The “92% of resolutions fail” statistic that opens half the productivity articles on the internet has no traceable primary study behind it. The best actual research says something more hopeful: 46% of resolvers were still succeeding at six months — versus 4% of people with identical goals who made no resolution. Resolutions don’t fail; unsupported resolutions do.
Every January the same number does the rounds: “92% of New Year’s resolutions fail.” It opens keynotes, coaching funnels and roughly half the productivity articles ever written — usually one paragraph before someone sells you the fix. While writing my own book about habits I went looking for the study behind it, because I keep a rule about not publishing numbers I can’t source. I never found it. What I found instead was a chain of blogs citing blogs citing blogs — and, under all that sediment, one real, careful study saying something almost nobody quotes, presumably because hope sells worse than doom.
Where did the “92% fail” statistic come from?
Nowhere traceable — that’s the honest answer. The “92% fail” figure, and its cousin “80% fail by the second week of February,” circulate with no primary study behind them; follow the links and you land on other articles pointing at each other, and the trail dead-ends in marketing copy. I won’t tell you who invented them, because I don’t know — and neither do the people quoting them. A number with no study isn’t a statistic. It’s a rumor with confidence.
What does the real research actually show?
The best study on the question followed people for six months: New Year’s resolvers versus people who wanted to change the same things but made no resolution. At six months, 46% of the resolvers were still continuously successful — versus 4% of the comparison group. Read that carefully: the resolution frame didn’t doom anyone; it multiplied the odds of still succeeding roughly tenfold. The honest caveats: outcomes were self-reported, and resolvers weren’t randomly assigned — people who are ready to change are the ones who resolve. Even so, the direction of the finding embarrasses the doom statistic.
Self-reported outcomes; resolvers weren’t randomly assigned — readiness to change differs.
Why does January work at all — and do you have to wait for it?
Because fresh starts are measurably real. Field data shows gym visits, “diet” searches and commitment contracts all spike after temporal landmarks — a new year, a new month, a new week, a birthday. The calendar keeps handing you clean slates, and behavior responds to them. Which also means January holds no monopoly: any Monday is a landmark, and so is the day after you read this. My own 67-day protocol starts whenever you do — that is precisely the point of it.
Why do resolutions collapse — and how do you make one that survives?
- Vague intentions with no start line. “Get healthy” has no first rep. My failed years were full of resolutions that never specified a day one.
- All-or-nothing restarts. One missed day becomes a lost quarter — not because of the miss, but because the restart ritual (wait for Monday, wait for the 1st) costs more than the miss ever did.
- No tracking. With nothing visible, week three feels identical to week zero, and quiet quitting follows.
- Everything at once. Six resolutions is six young habits competing for one adult’s attention. Mine used to collapse as a stack, by Thursday.
- Motivation as the plan. January’s emotion is launch fuel. Nobody budgets for February, when the resolution has to run on structure instead.
A resolution that survives looks like a project, not a wish: ONE goal, a hard frame with a start and a finish, a forgiving miss policy agreed with yourself in advance, and a visible record. The frame even has a research-backed length: new habits took a median of 66 days to become automatic in the best real-world study, with a range of 18 to 254. A resolution is annual; a habit is a roughly ten-week project. January’s job is not to fix your life. January’s job is to launch one ten-week project.
Interactive tool — estimate how long YOUR habit will take across the observed 18–254-day range (loads with the app version of this page).
| The claim you’ve seen | What actual studies measured |
|---|---|
| “92% of resolutions fail” | No traceable primary study — blogs citing blogs citing blogs |
| “80% fail by the second week of February” | Same: the citation trail dead-ends in other articles |
| “Resolutions are pointless” | Norcross 2002: 46% of resolvers still continuously successful at six months vs 4% of non-resolvers with the same goals |
| “January is arbitrary” | Dai et al. 2014: goal behaviors measurably spike after temporal landmarks — new year, month, week, birthdays |
| “A habit takes 21 days” | Lally 2010: median 66 days to automaticity, range 18–254 |
Quick answers
Do New Year’s resolutions work?
Better than the internet says. The key study followed New Year’s resolvers and a comparison group who wanted the same changes but made no resolution: at six months, 46% of resolvers were still continuously successful versus 4% of non-resolvers — roughly a tenfold difference from the frame alone. Honest caveats: success was self-reported, and resolvers weren’t randomly assigned, so some of the gap is readiness rather than the resolution itself. But the direction is unambiguous. Declaring a change, with a date attached, measurably beats privately wishing for one. The failure isn’t in resolving — it’s in resolving without structure.
Why do resolutions fail by February?
First, the famous version of that claim — “80% fail by the second week of February” — has no traceable study behind it; I looked. What is real is the mechanism that makes February dangerous: January supplies emotion, and by week three the emotion is spent while visible results haven’t arrived yet. If the resolution was vague, untracked and all-or-nothing, that gap is fatal — one missed day triggers a restart ritual, and the restart never comes. The fix isn’t more willpower in February. It’s structure built in January: one goal, a written policy for missed days, and a record that makes week three look different from week zero.
Is it better to set resolutions or goals?
It’s a false choice — a resolution is just a goal with a famous start date. The distinction that actually matters is supported versus unsupported. An unsupported resolution (“get fit this year”) has no start line, no finish line, no miss policy and no record; it fails identically whether you call it a goal, a resolution or an intention. A supported one is a project: one behavior, a ten-ish-week frame — the median in habit research is 66 days to automaticity — a forgiving rule for missed days, and something visible tracking the reps. Call it whatever you like. Build it like a project.
When is the best time to start a resolution?
The next temporal landmark you can reach — which is probably Monday. The fresh-start research found goal behaviors spike after landmarks of every size: new years, new months, new weeks, birthdays. January is simply the largest landmark, not a magically effective one, and waiting months for it costs far more than the bigger fresh-start feeling is worth. My own 67-day protocol is deliberately calendar-free: it starts the day you start, and day one does the same job January does — it draws a clean line between the old data and the new. Pick the nearest line and step over it.
The 92% stat is a rumor with confidence; the real study found resolvers succeeding at ten times the rate of non-resolvers. Resolutions don’t fail — unsupported ones do. Pick one goal, give it a frame and a record, and start at the next landmark. Monday works.
Ready to stop reading and start measuring?
Everything on this blog is one tap away in the app — the forgiving streak, the 67-day protocol, the proof.