Daily Active Users

For a long time the answer to "is the product doing well" was the cumulative signup count. A hundred and twenty thousand people signed up is a clear sentence and an easy one to put in a deck. The trouble is that the number can never go down. Every user could abandon the product tomorrow and the total would still read a hundred and twenty thousand, with the chart still pointing up and to the right. A number that looks good while things get worse is a number no decision can rest on.

Meetings went the way you would expect. Grace would ask how many people use the product right now, someone would answer a hundred and twenty thousand signups, and nobody could say how many of those were actually using it. Shipping a feature settled nothing either, because signing up happens once and everything after it leaves no mark on that total.

Daily active users (DAU) sidesteps the problem by changing the question. Not how many people arrived, but how many used the product yesterday. Anyone who skipped yesterday drops out of yesterday's number, so the value can fall — and only a number that can fall is worth deciding anything with.

Count the distinct accounts that performed the action you call active at least once during the day. Ten visits from one person still count as one. That is the whole calculation; the work that actually takes time is the four decisions made before any counting starts.

1. Decide what a person has to do to count. Including or excluding people who logged in and left moves the value by around 20 percent. Pagedeck counts only accounts that opened at least one document, because the product's value comes from documents — count people who came and went and the metric stops tracking whether the product was useful.
2. Decide where the day ends. Midnight is the cut, but which midnight has to be written down. Cut on Coordinated Universal Time (UTC) and everything a Korean user does before 9 a.m. falls into the previous day, which puts the daily figure a shift away from what local time would show.
3. Decide how many one person counts as. Accounts and devices give different answers, and one person with a work account and a personal account registers as two. Laptop plus phone on the same day is also two under a device rule.
4. Decide who to leave out. Without stripping employee accounts and automated monitoring accounts, a company of 45 people carries that much padding in the number every single day.

Once those four are settled the value comes out on its own, but a single day never gets read on its own, because the day of the week drives it hard. Products used for work commonly drop to a third of their weekday number on Saturday. So a seven-day moving average goes next to it — the mean of the last seven days including today, which contains each weekday exactly once and cancels the weekly shape out.

Comparing against another company's figure can't tell you what a good number is. What counts as active is defined differently everywhere, so two products both reporting ten thousand may be counting logins on one side and document opens on the other. On top of that, a product with a reason to be opened daily and a product with a reason to be opened monthly were never playing in the same range.

Three reference points do the job instead: your own number from last quarter, the number before and after you changed something, and the number you get by working backwards from what next quarter's plan requires. The first two compare measurements taken with the same ruler. The third is what turns a measurement into a target.

Not reading it alone is part of the standard too. If DAU climbs from 900 to 1,100 while monthly active users (MAU) climbs from 8,400 to 12,000, what grew is the crowd, not the attachment. It means advertising brought a lot of people in and most of them looked once and left. The ratio between the two reads this directly and is called stickiness. Nine hundred divided by 8,400 is 11 percent: roughly one in nine of the people who use the product in a month shows up daily.

Having an order for investigating a drop helps as well. Rule out the day of the week and public holidays first, check that the tracking itself didn't stop, and only then look at what changed in the product. Skip that order and a holiday week turns into a decision to roll back last week's release.

"DAU was 900 yesterday against 940 the same day last week. On the seven-day average it's still flat."

Read the moving average rather than treating a one-day dip as an event.

"Is this number document opens or logins?"

A metric with one name produces different values under different definitions, so settle the definition first.

"DAU went up but MAU went up more. Attachment actually went down."

Separating more people from more frequent people.

Pagedeck keeps this value on the meeting screen next to its seven-day average. What Mia watches each week is the slope of that average line, not yesterday's point. After a release, a vertical line goes on the chart at the release date and the question is whether the slope differs on either side of it. The week the comment notifications were reworked, the average rose about 60 over the following fortnight, and whether that came from the notifications or from a partner promotion the same week took a second look split by acquisition channel.

On weeks when the number falls, the calendar comes first. If a public holiday sits in the week, that week leaves the comparison and the next one gets the attention.

The most common misreading is treating DAU as the count of people who come every day. The word daily is right there in the name, so it sounds like it only counts the regulars. It is a snapshot of one specific day, and anyone who showed up once that day is in it. A steady 900 across a month is no guarantee of the same 900 people; it can just as easily be 8,400 people each dropping in on one day. Leave that belief alone and the difference between making people come daily and making someone come once today disappears, which tilts every plan toward sending more notifications.

Turning the value into a goal is the other frequent move, because it is an easy metric to raise. Send three notifications a day and next week's number will certainly go up, along with the count of people who switch notifications off. The visible number rises and the bill arrives a quarter later as retention, the measure of whether existing users keep coming back.