Corporate Wellness Programs That Actually Work (and What a Small Manhattan Company Should Do Instead)

Most corporate wellness programs are built around a health risk assessment, a portal, a few webinars and a gift card for finishing them. The two largest randomized trials of that model in the US found the same thing: people who joined reported better habits, and nothing else moved. Not cholesterol, not blood pressure, not health spending, not sick days, not performance. If you run a company of 10 to 60 people in a Manhattan office building, that is useful news, because the one thing those programs did change is the one thing you can run in a conference room on a Tuesday.

The short version

• The standard wellness program (screening, portal, modules, incentives) has now been tested in two large randomized trials, for up to three years. Self-reported exercise and weight management improved. Clinical measures, medical spending, absenteeism and job performance did not.
• The famous “three dollars back for every dollar” number comes from a 2010 review of mostly non-randomized studies at large companies. The same lead authors ran the big trial a decade later and found no savings.
• Participation is the problem nobody prices in: under half of employees get screened, a fifth or fewer of those flagged do anything, and about a quarter of employer-paid gym memberships are never used once.
• Exercise itself has strong evidence for mood, anxiety and stress. The workplace question is only whether you can get people to do it. Scheduled, live group sessions beat prerecorded videos and email content on adherence in the one small trial that compared them, and group exercise beats solo exercise in a 44-study meta-analysis.
• For a small Manhattan company the practical answer is a fixed weekly class at the office, Tuesday to Thursday, with the founder in the room. Pricing is by proposal and depends on team size, format and frequency.

What the big studies actually found

The BJ’s Wholesale trial (JAMA 2019, Health Affairs 2021)

Zirui Song and Katherine Baicker randomized 160 BJ’s Wholesale Club worksites, 32,974 employees in all, to a comprehensive wellness program or nothing. The program was eight modules on nutrition, physical activity and stress, run by registered dietitians, with gift cards for completion. At 18 months, employees at program sites were more likely to say they exercised regularly (69.8 percent versus 61.9 percent) and managed their weight (69.2 versus 54.7). That was it. Across 10 clinical markers, 38 spending and utilization measures, absenteeism, tenure and job performance, there was no significant difference. Average participation was 1.3 modules out of 8. Song and Baicker, JAMA, 2019.

The three-year follow-up in Health Affairs showed the same pattern, and about 28 percent of employees at program sites had completed even one module. The authors wrote that the results might “temper expectations of substantial improvements in health outcomes or financial returns on investment from wellness programs.” Song and Baicker, Health Affairs, 2021, figures via MedCity News.

The Illinois study (QJE 2019, JAMA Internal Medicine 2020)

At a large public university, 4,834 employees were randomized individually, with cash incentives, into a program of screening, assessment and wellness activities. More than 56 percent of those offered it signed up. The people who signed up already had lower medical spending and healthier habits than those who did not. After one and two years the program had no measurable causal effect on spending, 16 clinical biometrics, health behaviors (other than more screening and more people reporting a primary care doctor), productivity or self-reported health. The confidence intervals ruled out roughly 84 percent of the savings estimates in earlier published studies. NBER, Illinois Workplace Wellness Study.

The selection finding matters more than the null result. If your vendor shows you that participants cost less than non-participants, that is what you would expect even if the program did nothing, because the healthy people are the ones who join.

RAND for the Department of Labor (2013)

RAND’s review of employer data found that fewer than half of employees (46 percent) complete a screening or health assessment, and of those flagged for an intervention, “a fifth or less chose to participate.” Weight change over three years worked out to about one pound. The estimated cost difference was 157 dollars a year and not statistically significant. In a separate seven-year analysis of PepsiCo’s program, disease management for people with existing conditions returned about 3.80 dollars per dollar; lifestyle management, the exercise and nutrition side, returned about 50 cents. RAND RR-254 and RAND RB-9744.

Where “three dollars back” came from

The number every brochure quotes, 3.27 dollars in medical savings per dollar spent, is from a 2010 Health Affairs meta-analysis by Baicker, Cutler and Song. It pooled 36 studies, more than 90 percent at large firms, only nine of them randomized or matched, with costs imputed for seven studies that never reported them. The authors themselves flagged selection bias as “a major concern.” Two of those three authors then ran the BJ’s trial above and found nothing on spending. Baicker, Cutler and Song, Health Affairs, 2010. If a proposal promises you a return on claims, ask which randomized trial it is based on. There is not one.

The UK study on apps and resilience training (2024)

William Fleming at Oxford used survey data from 46,336 workers across 233 organisations to compare people who had used individual-level wellbeing offerings (apps, mindfulness, resilience and stress training, coaching, relaxation classes) with matched colleagues who had not. Participants “appear no better off.” Volunteering was the one offering associated with better wellbeing. The design is a single snapshot, so it cannot prove cause, and it did not look at exercise classes at all. Fleming, Industrial Relations Journal, 2024.

The participation funnel is the whole problem

Put the numbers in a row and the pattern is obvious. 46 percent of employees get screened (RAND). A fifth or fewer of those flagged take up an intervention (RAND). 28 percent of employees at BJ’s program sites completed at least one module over three years. Industry data from 2012 and 2013 put use of employee assistance program counseling at about 3 percent of employees, and use of the stress management services most EAPs include at about 5 percent (Business Insurance, 2014). In the UK, an analysis of public gym data found almost one in four employer-paid memberships was never used once (ukactive, 2019).

Large employers spent an average of 716 dollars per employee on wellness incentives in 2023, according to the Fidelity and Business Group on Health survey of 184 big companies (Fidelity, 2023), and the results above are what that money buys. A 30-person company does not need to replicate the funnel at a smaller scale. It needs to skip it.

The one thing that moved: exercise

Look back at the BJ’s trial. Across 29 self-reported outcomes, the two that improved were exercise and weight management. That is not a coincidence. Exercise is the component of a wellness program with real evidence behind it; the delivery model is what fails.

• Mental health. An umbrella review of 97 systematic reviews covering 1,039 trials and 128,119 people found physical activity reduced depression, anxiety and psychological distress with moderate effect sizes, “comparable to, and slightly larger than” medication and psychotherapy in trials. Higher intensity did more. This is general and clinical populations, not a workplace study. Singh and colleagues, British Journal of Sports Medicine, 2023.
• Exercise on company time does not cost output. In a cluster-randomized study at Swedish dental clinics, 177 women were assigned to 2.5 hours a week of mandatory exercise during paid work hours, to shorter hours with no exercise, or to a reference group. Production (patients treated) was maintained or increased in all groups; the exercise group reported higher work capacity and fewer sick days, with small improvements in blood glucose (effects the authors called small and varied). One industry, one country, small sample. von Thiele Schwarz and colleagues, 2008 and the 2011 follow-up.
• Class days are better work days. Two hundred desk workers at three Bristol companies rated their mood and performance on one day they exercised at work and one they did not. Mood and self-rated performance (on mental and interpersonal demands, output and time demands) were all better on exercise days. Self-selected and self-reported, so treat it as a pattern rather than proof. Coulson, McKenna and Field, 2008.
• Productivity evidence is real but thin. A 2019 systematic review of 39 workplace nutrition and physical activity studies found 14 with significant effects on productivity, performance or work ability; about 28 percent of the studies were high quality. Grimani and colleagues, BMC Public Health, 2019.

For scale: the WHO recommends 150 to 300 minutes of moderate activity a week (the CDC says 150) plus strength work on two or more days (CDC). One 45-minute office class is roughly a third of the moderate minimum and, if it includes strength work, one of the two strength days. It is a real contribution, not the whole answer, and it is the part of the week most people would otherwise skip.

What works when the team is small and the office is half full

The Partnership for New York City’s March 2025 survey of major Manhattan employers found 57 percent of office workers in on an average weekday, 76 percent of pre-pandemic levels, with three days a week the most common pattern. Firms with fewer than 500 employees ran higher, at 67 percent. (Partnership for NYC, 2025). Fridays are the dead day: REBNY’s weekday visitation data put midweek attendance at 73 percent of 2019 and Friday at 43 percent (REBNY, 2023). Design around that.

1. A fixed slot, not a library

In a small trial of 84 inactive adults, adherence over eight weeks was 75 percent for live scheduled (videoconference) classes, 40 percent for prerecorded videos and 31 percent for education emails (Dawson and colleagues, JMIR Formative Research, 2024). Pick one day, Tuesday to Thursday, one time, and put it on the shared calendar as a standing meeting. The point of a fixed slot is that it happens whether or not anyone felt like it that morning.

2. A true group, not a drop-in

A meta-analysis of 44 studies found that exercising in a “true group” (same people, same time, a shared identity) beat a standard class, and a standard class was no better than exercising at home with some contact, which in turn beat exercising alone (Burke and colleagues, 2006). A 20-person company already is a true group. Keep the same instructor every week so it stays one.

3. The founder shows up

In a HERO and Mercer analysis of nearly 450 employers, assessment completion ran 59 percent where leadership visibly supported the program versus 41 percent where it did not (HERO, 2011). That is employer self-report and correlation, not a trial, but in a small company it is also just common sense. If the person who signs the checks is on a mat in the conference room at 12:15, the class is real.

4. Add competition, not devices

Handing out step counters does little on its own; a Cochrane review of 14 trials found pedometer programs “may have no effect” once they end (Cochrane, 2020). Structure around them does more. In the STEP UP trial of roughly 600 Deloitte employees with overweight or obesity, people grouped in threes and ranked on a weekly leaderboard (the competition arm) added about 920 steps a day during the program and were still 569 steps ahead three months after it ended (Patel and colleagues, JAMA Internal Medicine, 2019). A quarterly step competition in teams of three costs almost nothing.

5. The cheap add-ons, honestly rated

• Stair prompts. Signs at the elevator raised stair use by a median 2.4 percentage points, a 50 percent relative increase, across 11 studies reviewed by the Community Preventive Services Task Force. Note the base: a 50 percent relative gain on a small number. Community Guide.
• Walking meetings. A pilot with 17 workers added about 10 minutes of moderate activity a week. Fine, and small. Kling and colleagues, 2016.
• Sit-stand desks. Cochrane’s 2018 review found they cut sitting by 84 to 116 minutes a workday, on very low quality evidence, with no harms found. They reduce sitting. They do not add exercise. Cochrane, 2018.

What a program looks like for a 10 to 60 person office

This is what we run for small companies in Manhattan office buildings. The details are on the fitness classes for small teams page; the shape is simple.

• One weekly class, 45 minutes, fixed day and time. Same instructor every week. Capped at 15 per session so it stays a group, not a crowd; larger teams split into two slots.
• Three formats. HIIT when the team wants to sweat, Pilates for people who sit all day, yoga or meditation for the week everyone is fried. Most companies pick one and stay with it; some rotate by month.
• Your space, not a gym. A cleared conference room, a roof or terrace, the nearest park, or your building’s gym if it has one. We bring mats and any equipment. Nothing is installed.
• Flat rate per class, whoever shows up. No per-head billing, no minimums that punish a quiet week. Pricing is quoted for your team size, format and frequency rather than off a menu, so it is on request.
• Insured, with the paperwork your office manager will ask for. Certificate of insurance on request, participation waivers handled by us.

If your building rather than your company is the one organizing, that is a different program: resident classes and private training paid for by the building, described on the trainer-led fitness for your building page.

No sales call. Tell us the team size, the days people are in and the room you have, and we will send a one-page overview of how it would work for your office. Request a proposal or text (917) 951-0111.

The rest of the wellness menu, rated

The old version of this post listed fourteen program types. Here is the honest version.

• Flexible and hybrid work. Already the norm in Manhattan (69 percent of surveyed employers are hybrid, Partnership 2025). Not a wellness program so much as a baseline; the question is what you do on the in-office days.
• Employee assistance programs. Worth having for the small number of people who need one; industry figures put use at single digits. Do not count on an EAP as your wellness strategy.
• Health risk assessments and biometric screening. The core of the model that the BJ’s and Illinois trials tested. Low take-up, no measured effect on outcomes, and incentives tied to results bring legal rules with them (below).
• Wellbeing apps, resilience and mindfulness training. In the Fleming data, users were no better off than matched colleagues. If people want an app, fine, but it is not a program.
• Gym stipends. The UK data suggests about a quarter of employer-paid memberships go unused; the ones that are used get used often, which points to the already active.
• Nutrition programs, financial wellness, sleep programs, flu shots, lunch and learns. Reasonable, low-risk, and mostly untested at small scale. Disease management for people with existing conditions is the exception: RAND’s PepsiCo data showed it did pay, about 3.80 dollars per dollar. That is a health plan question, not a wellness one.
• Volunteering days. The one offering in Fleming’s data associated with better wellbeing. Cheap, social, and worth a quarterly slot.

A note on incentives and the law

Programs that tie rewards to a health outcome (a biometric target, a weight goal) fall under HIPAA and ACA rules that cap the incentive at 30 percent of the cost of coverage, 50 percent where tobacco is involved, and require a reasonable alternative (Department of Labor FAQs). Programs that ask medical questions also sit under the ADA and GINA, where the EEOC’s 2016 incentive rules were vacated by a federal court effective 2019 and the proposed replacement was withdrawn in February 2021, with nothing finalized since (K&L Gates, 2021). A voluntary class that collects no medical information and pays nobody for their blood pressure sits outside all of that. This is general information, not legal advice; if you are attaching money to health data, talk to counsel.

Frequently asked questions

Do corporate wellness programs actually save money?

No rigorous trial has shown it. The two largest randomized studies (BJ’s Wholesale, 32,974 employees; Illinois, 4,834 employees) found no significant effect on medical spending after up to three years. The widely quoted 3.27 dollars per dollar figure comes from a 2010 review of mostly non-randomized studies at large firms.

What is the one thing a wellness program reliably improves?

Self-reported exercise. In the BJ’s trial, regular exercise rose from 61.9 to 69.8 percent at program sites and nothing downstream moved. Exercise itself has strong evidence for mood, anxiety and stress, so a program built directly around getting people moving is the version with the best odds.

How often should an office fitness class run?

Once a week, same day and time, is the realistic floor for a small company and the schedule that survives. Twice a week is better for fitness and still practical for teams that are in three or more days. Tuesday to Thursday, when Manhattan attendance peaks.

Do we need a gym or special equipment?

No. A cleared conference room fits 10 to 15 people for Pilates, yoga or a mat-based HIIT class. Roofs, terraces and nearby parks work from spring to fall. Your building’s gym works if it has one. We bring mats and small equipment.

What does an office fitness program cost?

It is quoted per team, as a flat rate per class based on team size, format and frequency, so pricing is on request. Request a proposal and we will send a one-page overview for your office.