The anatomy of well-being: what the five WHO-5 dimensions reveal about employee health

The WHO-5 Well-Being Index produces a single score from 0 to 100, and almost every report that uses it stops there. A company learns its average is 68, decides that sounds acceptable, and moves on. But the composite score hides the structure underneath it. The WHO-5 is built from five separate items, and those items do not move together.

We analyzed 65,626 WHO-5 responses from 2,912 employees across 74 companies over two years, from April 2024 to January 2026, and pulled the five dimensions apart. The question was simple: which parts of well-being are strong, which are weak, and which ones most sharply separate employees who are thriving from those who are struggling?

The answer is consistent enough to be called a structural feature. Across every one of the eight quarters in the study, the ranking of the five dimensions never changed. Employees report being most cheerful, most interested, and most energetic. They report being least rested and second-least calm. Rest and calm are not just low on average; they are the dimensions that collapse the hardest when overall well-being declines.

One point on the instrument helps before the findings. Each WHO-5 item asks how often, over the past two weeks, the respondent felt a particular way: cheerful, calm, active, rested, and interested in daily life. Items are rated 0 to 4, multiplied by 4 to give a 0 to 20 dimension score, and summed for the 0 to 100 composite. The recognized clinical thresholds are below 28 for likely depression, below 50 for low well-being, and 72 or above for high well-being.

How the workforce distributes

Before the dimensions, the headline distribution. The mean composite WHO-5 across the full sample is 67.9 (SD 16.5), with a median of 68. That places the typical employee comfortably in the moderate band. But the tail matters more than the average.

About 12.5% of the workforce, 365 employees, sits below the clinical threshold of 50. The 1.1% in the depression range is small as a percentage but real as a count: 32 people. A population-average view of well-being treats these employees as a rounding error. The dimension-level view, as the rest of this article shows, makes clear exactly what kind of support they need.

Part A — The dimension hierarchy holds for two straight years

We aggregated the five dimension scores by quarter, from Q2 2024 through Q1 2026. The sample size per quarter grew from 1,319 to 2,052 as the platform expanded, but the ordering of the dimensions did not budge.

The rest deficit

Rested is the floor of well-being. It is the lowest-scoring dimension in every quarter, with a two-year average of 12.97 out of 20, or 64.9%. At its worst point, Q3 2024, it fell to 12.21, a full 1.49 points below the next-lowest dimension, Calm at 12.55. Even at its best, in Q1 2026 at 13.74, it still trailed every other dimension. The pattern is unambiguous: employees feel more cheerful, more interested, and more energetic than they feel recovered. This is consistent with the broader occupational health literature, in which rest is typically the first casualty of demanding work.

The calm gap

Calm is the persistent second weakness. The item asks whether the respondent has felt calm and relaxed, and it averages 13.22 out of 20 across the full period, or 66.1%. That sits 0.78 points below the Cheerful average of 14.00 and 0.81 points below Active at 14.03. Employees can summon energy and hold their mood, but the underlying sense of peace lags. A workforce that scores high on energy and low on calm is best described as activated but not settled: productive, and carrying tension.

Every dimension is rising

The composite climbed from 65.7 in Q2 2024 to 71.6 in Q1 2026, a 9.0% increase. Every dimension contributed, and the dimensions that started lowest improved the most. Rested gained the most in absolute terms, up 1.29 points from Q2 2024 to Q1 2026, followed by Calm at +1.12. The gap between the highest and lowest dimension also narrowed slightly, from 1.11 points in Q2 2024 to 0.96 points in Q1 2026. The dimensions are converging, though slowly. As the limitations section notes, part of this upward trend may reflect the changing composition of the sample rather than improvement within a fixed population.

Part B — Which dimensions separate thriving from struggling

The trend data shows the dimension hierarchy. The next question is diagnostic: when an employee's overall well-being is low, which dimensions have collapsed? We compared the 1,190 employees with high well-being (composite ≥ 72) against the 365 with low well-being (composite < 50).

The recovery divide

Rested produces the widest raw gap of any dimension, 9.72 points, nearly a full point wider than Interest at 8.93 and almost two points wider than Cheerful at 7.73. When well-being collapses, rest collapses hardest. The low group's Rested score of 6.68 out of 20, 33.4% of the maximum, is the most alarming single number in the study. It means the average low-well-being employee rates their sleep and recovery at barely a third of full. This is not mild dissatisfaction. It is a population in chronic under-recovery.

Energy is the most reliable signal

Rested has the biggest raw gap, but Active has the largest standardized effect, a Cohen's d of 4.42. The reason is the unusually tight distribution of Active scores within the high group (SD 1.92). High well-being employees do not just have high energy on average; they almost uniformly have high energy, with very little variance. That makes Active the most reliable diagnostic of the five. If someone's energy is low, their overall well-being is, with near certainty, low too.

Mood degrades last

Cheerful shows the smallest gap between groups, 7.73 points. Mood is where the low-well-being group is closest to the high group. Even employees scoring below 50 on the composite manage Cheerful scores of 9.04 out of 20, 45.2% of the maximum, still below midpoint but meaningfully above their Rested (33.4%) or Calm (39.3%) scores. Mood is the most resilient dimension. People hold some degree of cheerfulness even when rest, calm, and energy are deeply compromised.

The low-well-being profile is not "everything is equally bad." It has a specific shape: mood is relatively preserved while physical recovery and inner peace are the most depleted. That points to a practical sequencing rule for the low group. Prioritize rest and stress management over mood-boosting activities; the mood tends to follow once the physiological foundations are restored.

Part C — The workplace factors behind the gap

Parts A and B describe well-being. Part C asks what work conditions accompany it. We compared the same high and low well-being groups across six workplace factors, with sample sizes varying by factor due to data availability.

Recognition is the strongest lever organizations control

Daily happiness has the largest single effect at d = 2.11, with the high group averaging 4.44 out of 5 against 3.40 for the low group. But daily happiness is close to a restatement of well-being itself; the direction of causality is ambiguous. Among the factors organizations can directly act on, recognition dominates. High well-being employees rate recognition at 0.661, near the best possible score, while low well-being employees rate it at 1.457, nearly midscale. The d of 1.59 is larger than for resources (1.47), manager quality (1.27), or feedback (0.69). If an organization wants the highest-leverage well-being intervention, recognition is where the data points.

Resources and manager quality follow close behind. Together with recognition they form a coherent cluster of organizational support factors. Feedback is the weakest of the six, but its d of 0.69 still clears the conventional "large" threshold. The likely reason it ranks last is that feedback is episodic; recognition, resources, and manager quality are experienced continuously.

The stress paradox

The most counterintuitive result in the dataset: high well-being employees report more stress, not less. The high group averages a stress level of 1.71, between slightly and moderately stressed, while the low group averages 1.27. The d of 0.73 is a large effect by conventional standards. This is not noise.

The finding echoes the eustress, or challenge-stress, literature. Moderate stress is not the enemy of well-being; it is a byproduct of engagement and meaningful work. The high well-being group is more recognized, better resourced, better managed, and happier, and it is also more stressed, most plausibly because its members are more engaged in demanding, visible work that produces productive tension. Low well-being employees, by contrast, appear to experience not more distress but less of everything: less recognition, fewer resources, less challenge. The same pattern appeared in our prior manager research, where managers, who scored higher on most well-being dimensions, also reported more stress than individual contributors.

What this means

The composite WHO-5 score is a starting point, not an answer. The dimension profile tells you what kind of support a team or an individual actually needs. Four decisions should use the dimensions directly.

A simple version of the energy rule for early warning: if you can track only one dimension, track Active. With a Cohen's d of 4.42 it is the most reliable signal of overall well-being status, far more diagnostic than mood, which can be maintained through social performance even while well-being collapses.

Limitations

This study describes the structure of well-being, not its causes. Several caveats shape how far the findings travel.

• Cross-sectional group comparison, not causation. Parts B and C compare groups defined by their WHO-5 score. The workplace factors that differentiate them may be causes, consequences, or co-symptoms of well-being differences.
• Self-report on every measure. Both WHO-5 and the workplace factors are self-reported. Shared method variance may inflate associations, since a person who feels good may rate everything more positively.
• Completeness requirement. Requiring all 5 WHO-5 items in the same month excludes partial responders. If struggling employees are more likely to skip questions, the sample may underrepresent the lowest-well-being group.
• Platform growth confounds the trend. The sample grew from roughly 1,319 to 2,052 employees per quarter. The upward composite trend may partly reflect compositional change rather than improvement within a fixed population.
• Low-group heterogeneity. The low group combines the 28–49 band (n=333) with the depression range below 28 (n=32). The small depression-range n prevents separate analysis.
• Scale directionality in Part C. Several workplace factors use inverted scales where lower is better, which complicates cross-factor comparison; effect-size comparisons across factors should be read with care.
• No demographic or role controls. The analysis does not adjust for seniority, tenure, job function, or demographics. Some associations may be confounded by role-related factors.