Streak

While notifications were the only way to produce a return visit, the handle sat entirely on the company's side. Send a message to someone who hasn't shown up, and send more if they still don't. Nothing inside the product gave them a reason to open it themselves.

What Mia found in the 30-day curve after signup shows the limit of that. People dropped out unusually hard between day seven and day eight. Split by first-week behavior, more than half of the users who came in on five or more of their first seven days were still there a month later; users who came in on two days or fewer held at 8 percent. How many days someone shows up in week one all but decides the rest, and the only tool for shaping week one was notifications. Open rates on those start falling by the third day.

A streak moves that handle over to the user. Put the number of consecutive days on the screen, and the reason to come back tomorrow stops being a message from the company and becomes the number they built up to yesterday.

It runs on an unwillingness to lose what has accumulated. For someone twelve days in, skipping a day is not skipping a day — it is resetting twelve to zero. Losses register more heavily than gains, so the same twenty-four hours carry a different weight.

Building one means settling four things.

1. What counts as a day. Opening the app, or editing at least one document. A loose rule makes the number easy to build and empty of meaning; a strict one breaks most people's streaks within the first week.
2. Where the day boundary falls. Cutting at midnight separates someone who arrives at 11:50 p.m. from someone who arrives at 12:10 a.m. by twenty minutes. Counting in the user's own time zone is the baseline, and some products push the boundary to 4 a.m.
3. How it gets displayed. A single number, or a month calendar with completed days filled in. The calendar makes the empty squares visible and invites filling; the number gains weight as it grows.
4. Whether to warn before it breaks. One evening reminder to anyone who hasn't done it yet. These work well, and they also go out every single day, so they compete for room inside the notification strategy that decides who hears what and when.

Building in a cushion is the part that decides whether the whole thing helps. A one-day grace period, a rule that skips weekends, a make-up option that lets tomorrow count twice. The reason sits in what people do after a break. Someone forty days in who misses a day and lands on zero often doesn't come back at all rather than restarting at one. Once the accumulated thing is gone, the same pressure runs in the opposite direction.

Two numbers report the result: the share of users holding a streak of seven days or more, and the share of broken streaks restarted within a week. Chasing the first number alone loosens the rule for what counts as a day. Reading both pushes the work toward the cushion instead, which is usually where the larger gain sits anyway.

Products where value comes from doing something daily are the home case — language learning, exercise logs, budgeting, journaling, anything where one day is the natural unit of use. There the streak acts as a bridge carrying users toward the habit moment, the point where use hardens into repeated behavior. Around twenty days in, the hand moves before the number does any work, and the counter has done its job whether or not anyone still looks at it.

It also serves as one component of gamification, the practice of putting points, badges and leaderboards into a product. What separates it from the other components is that the opponent is yesterday's version of the same person. A leaderboard works on the few at the top and quietly discourages everyone below them; a streak hands every user their own number, which is why it survives in products where a public ranking would be unwelcome.

The products it does not fit matter more. Attach it to a tool people need two or three times a month and it manufactures visits that last three seconds and exist only to avoid a break. Daily visits rise, the retention curve looks better, and nothing anyone gets out of the product changes. In a tool where daily users and before-the-meeting users are mixed together, the device has to be scoped to a segment rather than shipped to everyone: it fits a team writing meeting notes every day and is pure pressure on a team writing one planning doc a quarter.

"Daily visits are up 18 percent since the streak shipped and documents created is flat."

Visits went up and actual use did not.

"Users who break a 40-day streak don't come back the next week at all. Let's add one grace day."

Cushioning the drop-off that follows a break.

"This isn't a daily product. Adding that just drags people in for no reason."

The device and the product don't match.

Pagedeck ran a daily editing streak for all users over two weeks last quarter. Daily visits rose 18 percent while documents created rose 2 percent, and most of the added visits were under a minute, opened and closed.

Mia narrowed the audience rather than pulling the device. It stayed only for teams that had come in four or more days a week over the previous month and came off everyone else. In that segment, visits and documents created rose together. The same device manufactured use in one group and manufactured visits in another, which made the audience the variable rather than the device.

The most common misreading is that a longer number means a better state, because consecutive days look like a measure of engagement. But the number counts visits, not anything obtained from the product. If opening the app is what counts, a user 60 days into a streak may not have created a single document in those 60 days. Put consecutive days on the goal sheet in that condition and the rule for what counts gets looser, and six months later the average streak length is up while the cancellation rate has not moved.

The other one is that more pressure works better. The warning before a break performs well, so two or three a day should perform better. Daily nagging is the fastest route to notification fatigue, where users swipe messages away unread or switch notifications off entirely. Once someone turns them off, genuinely important news stops reaching them, and almost nobody turns the setting back on. Visit counts rise as the pressure rises, so this bill tends to arrive about a quarter late.