What Is MAU? Monthly Active Users Defined and Calculated

A.Three decisions the formula hides

The first is the qualifying action. A count of people who loaded a page and a count of people who did the thing your product exists to do are different populations, and the gap between them is not a rounding difference. Nothing about the product changes when you move that bar. The number moves anyway.

The second is the unit of deduplication. Dedupe on device or session and one person with a laptop and a phone counts twice, which means your MAU partly measures device proliferation. Dedupe at the person level and you need a durable identity across those devices, which most products only have for signed-in users. That is a real constraint rather than a configuration setting, and it is worth knowing which side of it your number sits on before quoting it.

The third is the window. A rolling 30 days recomputed daily and a calendar month that resets on the first behave differently for any product with a weekly rhythm: the calendar version has a sawtooth in it that has nothing to do with usage, and the rolling version smooths over month boundaries that finance cares about. Pick one, write it down, and never present the two side by side.

The word doing all the work in “monthly active users” is “active,” and it is the only word the formula does not define. Define it as an action that indicates the user received something: created or updated a task rather than opened the dashboard, ran a report rather than landed on the home screen, posted or commented rather than scrolled, sent a message rather than launched the app.

B.What a loose definition costs later

A definition that is easy to hit is not a harmless simplification, because it changes what the metric can detect. The looser the bar, the more of your MAU is made of people whose behaviour cannot deteriorate any further. Someone counted for opening the app is already at the floor. When engagement actually declines, they keep counting, and the metric holds flat through the exact deterioration it was installed to catch. A tight definition has room to fall, which is what makes it an instrument.

The second cost is comparability. MAU is only meaningful as a series, and a series is only meaningful if the definition and the window were frozen across it. Every redefinition creates a discontinuity, and because redefinitions are usually undocumented, the discontinuity gets read as growth or as churn by whoever inherits the chart. Two teams using different qualifying events for the same product produce two MAUs, both correct, that will be argued about in a meeting.

There is also a commercial incentive worth naming, because it pulls in the wrong direction. Many analytics vendors price on tracked MAU, which means a stricter definition of active reduces your bill. That is a pricing model quietly editing your measurement. KISSmetrics prices on events instead: 100,000 events a month free, $99 for 500,000 on Growth, $299 for 2 million on Silver. The definition of active stays a product decision rather than a budget one.

A.Reach read as intensity

MAU tells you how many people crossed a threshold once in thirty days. It says nothing about how many times, and the difference between a user who returned once and a user who returned twenty times is the entire difference between a product that has a habit and one that has an audience. The ratio that recovers the missing dimension is DAU/MAU, usually called stickiness: the share of your monthly population present on an average day. Fifty percent means half your monthly users are here today, ten percent means one in ten.

The failure mode this exposes is the common one. A product can grow MAU through acquisition while stickiness falls, which means the new arrivals are less habitual than the people they replaced. MAU rises the whole time. The reported picture is growth, the actual picture is a widening pool of shallower users, and the moment acquisition slows the two numbers cross.

The same argument applies to the choice of window. Track MAU for a product designed for daily use and the number stays healthy while daily engagement rots underneath it. Track DAU for a product used monthly and the number looks like a disaster even when every user returns reliably. The window is not a reporting preference. It is a claim about how often your product is supposed to be used, and it should match one.

WAU is the underused member of the family and often the correct primary for business software, because a seven-day window absorbs weekends and public holidays without absorbing a lapse. A user who skips a week has changed their behaviour. A user who skips a Tuesday has not. It follows that WAU/MAU is the gentler stickiness ratio for anything used inside a working rhythm, and that quoting DAU/MAU for a weekly product produces a number which looks alarming and means nothing.

B.One column of a grid, read as the whole grid

MAU in any month is a single figure standing in for a distribution of cohorts, each at a different age and each decaying at its own rate. A cohort grid holds all of it. MAU holds one number read down one column, which is why it can be flat while the underlying retention curve improves, and flat while it collapses.

Two consequences follow, and they are the reason MAU should never be a target. Add enough new users each month and MAU climbs however badly each cohort retains, because acquisition is a flow and retention is a rate, and a flow can outrun a rate for a long time. And because every user counts once regardless of what they are worth, MAU can hold steady while your highest-paying customers leave and free accounts replace them. Product-led growth metrics that tie activity to revenue are what stop that being invisible.

A.Count days, and decompose the movement

The binary flag is the root defect: a user is active or not, and twenty visits look like one. Counting active days out of the last 7, 14 or 28 restores the distribution for free, using data you already have. L28 is MAU with the information left in, and it splits your population into the shapes that matter: daily, weekly, occasional, and gone. Activation rate does the same job at the other end of the lifecycle, measuring the share of new sign-ups who reach a defined activated state rather than the share who appeared. Our activation rate guide has that framework in full.

The second move is to stop reporting MAU as a level and start reporting it as a balance. Any month-over-month change is the sum of three flows, and they have completely different implications.

Two more measures earn their place beside those. Feature adoption, the share of active users touching each capability, because depth predicts retention in a way headcount does not. And revenue per active user, which keeps the count tethered to economics: MAU up twenty percent with ARPU down thirty means you are adding people who are worth less than the ones you had. Our guide to growth metrics covers holding that connection over time.

The usual blocker is instrumentation rather than analysis. Defining active as “sent a message” requires a Message Sent event to exist, and waiting on an engineering ticket is where most good definitions die. Autocapture removes the dependency: KISSmetrics scans the product on install, records the interactions already happening, and the Cohorts and Product Usage reports read against them without a code change.

Resurrection is worth singling out, because it is the flow most teams never build. Everyone instruments new users, most instrument churn, and almost nobody can name the people who lapsed and came back or say what brought them. Those users convert on second contact far more readily than a stranger does, since they already know what the product is, and they cost nothing to reach. Holding them as a population that updates itself is the difference between knowing the flow exists and being able to act on it.

B.Where MAU is still the right number

None of the replacements do MAU’s actual job, which is to be one number that fits in a sentence. A board update, an investor letter, a category comparison and a capacity forecast all need a single scalar, and a decomposition is not a scalar. L28 distributions and cohort grids are diagnostic instruments; they are unreadable at the altitude MAU is usually quoted from, and demanding that everyone read a grid is how reporting stops happening.

There is also a class of product where the thirty-day window is not arbitrary but correct. Payroll, invoicing, tax, expense approval, monthly reporting: the natural cadence is one visit a month, so a user active once is a fully engaged user, and stickiness is meaningless by construction. For those products MAU is close to a complete picture of usage, and pushing the team toward daily engagement metrics manufactures a problem the product does not have.

Cross-industry benchmarks are the part to abandon rather than improve. Published MAU figures rarely state their qualifying action, their deduplication unit, or their window, which are the three things that determine the number. Comparing your MAU to a competitor’s is comparing two unlabelled definitions. Your own series, on a frozen definition, carries more information than any external figure.