The DAU/MAU Ratio: Benchmarks and Why 20% Isn't Universal
DAU/MAU — usually called stickiness — is your average daily active users divided by your monthly active users, expressed as a percentage. Whether a given number is "good" depends entirely on the category, because the ratio mechanically converts into visit frequency. 20% just means the average monthly user shows up on roughly 6 of 30 days. A product with a naturally monthly cadence, like payroll or tax software, can't reach 20% no matter how much people rely on it.
Formula: DAU/MAU stickiness = (average daily active users ÷ monthly active users) × 100.
What the DAU/MAU ratio actually measures
The short version: stickiness tells you how many days in a typical month your monthly users actually come back.
DAU is the count of distinct users active on a given day. MAU is the count of distinct users active across a 30-day window. You usually compute the ratio by taking the average DAU over a period and dividing it by the MAU for that same period. So if you averaged 4,000 daily active users last month against 25,000 monthly active users, your stickiness is 4,000 ÷ 25,000 = 16%.
Two details trip people up. First, "active" means whatever your product defines it to mean, so garbage-in problems here poison the ratio downstream. Second, MAU is almost always a rolling 30-day window rather than a calendar month, which smooths out weekly spikes. And both DAU and MAU count unique users, not sessions — someone who opens your app nine times in a day is one DAU, not nine.
The number is really a frequency reading
Here's the reframe that makes the rest of this click. The ratio isn't an abstract quality score. It's visit frequency wearing a costume.
Multiply the ratio by 30 and you get days-per-month. KPI Tree lays this out cleanly: a 20% ratio means the average user visits about 6 days per month (30 × 0.20), and a 50% ratio means about 15 days. That's it. That's the whole conversion.
Once you see it this way, "is 20% good?" becomes "should my average user open this product 6 times a month?" For a messaging app, 6 days is dismal. For something you touch when you file quarterly taxes, 6 days a month would be suspicious — you'd wonder what they were doing in there. The ratio can't escape the frequency it encodes.
Where the "20% is good" rule came from
The heuristic has a specific and slightly deflating origin. According to a Wudpecker explainer on the DAU/MAU ratio, in 2014 Sequoia Capital tweeted that a standard DAU/MAU is 10–20%. A tweet. Not a study, not a benchmark dataset — a tweet that got repeated until it hardened into a law.
What gets left out is that Sequoia's own later guidance walks it back. In their note on measuring product health, Sequoia writes that "the specific value of DAU/MAU is strongly dependent on the business or product and the expected usage," and that a low ratio "does not necessarily mean it isn't doing well." The people credited with the 20% number were telling you not to use it as a universal target.
The 20% figure is really a consumer-app average — a rough middle for social and messaging products where daily use is the whole point. Applied to a B2B tool or a monthly utility, it's a category error dressed up as a KPI.
Why "loved" and "sticky" aren't the same thing
Let me show the frequency ceiling with actual numbers instead of asserting it.
Take 12 users, tracked over a 30-day month. Product A is a daily-habit app. Product B is a payroll tool people run once a month, plus a mid-cycle correction now and then.
For Product A, the 12 users log in like this: most show up around 18 days out of 30, a couple around 10, one power user hits 27. Add up all their active-days and you get roughly 200 total user-days. Average DAU is 200 ÷ 30 ≈ 6.7. MAU is 12, because everyone was active at least once. Stickiness = 6.7 ÷ 12 = 56%.
For Product B, all 12 users are loyal — 100% renewed, nobody churned. But they use it twice a month on average: run payroll, fix one thing. That's about 24 total user-days across the month. Average DAU is 24 ÷ 30 = 0.8. MAU is still 12. Stickiness = 0.8 ÷ 12 = 6.7%.
So Product B has perfect loyalty and a stickiness under 7%. Judge it against the 20% rule and you'd call a beloved, fully-retained product a failure. Nothing is broken. The cadence is monthly, so the ratio ceiling is low by arithmetic, not by weakness. Gainsight makes exactly this point: products have a natural usage cadence, and expecting to lift a 10% ratio to 50% is unrealistic without "drastically changing its underlying utility and value proposition."
Category benchmarks worth trusting
If one number is wrong, the fix is segmented bands tied to how often the product is supposed to be used. The strongest recent numbers here come from Mixpanel's 2026 State of Digital Analytics report, which analyzed data across 12,000+ companies and eight industries. Where a cell isn't from that dataset, I've noted it.
| Category | Typical stickiness band | Implied visit-days / month | Natural cadence |
|---|---|---|---|
| Social / consumer | ~20% and up | ~6+ | Daily-to-multiple-daily |
| B2B SaaS | ~31% (Mixpanel 2026) | ~9 | Most workdays |
| AI products | ~21% (Mixpanel 2026, NA) | ~6 | Task-driven, efficient sessions |
| Fintech | Category-dependent, often mid | varies | Weekly-to-monthly |
| Utility / monthly tool | Single digits (see worked example) | 1–3 | Monthly or event-driven |
A few things worth flagging. The B2B SaaS figure of ~31% is below the long-cited 40%. Mixpanel is explicit that the 40% number "was a Gainsight estimate" and "always a rough heuristic," never a measured average. So a lot of teams have been failing themselves against a line nobody ever verified. And the consumer/social band is precisely where the 20% myth was born, which is why it travels so badly to every other row.
The natural-frequency correction factor
Instead of one target, judge stickiness against your product's own ceiling. Here's the correction I use, and I'll admit up front that I side with Sequoia and Gainsight over the flat-benchmark crowd on this.
Three steps. Estimate the product's natural visit-frequency in days per month — for payroll that might be 2, for a project-management tool maybe 12, for a chat app 25. Convert that to a ceiling ratio by dividing by 30. Then compute your correction factor:
correction factor = observed DAU/MAU ÷ cadence-implied ceiling.
For the payroll example, natural cadence is ~2 days, so the ceiling is 2 ÷ 30 ≈ 6.7%. Observed was 6.7%. Correction factor ≈ 1.0 — the product sits at its practical ceiling. It is as sticky as a monthly tool can be. Chasing 20% here would mean inventing reasons for people to open payroll software daily, which nobody wants.
This is also why Mixpanel's read on AI is worth sitting with. They found North American AI products at 21% stickiness, then argued the number "is misleading" because "mature enterprise AI users tend to accomplish more per session and return less frequently as a result." Lower frequency as a sign of efficiency, not disengagement. The correction factor catches that: if the natural cadence dropped because the tool got better at one-shot answers, your ceiling moved, and measuring against a fixed 20% would punish the improvement.
Where this goes wrong
Four failure modes I see over and over.
The first is chasing 20% for a monthly-cadence product. Teams add nudges, streaks, and daily digests to a tool nobody needs daily, mangling the roadmap in service of a ratio that arithmetic won't let them hit. You end up with notification fatigue and a metric that still sits at 8%.
The second is reading low enterprise or AI stickiness as disengagement. As Mixpanel's AI finding shows, fewer visits can mean users are getting what they came for faster. Before you sound the alarm, check whether session value went up.
Third is comparing ratios across categories. Your 15% next to a competitor's 40% tells you nothing if one of you is a weekly analytics tool and the other is a social feed. Different cadences, incomparable ceilings.
The fourth is the sneakiest. DAU/MAU is an average, and averages hide bimodal distributions. A 25% ratio can be a healthy middle. Or it can be 30% of your users logging in daily while the other 70% went dormant weeks ago. Same headline number, wildly different health. This is where retention curves and cohort views earn their keep, and where a proper {{METRIC-TREE}} view of input metrics beats staring at one blended ratio. I'd pair stickiness with a look at your North Star and its input metrics rather than treat it as standalone.
Computing it without hand-rolling SQL
Most product analytics tools compute stickiness natively, and increasingly you can just ask for it in plain language — for example, asking a chat-first tool like Kixo for your DAU/MAU by segment instead of writing the query yourself. Pushing the mechanics into the tool frees you to spend your time on the interpretation, which is where the real work lives anyway.
Quick reference recap
Stickiness = (average DAU ÷ MAU) × 100. It converts straight to frequency: multiply by 30 to get visit-days per month, so 20% ≈ 6 days and 50% ≈ 15 days. The 20% rule is a 2014 Sequoia tweet reflecting a consumer-app average, not a universal pass mark, and Sequoia itself says the meaningful value depends on expected usage. Mixpanel's 2026 data puts B2B SaaS near 31% and AI near 21%, both context-dependent. Benchmark against your cadence, not against 20%: estimate natural visit-frequency, convert to a ceiling, and measure how close you are to it.
FAQ
What is a good DAU/MAU ratio? It depends on your product's natural usage frequency. Consumer and social apps run around 20% and up; Mixpanel's 2026 report puts B2B SaaS near 31% and AI products near 21%. A monthly-cadence tool can be excellent at single-digit stickiness.
How do you calculate DAU/MAU? Divide average daily active users over a period by monthly active users for the same period, then multiply by 100. Both counts are unique users, and MAU is typically a rolling 30-day window.
Where did the 40% B2B benchmark come from? Per Mixpanel, it was a Gainsight estimate — a rough heuristic that got widely cited. Mixpanel's 2026 measured average across 12,000+ companies came in closer to 31%.
Does low stickiness mean my product is failing? Not on its own. Sequoia's own guidance says a low ratio doesn't necessarily indicate a weak product, and for monthly tools or efficient AI products, low frequency is expected rather than a warning sign.