Skillion is built for the first week: small tasks to start with, energy that stops you piling on fifteen things, and a nudge before a streak breaks. That design comes from a figure almost nobody has: when a habit really gets abandoned.
Half the habits anyone sets never reach day seven. We know because we measured it in our own data. I ran the query on a Sunday afternoon expecting an ugly number. It came out worse.
Almost everything you have read about habits comes from studies with a few dozen volunteers who stayed to the end. We see the opposite: the person who installs the app on a Tuesday, creates four habits, shows up twice and never comes back. That is why the app is built the way it is.
1 · What we are counting
- From 9 October 2025 to 30 August 2026.
- 7,070 people, 156,028 completed tasks and 9,923 repeating habits, from 2,323 people.
- All of it added up, no names. No group smaller than 50 people appears here.
- “A habit lived 90 days” means 90 days passed between the first time someone completed it and the last. Not that they did it every day. That is a generous way to count, and the numbers still come out low.
2 · Half of them never reach week two
Out of 9,923 repeating habits:
| How long they lived | |
|---|---|
| A quarter, less than | 3 days |
| Half, less than | 7 days |
| Three quarters, less than | 19 days |
16.8% get completed once and never again: born and dead the same day. At the other end, the oldest habit still alive is on 314 days.
3 · Why that figure misleads too
If a habit was born ten days ago, it cannot have lived 90. Throwing it into the “how many reach 90 days” count makes more of them look dead than really are. It is the most repeated mistake in charts like this.
If you look only at people who have been around 90 days or more, the picture changes completely:
| Still alive at… | Everyone (9,923) | People on 90+ days (1,333) |
|---|---|---|
| 7 days | 54% | 92% |
| 14 days | 34% | 85% |
| 30 days | 16% | 74% |
| 66 days | ~9% | 57% |
| 90 days | 6% | 48% |
| Half of them last | 7 days | 85 days |
Seven days against eighty-five. Both columns are true and neither is the right one.
The first answers “if I set myself a habit today, what are its chances?”. The second, “of the people still here after three months, how did it go?”. Together they say one thing: the filter is not in the habit, it is in the first week.
4 · The 66 days from the studies
There is a much-cited study that followed people for twelve weeks. Half of them took around 66 days for the habit to run without thinking, with a spread from 18 to over 250.
It does not contradict us. They measured how long something takes to become automatic in people who stayed three months. We measure how long people last before giving up.
Together they say something uncomfortable: most people quit in the first week on something that takes two months to settle. And since only 8.7% reach 66 days, hardly anyone finds out whether it would have worked.
5 · Nobody announces they are quitting
Of 7,070 people, only 801 ever marked a task as failed. That is 11%. It is not that the rest do everything. It is that giving up is not announced: people just stop opening the app.
The usage data confirms it. Looking at 31 groups of about 1,125 people each:
| Come back | |
|---|---|
| Next day | 58% |
| After 3 days | 27% |
| After 7 days | 17% |
| After 14 days | 10% |
| After a month | ~5% |
42% never come back the day after installing it.
6 · Promising “always” costs double
| Type | Created | Get done |
|---|---|---|
| One-off | 49,509 | 34% |
| Repeating | 60,283 | 17% |
Exactly half. And people still create more repeating tasks. When you create the task you are not promising to do it today, you are promising to do it always. And “always” looks easy from the sofa.
7 · The day of the week barely matters
The best day is Tuesday, with 23,402 tasks completed. The worst is Saturday, with 20,941. Best to worst is 12%. That is flat. If you are waiting for a Monday to start, Tuesday will go the same.
8 · What this data cannot tell you
Which skills people choose. The most common are discipline, intellect and health, but they are the ones the app sets up at the start. Their average level runs from 1.1 to 1.9. We are measuring what we put there, not what anyone picked.
How many habits to start with. The app sets up three at the start and hardly anyone adds more. The “four or more” group is 69 people, below the minimum we publish. The advice to start with three comes from the studies, not from this data.
Whether planning your day works. People who put tasks on a slot in the day complete 26.6 on average, against 20.5 for people who do not. It could be the other way round: the consistent people are the ones who plan. Our data cannot separate them.
What time of day people deliver. We store universal time, not each person’s zone. The spike at two in the morning is almost certainly the afternoon in Latin America.
9 · What we do with this
One conclusion: the problem is not keeping a habit going, it is reaching week two. Skillion is built around that.
When you start, it gives you three-minute tasks so day one is a sure win. Energy, 100 points a day, stops you piling on fifteen things. And if a streak of three days or more is about to break, it nudges you before midnight.
The advice for you comes from the same numbers: make the habit ridiculously small for seven days. Not to build the habit. To cross the stretch where half of them die.
Added-up figures from Skillion’s database and usage analytics, queried on 30 August 2026. No group smaller than 50 people and no individual data.