Why Company-Wide Engagement Numbers Can Mislead You

Your organization may report a 45% wellness participation rate and still have one location at 75% and another at 12%.

Those are not small variations around an average.

They are very different employee experiences.

This is why company-wide engagement numbers can be useful for reporting and weak for diagnosis.

An average tells you what happened across the population. It does not tell you where the strategy worked, where it failed, or what employees experienced along the way.

We like to segment engagement by the parts of the organization that actually shape daily life: location, department, shift, role, manager, eligibility group, or work environment.

The right segment depends on the workforce.

In manufacturing, site and shift may matter most.

In a public-sector organization, police, fire, transportation, administration, and field services may have completely different routines and communication patterns.

In a distributed workforce, geography, access method, and supervisor structure may explain more than the corporate average.

Propel analyses suggest that participation often clusters at the closest meaningful work level, such as a department or location.

This fits a broader behavioral idea: people experience culture locally.

Your corporate brand may still matter overall, but to a particular employee, their team matters every day.

Segmenting results helps us ask better questions.

If one plant outperforms another, what is different?
Did managers communicate differently?
Was there a stronger champion?
Did employees have easier access?
Was the challenge structure more relevant?
Did one site have more time or opportunity to participate?
Was the local culture more supportive?

The same logic applies to low-performing segments.

A shift that barely participates may not be disengaged. The communication could just be arriving while they are asleep.

Employees may lack work email. The competition may favor desk workers. The activity may be incompatible with the job.

Once you see the segment, you can investigate the mechanism.

A recent example:

We were speaking with a company that was struggling with engagement among a particular segment of their employees.

The benefits leader said, “I went down there myself and told them about the program. Nobody had any idea. It seems like nothing is getting through to them.”

When we asked what their workday looked like, he said, “Oh they meet for about 15 minutes first thing in the morning, then head off to different sites.”

Bingo.

This one segment wasn’t working like the others because their day was completely different. They had no immediate social proof with the program, limited time to hear about what was available, and likely had managers that couldn’t get the point across in that small window of time.

There is also an important fairness issue. Comparing groups with fundamentally different opportunities to participate can turn a leaderboard into a measure of job design rather than motivation.

Company-wide reporting still has value. Leaders need a simple view of performance.

But the next layer of the dashboard should answer a different question: where is the experience meaningfully different?

Segmentation is especially useful when it leads to qualitative follow-up.

Pick one high-performing unit and one low-performing unit and ask the same five questions in both:

How did employees hear about the program?

What did the manager say?

Was participation feasible during the workday?

What part felt most valuable?

What part was confusing?

The contrast can surface local practices worth scaling and barriers worth removing. Do not assume the high-performing group is simply more motivated. Their environment may be doing more of the behavioral work for them.

The goal is not to create dozens of tiny dashboards. Segment only where the distinction can change a decision and where privacy can be protected.

A few well-chosen cuts of the data are usually enough to show whether the organization truly has one engagement problem or several local ones hiding inside the average.

Want more? Check out our full library of past issues here.