You're looking at the quarterly dashboard. 68% participation in the health risk assessment. 42% biometric screening completion. 1,400 logged gym visits. Your vendor calls it a record quarter. Your CFO nods approvingly. You feel pretty good.
Those engagement numbers are among the most deceptive metrics in a benefits stack. They can paint a picture of health improvement while hiding a simpler fact: the dashboard counts activity, and activity is not the same as better health.
The most heavily promoted wellness programs, judged by today's standard metrics, are often the ones that have accidentally built the strongest Behavioral Artifact Generation Systems, or BAGS. Employees learn to produce the signals the interface rewards: logging points, finishing challenges, collecting incentives. The dashboard records the signals. Health stays off the dashboard.
Engagement metrics measure system use
Logins, points earned, and challenges completed measure system interaction. A health outcome is a different unit: a lower blood pressure reading, a completed colonoscopy, a filled prescription for a chronic condition. The two get conflated constantly, and most benefits leaders don't notice the conflation.
Incentives make the conflation worse. In KFF's analysis of the employer survey, 50% of workers at large firms completed a health risk assessment when an incentive was attached, versus 31% at firms without one. Participation climbs when the program pays people to show up, and climbing participation reads like progress.
Consider the employee who opens the wellness portal three times a week to log water intake, track steps, and claim a $10 coffee voucher. The CRM flags that person as a power user. What happened is that the employee found the fastest path through the interface to rack up points with minimal physical effort. The system recorded engagement with the reward interface. It captured nothing about the employee's health.
The failure is easy to state. You are tracking compliance with the program's interface. A user who types "15 cups of water" every day, regardless of how much they drink, optimizes that interface and shows up as highly engaged in a Power BI dashboard. The employee who quietly quits smoking and sees a therapist twice a month, without logging a single point, shows up as disengaged. The dashboard can't tell the two apart.
High engagement can hide poor outcomes
Put the two on a ratio:
Outcome-per-click = reduction in biometric risk ÷ increase in system clicks
In a healthy program, genuine improvements in blood pressure, BMI, or stress should come with a moderate increase in clicks, from logging coaching calls or updating food diaries. Meaningful change with efficient system use produces a high ratio.
The randomized evidence puts that ratio near zero. In a 2019 JAMA trial covering 32,974 employees at a US warehouse retailer, worksites with a wellness program reported higher rates of exercise and weight management, but clinical markers, health care spending, and utilization showed no significant change after 18 months. A randomized trial at the University of Illinois found no significant effects on 40 of 42 outcomes in the first year, including medical spending and productivity.
That is the pattern to watch for. A vendor report showing 70% engagement is a measure of logins. The login count can rise while health stays flat, and the user who keeps clicking without changing health is the one who keeps the engagement number high for your CFO. The program quietly becomes a gamified data-entry system.
Three fixes for misleading engagement data
Three changes can surface real impact instead of noise.
1. Track silent improvement over active participation
Look for the employee who declines every challenge but whose biometric readings and claims data show steady improvement. Flag high-log, low-outcome users as a liability. Quiet improvement is worth more than loud logging.
2. Audit for data entropy
Run a backend query on your wellness platform. How many users log water intake that matches the daily recommendation exactly, every single day? That pattern reflects robotic data entry. Irregular, imperfect logging is a better sign of a real person reporting real behavior.
3. Ask for the outcome correlation
Ask your vendor for the Pearson correlation between engagement clicks and the reduction in high-cost claim codes. Randomized trials have found only weak links between program activity and clinical or spending outcomes, so expect a number close to zero. When that is what comes back, the program is generating activity rather than change.
Measure verified preventive actions
The useful response is a different unit of measurement. A completed screening that a lab or provider filed, a filled prescription, a documented biometric improvement: those are verifiable events. A self-reported step count is not verifiable.
The incentive research points the same way. A literature review found that incentives raised self-reported health assessment participation but could not establish an effect on actual screening completion. Self-reported activity moves when you pay for it; verified completion is harder to fake.
WellthCare™ builds its Health-to-Wealth™ benefit on that distinction. Preventive actions are verified through standardized preventive care codes, and every plan of care is AI-drafted, then reviewed by a nurse practitioner and a physician before it reaches the employee. Rewards are tied to the verified action.
Any employer can apply the same test to a wellness or prevention program. Ask which metric moves when someone's health changes, and which metric moves when someone opens the app more often. Fund the first kind.
Start with the outcome data
High engagement is easy to celebrate and easy to misread. The question that matters is whether the people behind the numbers are getting healthier. Low participation is a visible problem. High participation that tracks activity instead of health is a quieter one, and it is more expensive to leave in place.
The dashboard can report strong activity while the workforce gets no healthier. Check the outcome data before you trust the engagement score.