glossary · metrics
DAU/MAU ratio
Average daily active users divided by monthly active users, expressed as a percentage. A stickiness proxy that measures how often users return within a month.
DAU/MAU is the fraction of your monthly user base that is active on any given day. Expressed as a percentage, it tells you how often users in the denominator are choosing to return. A 30% ratio means users are active roughly 9 days per month on average; 20% is roughly 6 days. The formula is simple. The judgment required to use it correctly in 2026 is not.
Formula and plain definition
DAU/MAU = (average daily active users / monthly active users) × 100
Both the numerator and denominator depend on how you define “active.” This is the most probed point in any metrics interview. Opening an app is not an active event. Completing a task, sending a message, editing a file, or generating a report is. The definition must name a value-delivering action specific to the product, otherwise you are measuring whether your app icon is visible, not whether it is useful.
Benchmarks by category
The “20% means healthy” shorthand is outdated and category-agnostic. Use these reference points instead.
| Category | DAU/MAU range |
|---|---|
| Consumer messaging (WhatsApp) | ~70% |
| Consumer social (Instagram) | ~60% |
| Consumer social (Facebook) | ~50% |
| Content and social discovery (Twitter/X) | ~30% |
| B2B collaboration (Slack) | ~30% |
| B2B SaaS (North America/EMEA) | ~31% |
| B2B SaaS (APAC) | ~33% |
| AI products (North America) | ~21% |
| AI products (LATAM) | ~37% |
| Ecommerce | 20-23% |
| Periodic-use tools (payroll, billing) | 5-15% |
Source: Mixpanel 2026 State of Digital Analytics (3.7 trillion events, 12,000+ companies). The 31% B2B SaaS figure replaces the informal “20% floor” that circulated for years. The Sequoia heuristic still holds as a rough frame: 10-20% is standard, 50%+ is exceptional, but neither number means anything without a category anchor.
WhatsApp sits near 70% because messaging is designed for daily habit loops. Slack and Twitter/X both land around 30% despite being “daily-use” products, because their natural cadence includes gaps. Periodic-use tools at 5-15% are entirely healthy when the job to be done (quarterly report, monthly payroll) does not require daily return. Comparing those products to WhatsApp is a category error, and interviewers specifically test whether you know that.
WAU/MAU as the right window for weekly-cadence products
DAU/MAU only makes sense when users are expected to return daily. For products with a natural weekly cadence (project management, code review, planning tools), WAU/MAU is the more honest stickiness measure. Forcing a daily lens onto a weekly-cadence product manufactures a low-looking number that tells you nothing useful. The right question before picking a measurement window: what is the natural cadence of the job this product performs?
The 2026 complication: AI agents inflate the number
AI agents making API calls on a user’s behalf do not take weekends or holidays. An account with significant agentic activity will generate DAU events every day regardless of whether a human opened the product. A 60% DAU/MAU can mean your automation layer is running, not that humans find the product worth returning to.
Before reporting stickiness, segment human sessions from agent-triggered events and report human DAU/MAU separately. Interviewers at companies building AI products will probe this explicitly. Candidates who treat all events as equivalent are signaling they have not thought about what the metric is actually measuring.
The deeper question the metric is supposed to answer: is the user returning because the product is genuinely worth returning to (they choose to come back) or because an agent is triggering on their behalf? Those two situations require entirely different interventions.
DAU/MAU is a lag metric
A declining ratio reflects activation, onboarding, and habit-formation decisions made weeks or months earlier. By the time the number moves, you are already behind. Pair DAU/MAU with leading indicators: activation rate (did users reach the key behavior in session one?), task recurrence interval (how long between first and second meaningful action?), and feature adoption rate (are users finding the habit-forming feature, not just the entry point?).
When diagnosing a drop, segment before you conclude. DAU/MAU can fall because new-user acquisition is growing (denominator up, same numerators) while core users are stickier than before. Cohort-level DAU/MAU by signup month strips that artifact and isolates whether the problem is with new users, existing users, or both.
Weak vs strong interview answer
weak
"DAU/MAU above 20% means you have a sticky product." This fails on every dimension: it ignores category context (20% is weak for a social app and reasonable for B2B SaaS), it assumes "active" is self-evident, it treats the metric in isolation without pairing it with retention or leading indicators, and it does not acknowledge that AI agent activity can inflate the number without reflecting human engagement. Interviewers recognize this answer as pattern-matched from a prep site.
strong
"I start by asking what the natural use cadence of the product is, because DAU/MAU only makes sense when users are expected to return daily. For a weekly-cadence product, I'd use WAU/MAU. Then I define 'active' as a value-delivering action specific to this product: for Slack that's a message sent, for Figma it's an edit, not a login. I apply the formula, then benchmark against category: consumer social targets 60%+, B2B SaaS sits around 31% in North America per Mixpanel's 2026 data, and periodic-use tools can be healthy at 5-15%. In 2026, I would explicitly ask whether the user base includes AI agents or scheduled automations. If so, I segment human sessions before reporting stickiness, because agentic activity inflates the number without reflecting human habit formation. Finally, I treat DAU/MAU as a lag indicator and pair it with activation rate and task recurrence interval so I can act before the stickiness number moves."
For the full diagnostic flow when this metric drops, see DAU dropped: find the root cause. Related glossary entries: retention and cohort analysis.