The 21/90 Rule: Does a Habit Really Take 21 Days?
On this page
The 21/90 rule says it takes 21 days to build a habit and 90 days to make it a permanent part of your lifestyle. Research doesn’t support a fixed number like that. In a study of people repeating one new daily behavior, the time it took them to get close to their peak automaticity ranged from 18 to 254 days (Lally et al., 2010), and the average was around 66 days (Gardner et al., 2012). So keep 21 days as your first checkpoint and plan for two to three months. Some habits will take longer than that, and that’s normal.
What the 21/90 rule says
The rule has two parts. Do something every day for 21 days and it becomes a habit. Keep going until day 90 and it becomes part of who you are. It’s easy to remember, and three weeks sounds doable, which is a big part of why it spreads.
The trouble is what happens on day 22. If the new habit still takes effort, the rule makes it look like you failed. Usually you haven’t. You’re just on a more typical timeline.
Where the 21 days came from
Habit researchers trace the 21-day idea to Maxwell Maltz’s 1960 book Psycho-Cybernetics. It grew out of anecdotal observations of plastic surgery patients, who usually got used to their new appearance in about three weeks (Gardner et al., 2012). That’s about getting used to a new face in the mirror, which is a different thing from making a new behavior automatic.
The 90-day half has no clear source. It does land closer to what studies found, as you’ll see below.
What the research actually found
The habit study behind “66 days”
Researchers at University College London asked 96 volunteers to pick one eating, drinking or activity behavior and do it daily in the same situation, such as after breakfast, for 12 weeks. Each day, participants rated how automatic the behavior felt and recorded whether they did it. Data from 82 people was complete enough to analyze (Lally et al., 2010).
Automaticity rose along a curve. The gains came fastest early on and then slowed to a plateau (Gardner et al., 2012). The time it took people to reach 95% of their own plateau ranged from 18 to 254 days. The authors said this shows how much the timeline varies and that it can take a very long time (Lally et al., 2010). On average, automaticity leveled off around 66 days after the first day (Gardner et al., 2012).
A 2024 review of 20 studies
A systematic review pooled 20 studies with 2,601 adults, covering behaviors like physical activity, diet, flossing and drinking water. Four of the studies reported how long habits took to form. Their median times ranged from 59 to 66 days and their averages from 106 to 154 days, with individual times from 4 to 335 days. The authors wrote that this refutes the popular idea that habits form in about 21 days (Singh et al., 2024). They also rated most of the studies at high risk of bias, so the numbers are a guide, not a promise.
How the rule compares
| The 21/90 rule says | The research found |
|---|---|
| 21 days to build a habit | An average of around 66 days, and a range of 18 to 254 days in one study |
| 90 days to make it a lifestyle | Median times of 59 to 66 days and averages of 106 to 154 days across four studies |
| Every day counts, so don’t break the chain | Missing one day did not materially affect habit formation |
Writing for family doctors, the UCL researchers suggest telling patients to expect a daily habit to take around 10 weeks (Gardner et al., 2012). If you want a single number, that’s a more honest one than 21.
Missing a day doesn’t reset you
This may be the most useful finding of all. In the UCL study, missing one opportunity to do the behavior did not materially affect how the habit formed (Lally et al., 2010). Progress soon picked up again after one missed day (Gardner et al., 2012).
So the all-or-nothing rule in some challenges, where one slip sends you back to day 1, is stricter than the research calls for. If you miss a day, do it at the next chance and keep your count going. (If you like hard challenges but want rules you can keep, see these 75 Hard alternatives.)
How to build a habit on a realistic timeline
- Choose it yourself. In the 2024 review, habits people chose for themselves formed more strongly (Singh et al., 2024). Pick something you actually want to do.
- Make it small enough to do every day. The UCL team’s advice to patients was to choose a simple action you can do daily (Gardner et al., 2012). Two pages beats a chapter you skip. One push-up beats a workout you dread.
- Attach it to the same moment every day. Pick a time and place you meet daily, and do the action every time you meet it (Gardner et al., 2012). “After I pour my coffee” works better than “sometime in the morning.”
- Do it often. More frequent practice sped up habit formation in the review (Singh et al., 2024). A daily habit gets more repetitions than a three-times-a-week one.
- Track it, and check how it feels. Tick off each day, and at the end of each week rate how automatic it feels so you can watch it get easier (Gardner et al., 2012).
- Miss once, then come straight back. One missed day is not a reason to start over.
A timeline you can plan around
Here’s a way to read the research without pinning yourself to one number.
- The first few weeks are when you’ll notice the biggest jump, because gains in automaticity come fastest early on. It still takes effort, and that’s expected.
- Around day 21 is a good first checkpoint. Ask whether the cue is working and whether the action is still small enough. Adjust before you quit.
- Around 10 weeks is a realistic point to expect many daily habits to feel close to automatic.
- Past three months is normal for some habits and some people. The longest times in the studies ran well beyond 200 days.
If you want a picture of where you’re headed, writing to the person who has already built the habit can help. These future-self journaling prompts are a good place to start.
Today’s small step
Pick one habit and write it as a single sentence: “After [something I already do every day], I will [tiny action].” Put a mark on today’s date when you do it, and put a note on your calendar 10 weeks from today to check how automatic it feels. Start there.
Frequently asked questions
Is the 21/90 rule true?
Not as a fixed rule. In a study of people repeating one new daily behavior, the time it took them to get close to their peak automaticity ranged from 18 to 254 days, and the average was around 66 days. A 2024 review of 20 studies found median times of 59 to 66 days. Twenty-one days is a fine first milestone, but most people need longer, and some need much longer.
What is the 66-day rule?
It comes from a University College London study in which volunteers repeated one daily behavior in the same situation for 12 weeks. Their sense of doing it automatically rose and then leveled off on average around 66 days after they started. It's an average, so treat it as a rough guide. In the same study, some people got there in 18 days and others were estimated to need 254.
Does missing one day ruin a new habit?
No. In the habit study behind the 66-day figure, missing one opportunity to do the behavior did not materially affect how the habit formed, and progress soon picked up again after a single missed day. The useful rule is to do it at the next chance instead of restarting your count.
How long does it take to break a bad habit?
The studies here measured how long it takes to form a new habit, not how long it takes to drop an old one, so they don't give a number for that. Habit researchers suggest focusing on a new behavior you want to do, tied to a cue you meet every day, rather than only trying to stop something.
Sources
- How are habits formed: Modelling habit formation in the real world, European Journal of Social Psychology (Lally et al., 2010). Accessed October 30, 2026.
- Making health habitual: the psychology of 'habit-formation' and general practice, British Journal of General Practice (Gardner, Lally & Wardle, 2012), via PubMed Central. Accessed October 30, 2026.
- Time to Form a Habit: A Systematic Review and Meta-Analysis of Health Behaviour Habit Formation and Its Determinants, Healthcare (Singh et al., 2024), via PubMed Central. Accessed October 30, 2026.