What to Measure

Wellbeing measurement is dominated by lagging indicators and composite scores. Neither tells you what to change.

Leading indicators

These move before outcomes do, and each points at a specific condition.

Hours worked beyond contracted, by team.

Leave taken as a proportion of entitlement, and whether it is taken in usable blocks.

Proportion of time in meetings.

Uninterrupted time available, which is the complement and is frequently near zero for some roles.

Out-of-hours message activity, in aggregate.

Vacancy duration in the team, which is a workload signal.

Manager one-to-one frequency, actually held rather than scheduled.

Time from a concern being raised to a response.

Lagging indicators

Useful for confirming, useless for steering.

Turnover, especially voluntary and especially by team.

Absence, including short-term frequency, which is a better signal than total days.

Grievances and formal complaints.

Recruitment difficulty.

By the time these move, the condition has been present for a long time.

Perception measures worth taking

Short, specific, and mapped to the six areas.

Is your workload manageable most weeks?

Do you have enough say in how you do your work?

Is your contribution noticed?

Do you have support from colleagues when you need it?

Are decisions here made fairly?

Can you do your work without compromising what you think is right?

Six questions. They diagnose. A single engagement score does not.

The composite score problem

Most organisations track one headline number.

It averages across teams with entirely different problems, producing a figure that moves slightly and means nothing.

It cannot be acted on. Nobody can improve a composite.

It invites target-setting, which produces score management rather than change.

Report the components, segmented. The variation between teams is the entire value of the exercise.

Segmentation

By team, always. This is where the actionable variation is.

By manager, carefully — manager quality is a real variable and naming it publicly produces defensiveness.

By tenure. New joiners and long-servers have different problems.

By level. Middle managers frequently score worst and are least likely to be asked.

By working pattern. Part-time, shift and remote workers have distinct experiences that averages erase.

Protect anonymity in small groups by aggregating, and say how you are doing it. A survey people believe is identifiable produces reassuring results and no information.

What not to measure

Individual health data. Steps, sleep, heart rate, app usage. Collecting this crosses a boundary, carries legal exposure in several jurisdictions, and tells you nothing about working conditions.

Attendance as a wellbeing metric, which measures presence rather than capacity.

Programme participation. Sign-ups measure marketing, not effect.

Anything you are not prepared to act on, which is the subject of its own note and is the most common failure in the whole field.

Building the picture from data you already hold

Most of the useful leading indicators are already in systems somebody owns.

HR system: leave utilisation, absence frequency, turnover, vacancy duration, tenure.

Calendar system: meeting hours, largest uninterrupted block, out-of-hours scheduling.

Messaging system: activity by hour, in aggregate.

Recruitment system: time to fill, offer decline reasons.

Finance: overtime and contractor spend, which is a workload signal.

Join them by team. That single join produces a picture most organisations have never seen and requires no new collection.

Start there before commissioning any survey, because it costs nothing and it tells you which teams to ask about.

Reporting the components, not the composite

How to present measurement so that it produces action.

One page per team, not one page for the organisation.

The six areas separately, with the team's score and the organisational distribution.

Two or three leading indicators for that team: hours, leave, vacancy duration.

The trend, which matters more than the level.

No composite score and no target, because both invite management of the number.

Sent to the manager first, with the authority to act on it locally.