Nobody Quits in February
The gym empties after the resolution crowd thins, and everyone draws the obvious moral. Then a dataset of 389,481 training records puts the median dropout at week nineteen — long after the hard part is over.
Walk into any gym in the middle of February and somebody will tell you the story. Look at it now, they say — you should have seen it three weeks ago. The resolution crowd has thinned, the queue for the rack is gone, and everyone still present feels quietly superior for having survived a cull.
The cull is real. February does empty rooms, and the people it empties them of did mean it in January. But the story stops there, and the part it leaves out is the part worth having.
In 2026 Conti, Marzagão, Galpin and Schoenfeld published an analysis in Frontiers in Sports and Active Living covering 389,481 adults using a training app, with workout records running from 2016 to May 2025. It is a far larger sample than this question usually gets. The median time to dropping out was nineteen weeks.
Nineteen. Not six. From a January start, that is the middle of May.
February is a test of willingness. May is a test of whether the thing has a reason — and nobody sets out in January with one built to last that long.
What is actually happening at week nineteen
Consider what week nineteen is like, because it is nothing whatever like week six.
By nineteen weeks the hard part is behind you. You are no longer sore in the way that makes stairs an event. You know where the equipment lives and what to do with it. You are not self-conscious. The lifts have gone up. Everything the February story says defeats people — difficulty, discomfort, humiliation — was survived months ago. And then it stops.
The same study makes the usual explanation harder to hold. Among beginners, twelve-month adherence rose steadily with age: roughly eight in a hundred of the eighteen to thirty-year-olds were still training a year later, against thirteen in a hundred of the over-fifties. The people with the least vigour and the fullest diaries lasted longest. Whatever holds a practice together, it is not energy and it is not enthusiasm, both of which run the other way.
One more finding reframes the early weeks rather than dismissing them. Training consistency across the first twenty-eight days was the single strongest predictor of still being there a year on — but the authors note the protection "attenuated over time". The good start does real work and then wears off. It is a loan against the middle of the year, not an endowment.
So what expires at week nineteen is not resolve. It is the reason. The goal that got you through the door has by then either been met, or revealed itself as much further off than advertised, or simply gone stale through familiarity — and if nothing has quietly replaced it in the meantime, there comes a Tuesday in May when the session has no argument behind it and does not happen. Then another.
None of which makes February innocent, and it would be a cheap trick to pretend otherwise. A median is a midpoint: half of all the leaving happens before week nineteen, and a good deal of that is early. The beginner cohort sheds badly and sheds fast — barely one in ten was still logging sessions after a year, against one in four of the advanced. And the first-month finding vindicates the least glamorous advice in the field, which is to turn up, unremarkably, for four weeks, before you are any good and before it means anything at all. That advice was right. It is simply not sufficient, and it has been sold as though it were. It is worth adding that this is app data rather than a trial: people who stop opening a training app have not necessarily stopped training, and anyone who installs one is already a particular sort of person. The shape of the finding is sturdier than any single percentage inside it.
Planning for May
Plan for the month it gets boring rather than the month it hurts. In practice that means settling now, while the momentum is still there, what the practice is going to be for once the original goal has gone — because the question arrives in May and cannot be answered in May, by which point the honest reply is I don't know and the session has already been skipped.
The study offers one concrete hint. Greater variety in the equipment people used was associated with a lower risk of dropping out. Read that not as a programming tip but as a warning about narrowness: a practice with exactly one expression — one lift, one route, one distance — has exactly one way of becoming boring, and nothing to fall back to in the week it does.
And put something in the diary for week nineteen. Not a test, not a weigh-in. A deliberate hour spent on what this is now for, held at the point where it has stopped being obvious. That question gets asked in May whatever you do. It is only ever a matter of whether you are present when it happens.