week
- Baseline: if D1 < 25% and DAU/MAU < 10%, start an onboarding A/B within 7 days. Typical effort: 20-80 engineer-hours and 1-2 analyst days to instrument and analyze.
- Momentum: reserve ~10% of sends as holdout for 14 days, measure incremental ARPU vs control before scaling. Expect 1-3 analyst days to set up and 10-40 engineer-hours to integrate.
- Habit: run small frequency-cap tests for 2 weeks; revert if ARPU per DAU falls or uninstall/support spikes exceed thresholds below.
How do you calculate and validate DAU/MAU?
*A checklist for running an onboarding A/B test: hypothesis statement, experiment_id and sample-size target, KPIs (D1, day-14 cohort DAU/MAU), data tags required, analysis window, and explicit rollback rules (D7 fail, novelty fade, instrumentation alerts).*Calculate DAU/MAU consistently, validate identity, and annotate exports before acting on small changes. Run a short audit to decide if you can trust the signal.
Formula, window selection, and export checklist
Use DAU/MAU = unique users with an event on a UTC day ÷ unique users with at least one event in the trailing 28 or 30 days; pick one and keep it consistent.
Rolling DAU/MAU export
Export a rolling series covering the last 30 days with your chosen window tag.
Retention cohorts
Export D1 and D7 retention for the last 6 daily cohorts to spot early leaks.
Session and engagement metrics
Export sessions per user/day, median session length, and top event counts.
Annotation fields
Include experiment_id and marketing_campaign tags; flag campaign days and exclude or annotate them in trend analysis.
Event and user filters
Filter test accounts, bots, and the top 1% by event volume to reduce noise.
Timeline snapshot you can run in two weeks:
- Day 0: one-hour audit - export rolling DAU/MAU and cohort tables.
- Days 1-7: run D1/D7 checks for recent cohorts and flag deltas.
- Days 8-14: compare app-open DAU vs meaningful-action DAU and lock window choice.
Identity hygiene, dedupe, and event-based validation
Aim for >= 90% of activity mapped to a stable user_id; below 90% expect noisy DAU/MAU and plan identity reconciliation (typical effort 1-4 weeks). Dedupe test accounts and top-volume users. Trigger an instrumentation audit if DAU/MAU shifts > 5 percentage points after filtering. A gap > 40% between app-open and meaningful-action DAU often signals session inflation.
Sample-size and uncertainty
For directional trends, 1k MAU gives basic stability; to detect a +3 to +5 percentage point absolute lift, plan for 10k+ MAU or run a simple power calculation (baseline rate, target lift, alpha 0.05, power 0.8). Detectable effect size depends on variance and allocation; smaller apps may need longer windows or larger pooled segments.
Mid-article: Get the one-hour DAU/MAU audit checklist
The audit tells you whether the signal is clean for quick experiments or if you need an instrumentation sprint first.
How can you lift DAU/MAU in 2-6 weeks and when should you rollback?
You can run structured experiments with clear rollback rules in 2-6 weeks, but outcomes depend on instrument quality, sample size, and team bandwidth.
Prerequisites, team, and timelines
Minimum tooling: clean user_id mapping, 30-day event export, experiment platform, and campaign annotation. Team: 1 analyst + 1 product owner + 1-2 engineers for 1-3 weeks for onboarding tests; habit work may need 4-12 weeks and more staff. Engineering-hours: onboarding tests 20-80 hours; habit/frequency experiments 80-320 hours. Expect a 30-60% soft failure rate for early hypotheses; plan for follow-up iterations.
1 - 3 week onboarding experiments (tests and measurement rules)
Remove one onboarding step
A/B test removing a single screen or field. KPI: D1 retention and 14-day cohort DAU/MAU. Sample target: 10k MAU for medium apps. Measurement window: 14 days. Roll back if D7 shows no improvement or novelty fades.
First-success checklist
Add a tracked 'first success' event and guide users to it. KPI: day-0 to day-3 active rate and session quality. Require clean instrumentation for the event.
Micro-copy and contextual tooltips
Single-variable copy tests around the critical action. KPI: D1 lift and session depth. Measure 7-14 days and revert if D7 shows no lift or support tickets rise.
Operational checklist: annotate experiment_id and campaign tags, schedule a mid-test instrumentation health check, reserve a control holdout, and define rollback steps (disable feature, stop campaign, reroute traffic).
Monitoring triggers and rollback thresholds
Use concrete thresholds and clear owners.
- Auto-alert if D1 or D7 drops > 5 percentage points versus baseline.
- Alert if uninstalls per 1k DAU increase > 2 absolute or relative increase > 50%.
- Alert if support tickets per 1k DAU increase > 5.
Ownership: send alerts to the product incident Slack channel; product owner evaluates within 2 hours and engineering on-call should execute rollback within 4 hours. These are conservative defaults; adjust to your SLA and risk tolerance.
One caution: a welcome modal we tested boosted D1 by 6 points but reverted by day 14 and increased support volume; we rolled it back and reworked the flow.
