The honest answer is a range, not a number: somewhere between about three weeks and eight months, with a median around two months, depending almost entirely on how hard the behaviour is and how consistent the cue is.
The 21-day figure everyone quotes is a misreading. It comes from a 1960 plastic-surgery book. Maxwell Maltz noticed that patients took "a minimum of about 21 days" to stop seeing the old face in the mirror, an observation about adjusting to a change in appearance, not about forming a habit. The self-help industry dropped the "minimum of about" and kept the number.
The number worth knowing instead comes from Lally et al. (2010), which followed 96 people forming a daily habit and measured how long it took the behaviour to reach automaticity. The median was 66 days. The range was 18 to 254.
What the 66 days actually measures
Three details from that study change how you should read the number.
It measures automaticity, not effort. The endpoint was the point at which the behaviour stopped requiring a decision. You will be doing the thing long before that; what changes at the end is that you stop having to choose to.
The difficulty of the habit dominates. Drinking a glass of water after breakfast automated far faster than doing fifty sit-ups. The spread from 18 to 254 days is mostly this, not individual variation. If you want a habit to form quickly, the lever is making it smaller, not trying harder.
Missing one day did not measurably affect the outcome. This is the finding that matters most and gets quoted least. The automaticity curve was not disrupted by an occasional miss. What disrupted it was a run of misses.
The shape of the curve
Automaticity does not build linearly. It rises steeply at first and then flattens, which produces two predictable experiences, both of which people misread.
Days 1–10 feel like fast progress, because they are. Most of the total gain in automaticity happens early. This is why the first week of a new habit feels easy and encouraging.
Weeks 3–8 feel like nothing is happening. The curve has flattened; the day-to-day difference is now too small to notice. This is where almost everyone quits, and they quit for the same reason: they were measuring by how it felt, and how it feels stops changing long before the habit is done forming.
Knowing the curve's shape is useful. The flat stretch is not failure, it is the middle.
What actually shortens it
In rough order of effect:
Make the behaviour smaller than you think it should be. Two minutes, not twenty. The size of the habit is the strongest predictor of how fast it automates, and a two-minute version that automates in a month can be grown afterwards. A twenty-minute version that never automates cannot be shrunk into existence.
Fix the cue and never move it. Automaticity is a cue-response link, so a habit that happens at a different time each day is building a weaker link every day. The most reliable cue is the completion of something you already do without thinking: after the kettle boils, after the laptop closes, after you park.
Same context, every time. Same room, same order, same trigger. Context is part of the cue, which is why habits famously collapse on holiday and after a house move: the cue was partly the place.
Reduce the friction to near zero. Shoes by the door, book on the pillow, app on the first home screen. Every second of setup between the cue and the behaviour is a place where the chain breaks.
What doesn't shorten it
Motivation. It is real and it is front-loaded, which is precisely the problem: it is highest on day one and lowest during the flat stretch when you need it. Systems that depend on feeling motivated fail on a predictable schedule.
Tracking harder. A more elaborate tracker does not form the habit faster. It does one useful thing: it tells you the truth about your consistency, which your memory will not. Adding more to it than that is a way of doing something adjacent to the habit instead of the habit.
Streak pressure. Streaks are motivating right up until the day one breaks, and then they are the reason people quit outright over a single missed Tuesday. Which brings us to the number that should replace them.
Count the window, not the run
Given that the research says missing a single day doesn't matter, the metric should be one that agrees: completions in a fixed window, not consecutive days.
"22 of the last 30" survives a missed Thursday. "Streak: 0" does not, and it is also less accurate, because it reports that nothing happened, when in fact twenty-two things happened.
This is why a fixed 30-day frame is a more useful unit than an open-ended counter. It is long enough to see the curve flatten, short enough to actually finish, and finishing is a thing that can happen. An infinite counter can only ever be broken.
Every habit gets its own 30-day grid rather than an endless list, and the schedule options mean a four-times-a-week habit has no broken streak on a Tuesday. The yearly heatmap is the honest long view, because it shows the pattern of misses, which is the thing that predicts whether a habit is forming.
What it doesWhat to expect, week by week
A rough calendar for a small daily habit with a fixed cue:
| Week | What is happening |
|---|---|
| 1 | Novelty is doing the work. Feels easy. Do not scale up. |
| 2–3 | The first genuine misses. This is normal and not a signal. |
| 4–6 | The flat stretch. Nothing feels like it is improving. Most quitting happens here. |
| 7–9 | You notice you did it without deciding to. That is the endpoint. |
| 10+ | Maintenance. Now it is safe to make it bigger. |
If you are at week ten and it still takes a decision every time, the usual cause is a cue that moves: the behaviour is happening, but not attached to anything that reliably triggers it.
After it forms
A formed habit is not permanent, it is cheap. It still breaks under a context change (new job, new house, illness, travel) because part of the cue was the context, and the context left.
The fix is not a new resolution. It is re-anchoring the habit to a cue that exists in the new context, and expecting a couple of weeks of it costing effort again. That is much shorter than starting over, which is the reasonable thing to tell yourself when a two-month habit disappears during a fortnight away.