Behavior science tested its own tricks at scale, and rewards and repetition held up

Key takeaways

• In a megastudy of 61,293 gym members, 45% of 54 programs raised gym visits during their four weeks, but only 8% still had a measurable effect afterward.
• When two government “nudge units” ran 126 trials with 23 million people, the average nudge worked about one-sixth as well as nudges in published academic studies.
• A 2025 Cochrane review rated the evidence as high certainty that rewards help people quit smoking, and the benefit lasted after the rewards stopped.
• Studies that tracked new health habits found typical times of about two months, with individuals ranging from 4 to 335 days.
• In a weight-loss drug trial, people who stopped both the medicine and lifestyle counseling regained about two-thirds of the lost weight within a year.

Thirty scientists from 15 U.S. universities once took on a single goal: get members of the 24 Hour Fitness chain to the gym more often¹.

Together they built 54 different four-week programs, using reminders, rewards, planning tools and more. More than 61,000 members took part, and the results were published in Nature in 2021¹.

The results were humbling. Almost half the programs raised weekly gym visits while they ran, by 9% to 27%¹. But once the four weeks ended, only 8% still had a measurable effect¹.

Then came the twist. Professors, practitioners and ordinary people who were asked to predict the winners failed to pick them¹.

That kind of experiment, called a megastudy, shows where the science of behavior change stands today. Researchers now test their favorite ideas at huge scale, and many shrink. A few hold up well: rewards, repetition and patience. This article covers what’s new and what it means for your own habits.

Megastudies test dozens of ideas at once, and experts guess wrong

A megastudy tests many approaches in the same group of people, with the same yardstick, at the same time¹. That makes it fair to compare one idea against another, which separate small studies can’t do well.

The gym study’s top performer was simple. It gave small rewards to members who came back after missing a workout¹. Rather than punishing a slip, it paid people to recover from one.

A second megastudy, published in 2022, sent text reminders about flu shots to 689,693 Walmart pharmacy patients². Researchers tested 22 different messages against no message at all².

On average, the texts raised vaccination by 2.0 percentage points, a 6.8% boost over the no-text group². The best version sent two texts three days apart saying a flu shot was “waiting for you”². Again, neither experts nor everyday people predicted the winner².

The lesson reaches beyond labs. Expert hunches, including the ones behind a lot of wellness advice, are a weak guide. Testing beats guessing.

Small nudges shrink once every study counts

A nudge is a small change in how a choice is presented, such as a reminder, a default option or a friendly message, that leaves people free to choose. Nudges became famous because published studies made them look powerful.

Then economists Stefano DellaVigna and Elizabeth Linos looked at every trial run by two of the largest U.S. government “nudge units,” published or not³. That covered 126 trials and 23 million people³.

In academic journals, the average nudge raised the target behavior by 8.7 percentage points³. In the nudge units’ complete set, the average was 1.4 points³. That’s still real and statistically solid, but about one-sixth as large³.

Why the gap? The authors estimated that about 70% of it came from selective publication³. Small studies with exciting results got published, while disappointing ones stayed in the drawer. Scientists call this publication bias.

A similar debate played out in the Proceedings of the National Academy of Sciences. A 2022 analysis there pooled 212 published papers and found that nudges had a small-to-medium effect overall⁴. Its authors flagged signs of publication bias themselves and said the true effect was likely smaller than their estimate⁴. In May 2022, the journal posted a correction that updated the paper’s text, figures and supporting data⁴.

Another team then reanalyzed the corrected data using methods that adjust for missing studies. The average effect shrank to near zero⁵. Their statistics were inconclusive for nudges overall but pointed more clearly to no effect for health nudges⁵.

That is not proof that every nudge fails. The reanalysis authors noted that results varied enough that some nudges may still work⁵. Taken together, these studies suggest a single clever trick rarely changes health by itself.

Paying people to quit smoking has some of the strongest evidence

Financial rewards tell a different story. In January 2025, Cochrane, an international network known for careful evidence reviews, updated its review of incentives for quitting smoking⁶. Not counting trials in pregnancy, it covered 48 studies with more than 21,924 people⁶.

People offered incentives were about 50% more likely to be smoke-free six months or more later⁶. The reviewers rated that evidence as high certainty, their top grade⁶.

The obvious worry is that people relapse once the money stops. The review found the benefit held for at least six months after the rewards ended⁶.

The effect was even larger in pregnancy. Across 13 trials with 3,942 pregnant people, incentives roughly doubled quit rates: about 13 of every 100 quit with rewards, compared with 6 of 100 without⁶.

Programs varied widely, from people putting up their own money to rewards worth $45 to $1,185 in total⁶. The review found no clear difference between low and high amounts⁶.

Rewards now sit inside modern treatment, too. The first prescription app for substance use, reSET, was tested alongside a system that rewarded clean drug tests, a method called contingency management¹². Our article on digital therapeutics covers that story.

A new habit takes about two months, and sometimes most of a year

You may have heard it takes 21 days to form a habit. That figure traces back to Psycho-Cybernetics, a self-help book by Maxwell Maltz, and the research doesn’t support it⁷.

A 2024 review by researchers at the University of South Australia pulled together 20 studies with 2,601 people⁷. In this research, a habit is a behavior that has become automatic, set off by a cue without much thought.

Four studies tracked how long that took. Typical times, measured as the median, ran 59 to 66 days, and average times ran 106 to 154 days⁷. Individuals ranged from 4 days to 335 days⁷. In one study, only 23% of people reached the researchers’ bar for a fully formed habit⁷.

The best-known study, by Lally and colleagues in 2010, followed people building healthy eating, water-drinking or exercise habits. The median person took 66 days to get 95% of the way to their peak level of automatic behavior, with a range of 18 to 254 days⁷.

The review also found patterns. Habits people chose themselves grew stronger than assigned ones, and enjoying the behavior helped⁷. In one study, morning stretching became a habit faster than evening stretching, averaging 106 days versus 154⁷. The authors also pointed to a stable setting and a specific plan for when and where⁷.

The authors flagged limits. Eleven of the 20 studies were rated at high risk of bias, samples were small, and most results came from people’s own reports⁷. Treat these numbers as rough guides, not deadlines.

Coaching that reacts in the moment is still being built

Phones and wearables make a new kind of program possible. Researchers call it a just-in-time adaptive intervention, or JITAI: support that changes based on what’s happening to you right now⁸.

The app watches for signals, such as your location or how long you’ve been sitting, and decides whether to step in⁸. One alcohol-recovery app alerted users when GPS showed them near a place they had flagged as risky⁸. An office program prompted workers to move after 30 minutes of nonstop computer use⁸.

A key design rule is that sometimes the best choice is to send nothing at all⁸. A 2018 paper in the Annals of Behavioral Medicine said this research was in its early stages, and that many such programs had been built with little evidence behind them⁸.

Behind the scenes, the National Institutes of Health has tried to make behavior research work more like drug research. From 2009 to 2019, its Science of Behavior Change program funded studies to pin down the mental and social processes that drive behavior, so programs could target them⁹.

It focused on three: self-regulation, which is roughly self-control; how people react to and recover from stress; and relationships with other people⁹. Its library of more than 100 research measures remains available, and seven NIH institutes have co-funded five more years of a coordinating center⁹.

Weight-loss drugs changed the math, not the need for habits

GLP-1 medicines such as semaglutide (Wegovy) have shaken up weight care. They also show how medicine and daily habits fit together.

In the STEP 1 trial, everyone got lifestyle counseling every four weeks, aiming for 500 fewer calories a day and 150 minutes of activity a week¹⁰. After 68 weeks, people in a follow-up group who took semaglutide had lost 17.3% of their weight, compared with 2.0% on placebo¹⁰.

Then the drug and the counseling both stopped. A year later, the semaglutide group had regained about two-thirds of what they had lost¹⁰. Blood pressure and other heart-health markers drifted back toward where they started¹⁰. The drug’s maker, Novo Nordisk, funded the trial¹⁰.

The researchers suggested that the lack of structured lifestyle support after stopping may have added to the regain¹⁰. The National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) says these medicines work best with a lifestyle program and don’t replace healthy eating or activity¹¹.

At one year, people who combine medicine with a lifestyle program lose 3% to 12% more of their starting weight than people in the program alone, NIDDK says¹¹. If you stop, you’ll probably regain some weight, though healthy habits may limit it¹¹.

These drugs aren’t for everyone. NIDDK says semaglutide and tirzepatide aren’t for people with a personal or family history of medullary thyroid cancer or a condition called MEN 2, and weight-loss medicines aren’t advised during pregnancy¹¹. Don’t start or stop one without your prescriber.

What the evidence supports

Approach	What studies found	Evidence level
Cash or voucher rewards to quit smoking	48 studies: about 50% more quitters at six months or later; benefit lasted after rewards ended6	Strong evidence
Rewards to quit smoking during pregnancy	Quit rates roughly doubled6	Strong evidence
Single nudges, such as reminders, run at scale	Small but real average effect of about 1.4 percentage points3	Strong evidence of a small effect
The large nudge effects reported in journals	Mostly disappear once missing studies are accounted for35	Evidence says no to big effects
Short programs to boost gym visits	45% worked while running; 8% still worked afterward1	Promising but early for lasting change
Daily repetition to build a habit	Habit strength rose; about two months typical; many studies at high risk of bias7	Promising but early
Real-time adaptive app coaching	Research in early stages as of 2018; many programs built with little evidence8	Promising but early
Semaglutide plus lifestyle counseling for weight loss	Follow-up group: 17.3% loss vs 2.0% at 68 weeks; about two-thirds regained within a year of stopping both10	Strong evidence while treatment continues

Natural health has always run on practice

None of these findings would surprise a grandparent. Daily practice, a routine tied to the time of day, rewards for doing the hard thing and patience with slow progress are old ideas. What’s new is the measurement.

That measurement cuts both ways. It confirms that repetition matters more than any single trick. It also warns against anyone, from an app maker to a wellness influencer, who promises a fast fix for habits, since even experts couldn’t predict which tricks would work¹².

It also fits with medicine rather than fighting it. A prescription can drive weight down, but in the STEP 1 follow-up, the weight came back when the medicine and support ended¹⁰. Habits are what you keep.

A habit plan built from findings that held up

This routine draws on the studies above. Write your answers down, and keep them where you’ll see them.

1. Pick one behavior you chose and like. Self-chosen, enjoyable habits grew stronger in the research⁷. “Walk 10 minutes after lunch” beats “exercise more.”
2. Tie it to a daily cue. Fill in: “When I _ (something that happens every day, in the same place), I will _.” Studies in the review pointed to specific when-and-where plans as helpful⁷.
3. Try mornings first. In one study, a morning routine became automatic faster than the same routine at night⁷.
4. Plan for about two months, and allow five. Mark a date 66 days out to check how automatic it feels⁷. Don’t give up at day 21.
5. Reward the comeback. After a missed day, give yourself a small reward for getting back on track, the approach that topped the gym megastudy¹.
6. Add stakes for the hardest habits. If you smoke, ask your clinician, employer or a quitline whether they offer an incentive program⁶.
7. Expect a fade. Most program boosts faded once the program ended¹. Decide in advance what will keep you going when a challenge or app subscription ends.
8. Pair medicine with routine. If you take a weight-loss medicine, build eating and activity habits now, and talk with your prescriber before you stop¹⁰¹¹.

For free coaching to help you quit smoking, the CDC points people to 1-800-QUIT-NOW (1-800-784-8669)¹³. If you’re struggling with alcohol or drugs, talk with a clinician about treatment, including programs that use rewards¹².

Questions readers ask

The bottom line

Behavior science has become more honest about itself. Huge experiments and complete records of government trials show that most single tricks produce small, short-lived changes, while rewards, repetition and time do much of the real work.

For your own health, pick one habit you enjoy, tie it to a daily cue, reward yourself for coming back after a slip and give it months, not weeks. If you use medicine, build the habits alongside it.

Related reading

• Learning health skills together works when the group checks its facts
• Pressure makes loved ones hide their habits, and other kinds of help work better
• Wellness apps help a little if you keep using them, and most people don’t
• Virtual coaching can work once you find out who, or what, is coaching you
• Some apps now come by prescription, with evidence that ranges from strong to thin

Sources

1. Milkman KL, Gromet D, Ho H, et al. “Megastudies improve the impact of applied behavioural science.” Nature, 2021. nature.com
2. Milkman KL, Gandhi L, Patel MS, et al. “A 680,000-person megastudy of nudges to encourage vaccination in pharmacies.” Proceedings of the National Academy of Sciences, 2022. researchdiscovery.drexel.edu
3. DellaVigna S, Linos E. “RCTs to Scale: Comprehensive Evidence from Two Nudge Units.” Econometrica, 2022. econometricsociety.org
4. Mertens S, Herberz M, Hahnel UJJ, Brosch T. “The effectiveness of nudging: A meta-analysis of choice architecture interventions across behavioral domains.” Proceedings of the National Academy of Sciences, 2022. pubmed.ncbi.nlm.nih.gov
5. Maier M, Bartoš F, Stanley TD, Shanks DR, Harris AJL, Wagenmakers E-J. “No evidence for nudging after adjusting for publication bias.” Proceedings of the National Academy of Sciences, 2022. pnas.org
6. Notley C, Gentry S, Livingstone-Banks J, et al. “Incentives for smoking cessation.” Cochrane Database of Systematic Reviews, 2025. cochranelibrary.com
7. Singh B, Murphy A, Maher C, Smith AE. “Time to Form a Habit: A Systematic Review and Meta-Analysis of Health Behaviour Habit Formation and Its Determinants.” Healthcare, 2024. pubmed.ncbi.nlm.nih.gov
8. Nahum-Shani I, Smith SN, Spring BJ, et al. “Just-in-Time Adaptive Interventions (JITAIs) in Mobile Health: Key Components and Design Principles for Ongoing Health Behavior Support.” Annals of Behavioral Medicine, 2018. pubmed.ncbi.nlm.nih.gov
9. National Institutes of Health Common Fund. “Science of Behavior Change.” NIH, reviewed July 2025. commonfund.nih.gov
10. Wilding JPH, Batterham RL, Davies M, et al. “Weight regain and cardiometabolic effects after withdrawal of semaglutide: The STEP 1 trial extension.” Diabetes, Obesity and Metabolism, 2022. doi.org
11. National Institute of Diabetes and Digestive and Kidney Diseases. “Prescription Medications to Treat Overweight & Obesity.” NIH, reviewed June 2024. niddk.nih.gov
12. U.S. Food and Drug Administration. “De Novo Classification Request for reSET (DEN160018): Decision Summary.” FDA, 2017. accessdata.fda.gov
13. Centers for Disease Control and Prevention. “Five Reasons Why Calling a Quitline Can Be Key to Your Success.” CDC Tips From Former Smokers, updated November 12, 2024. cdc.gov