The best data on resolutions comes from a randomized trial, not a poll. Oscarsson, Carlbring, Andersson, and Rozental recruited 1,066 people from the general public, let them write their own resolutions, randomized them into three levels of support, and followed up for a year. They published it in PLOS ONE in December 2020.
Two findings are worth building a plan around. At the one year follow up, 55 percent of responders considered themselves successful at sustaining their resolution, which is far higher than the 8 percent figure that circulates in January listicles with no source attached. And participants with approach oriented goals succeeded at 58.9 percent versus 47.1 percent for those with avoidance oriented goals.
approach versus avoidance, concretely
An avoidance goal is phrased as removal: stop eating sugar, quit skipping the gym, cut out late snacking. An approach goal is phrased as an action you add: eat fruit at breakfast, train Tuesday and Thursday at 7, cook dinner four nights a week.
The difference is not motivational packaging. An approach goal names a behavior you can perform and verify. An avoidance goal names an absence, and you can only confirm an absence once the day is over, which means the only available feedback is failure. Eleven percentage points of success rate separated the two framings in the trial.
Rewriting works. “Stop drinking soda” becomes “drink sparkling water with lunch.” “Lose 20 pounds” becomes “eat 100 grams of protein a day and lift three times a week,” which is a behavior, while the 20 pounds is an outcome you do not directly control.
the 21 day claim is invented
The 21 day habit number came from a 1960 book by a plastic surgeon, Maxwell Maltz, who observed that his patients took about 21 days to adjust to a changed appearance. It was never a study of habit formation.
The real measurement is Lally and colleagues at University College London, published in the European Journal of Social Psychology in 2010. They tracked 96 people adopting a new daily behavior and measured automaticity. The median time to reach an asymptote of automaticity was 66 days, with a range from 18 to 254 days depending on the person and the behavior.
Read the caveat too, because it matters more than the headline. The 66 day median came from the subset of participants whose data actually fit a habit formation curve, roughly 39 of 82 with usable data. So it describes people who succeeded, not people who tried. A realistic expectation is months, and a complicated behavior like cooking dinner will take considerably longer to automate than a simple one like drinking a glass of water after waking.
writing it down, and what that does
Writing goals down is standard advice, and the mechanism that has held up in research is narrower than the advice suggests. What produces results is not the writing itself but specifying when and where. Gollwitzer’s work on implementation intentions, summarized in a 2006 meta-analysis with Sheeran covering 94 studies, found that if then plans specifying a situation and a response produced a medium to large effect on goal attainment.
“Exercise more” is a wish. “Monday, Wednesday, Friday at 6:30 a.m., gym bag in the car the night before” is an implementation intention. The second one has a trigger, a time, and a removed obstacle.
The support finding in the 2020 trial has an odd wrinkle worth reporting honestly. The group that got some support did significantly better than the others. The group that got extended support did not outperform the minimal control. The authors could not fully explain it, and more information did not translate into better outcomes. That is a reason to be suspicious of elaborate systems, tracking apps, and 40 page plans. They feel like progress and may be consuming the effort the behavior needed.
a 30 day challenge is a start date, not a plan
Short challenges get people moving, and there is nothing wrong with a 30 day structure as an entry ramp. The failure mode is that 30 days ends. Day 31 has no structure attached to it, the intensity of the challenge was never sustainable at the pace you were holding, and the habit has not had 66 days to become automatic.
If you run one, decide in advance what day 31 looks like, and make it smaller than the challenge. Someone who trained six days a week for 30 days should plan three, because three completed every week for a year beats six completed for a month and then abandoned.
the number of resolutions to make
One. Possibly two if the second is trivially easy.
Self regulation research is mixed on whether willpower depletes the way the ego depletion literature once claimed, and several large replication attempts have failed to reproduce the classic effect. But the practical argument does not need that theory. Changing a behavior requires attention, planning, and rearranging your week. Attempting five simultaneous changes means five half built routines competing for the same evening hours, and the most common outcome is that all five collapse together and the conclusion drawn is a personal one about discipline rather than a structural one about load.
Pick the change that makes the others easier later. For most people that is sleep or training frequency, because both improve the odds that the food decisions go well without requiring separate effort.