The hardest day of a daily habit is day two

Advice about building a daily habit tends to focus on week three, or on the famous idea that it takes 21 days, or 30, or 66. We run a daily geography game and can watch thousands of people try to build exactly this kind of habit, and our data says the interesting moment is much earlier than any of those.

It is the second day. Almost everyone who is going to quit quits there.

The curve has one cliff and then a wall

Take everyone who has ever finished one of our daily puzzles and ask a simple question at each step: of the people who got this far, how many came back for one more day?

Read it left to right and the story is stark. Of everyone who finishes a first daily, 26% ever play a second. Of the people who make it to two, 65% reach three. By four days, 83% continue. By a week, 92%. Past two weeks it is above 90% at every step, and the people who reach twenty days almost never stop.

So the thing that looks like a long slow climb to a habit is really one enormous filter at the start followed by a fairly flat plateau. Getting somebody to day two is worth more than everything you can do for them in week three.

Why day two is so expensive

Day one has everything going for it. Someone clicked a link, they were curious, the novelty carried them. None of that exists on day two. There is no novelty left, nothing has been invested yet, and no streak exists to protect. The only thing that brings a person back is either a reminder that arrives at the right moment, or the memory of having enjoyed it.

It is also why the standard habit numbers are misleading if you read them as a countdown. The often-quoted research by Lally and colleagues found habits reached automaticity somewhere between 18 and 254 days, with a median around 66. That is a real finding about how long automaticity takes, but our curve says almost nobody is still deciding at day 66. They decided on day two.

What this changes about advice

Most self-improvement advice is written for the middle of the curve: don't break the chain, track your progress, reward yourself at 30 days. Useful, but aimed at people who were going to continue anyway. If the cliff is at day two, the useful advice is narrower:

• Set up the return before you need it. On day one, while you still care, do the one thing that makes day two happen on its own: a reminder, a home screen icon, a fixed slot in your morning. On day two you will not be motivated enough to arrange it.
• Make day two small. The second session should be shorter and easier than the first, not a step up. Anything that takes more than a couple of minutes is competing with a person who has not yet decided they care.
• Treat a single missed day as nothing. Our returning players miss days constantly and carry on. The break in the chain is not what ends habits; the absence of a reason to start again is.
• Stop counting to 21 or 66. If you are past your first week, the data says you are far more likely to continue than to stop, whatever the number says.

A note on what this data can and cannot say

These are people who chose to play a geography game, which is not a random sample of humanity, and the figures count anyone who ever returned rather than who returned the very next morning. A newcomer who started this week has had less chance to continue than one who started in June, so the later steps lean toward committed players by construction. The shape is robust across every way we have cut it, but treat the exact percentages as ours rather than universal.

If you want to try the habit itself

Five countries a day takes about two minutes. If it sticks, the effect on what you actually know is measurable: we have written up how long it takes to learn the world map and the spaced repetition method behind it. And if you want to know what you are likely to be missing, there is the world's blind spots.

Play today's five countriesfree, no signup, about two minutes