Weekly Active Users

For a long time the number companies quoted was registered accounts. Twelve thousand signups collected over three years only ever goes up: nobody came in yesterday, half the old users have drifted away, and the figure is still twelve thousand. So teams moved to counting only the people who showed up inside a fixed window. The first two windows anyone tried were a day and a month, and both have a catch on a meeting screen.

The daily count, daily active users (DAU), swings hard with the day of the week. Someone opens Monday's review and asks why yesterday was 340 when the previous Thursday was 910. The only honest answer is that yesterday was a Sunday. Work software routinely drops to a third of its weekday level on weekends, so a single day often tells you which day it was rather than anything about the product.

The monthly count, monthly active users (MAU), has the opposite problem: it is blunt. Somebody who logged in once on the second is in there alongside somebody who logged in daily. And a fix shipped last week only surfaces at the start of next month, by which time three other changes have landed and nobody can say which one did the work.

Weekly active users sits in the gap. Bundle Monday through Sunday into one figure and the weekend dip gets absorbed by the weekdays around it, so the day-of-week effect cancels out inside the window. A week is also short enough that people will wait for it: something you changed this week shows up in next week's number, while the two are still connected in everyone's memory.

Take the week you have defined, count every person who was active in the product during it exactly once, and report how many people that is.

Counting each person once is what gives the metric its character. Someone who logged in five times counts as one, and so does someone who logged in on Wednesday and never came back. A week with 3,200 visits can easily resolve to 1,100 people once repeats are collapsed. If visits are what you care about, count sessions — one stretch of use from arrival to departure — separately.

Four things have to be written down and agreed on before the number means anything.

1. Decide what counts as active. Whether you include people who logged in and did nothing else moves the result by around 20 percent. Pagedeck counts an account only if it opened at least one document.
2. Decide where the week starts and ends. A calendar week from Monday to Sunday compares directly against last week; a rolling window of the last seven days gives a fresh number every day. Pick one and keep it.
3. Decide how you recognize the same person. Counting by account and by device give different answers, and one person with a work account and a personal account shows up as two.
4. Decide who gets excluded. Leave staff accounts and automated monitoring accounts in and a 45-person company quietly adds 45 to its own number every week.

Change any one of those four later and earlier weeks stop being comparable. Record the date of the change and recompute a few earlier weeks the new way.

There is no way to judge the figure on its own. Whether 1,100 people is a lot depends on the market and the price. For a tool sold to companies at $400 a seat per year it is a real business; for a consumer service it is barely a start.

So the reference points come from inside, and three of them work: your own number last quarter, the number before and after a specific change, and the number next quarter's plan would need, worked backwards from that plan.

One week on its own also does not set a direction. A week with a public holiday comes in low; a week when a large customer ran internal training comes in high. Four weeks going 2,600, 2,750, 2,940, 3,100 says far more than any single week that rose.

Put the figure next to the monthly one and something extra appears. Divide 3,100 by 8,400 and you get 37 percent: close to four in ten of the people who touched the product this month also touched it this week. Above half means almost nobody drops in once a month; below a quarter means most users are occasional visitors. Dividing the daily figure by the monthly one gives stickiness, the same question asked about people who return every day.

"WAU is 3,100 this week, up from 2,940. There was a holiday in there, so let's wait for next week before calling it."

A single week is not enough to declare a direction.

"Only 4 percent of WAU has tried that feature. Against all signups it reads as 1 percent."

This is the moment a team agrees on the denominator for feature adoption.

"WAU went up but MAU went up more. The new people looked once and left."

Scale grew, engagement per person did not.

At Pagedeck, Mia keeps this figure in the first cell of the Monday morning dashboard. When Ravi has run ads the week before, the effect is mixed in, so they split it by acquisition channel first. In a week when Theo shipped a first-load speed change, they mark the deploy date with a vertical line and compare the weeks on either side.

Outward-facing documents use a different figure. The deck Grace sends to investors reports monthly active users, because that is the convention and the larger window flatters it. The team needs to know the inside and outside numbers are deliberately different, or the two land in the same meeting and the conversation becomes an argument about which is wrong.

The most common error is assuming this is the daily figure times seven. The name makes it sound like seven days added together, so the arithmetic feels obvious. But repeat visitors collapse to one, so a product with a DAU of 900 typically lands somewhere between 2,000 and 3,000 for the week, not 6,300. Teams that never check this set targets they cannot reach, then record the quarter as a miss.

The second error is reading any increase as proof the product is working. More people is more people, so the reading is natural. But the figure rises the week you spend on ads whether or not those people ever come back. The harder case is a flat number: 3,100 one week and 3,100 the next can hide 500 people leaving and 500 arriving underneath. Miss that and you miss the window to find out why they left, which is why this metric belongs next to retention, the measure of whether existing users keep returning.