How to measure employee wellbeing without surveys
You measure it with objective, continuous data: sleep, stress, recovery and activity, collected every day from the wearable people already own, aggregated into groups of at least 30 people and compared against a baseline. Surveys remain useful for understanding why. Physiological data tells you what is changing, and when.
Why aren't surveys enough?
An engagement survey measures how people say they feel on the day they fill it in. It is an annual snapshot, quarterly at best in its “pulse” form, with three limits that no well-written question can fix.
The first is time. Weeks pass between completion and results; months pass between rounds. A wellbeing programme launched in March produces no comparable number before the autumn.
The second is self-reporting. People answer as they believe they are, as they wish they were, or as they think it is wise to appear. This is not dishonesty: it is how memory of your own sleep or stress actually works.
The third is participation. Response rates fall with every round, and the people who stop answering are rarely the ones doing well.
Surveys have their place. They are the right instrument for causes: the relationship with a manager, a sense of purpose, etc. They are the wrong instrument for knowing whether people sleep better since the programme started. That needs objective measurement, and measurement is what makes a programme one of measurable employee wellbeing.
What can be measured objectively?
A watch, a ring or a band records a handful of physiological signals every day, with no effort from the person:
Sleep: duration, regularity of timing, estimated stages.
Resting heart rate and heart-rate variability (HRV), from which stress and recovery scores are derived.
Activity: steps, active minutes, workouts.
With weight and height entered manually, BMI as well.
These are consumer devices, not medical instruments. A single measurement on a single night may not always be precise to the minute. But a corporate programme does not work with single nights: it works with group averages and trends over weeks. At that level sensor error averages out and the signal remains.
How is data aggregated without seeing the individual?
This is the part that decides whether the programme is acceptable to employees and to the data protection officer.
The principle is simple: the organisation must not be able to identify a person. In practice that takes several rules, written into the software rather than into a policy document. Three of the many rules:
Voluntary enrolment. Each person activates the app with a personal code if they want to, and can leave whenever they like.
A minimum group size. HR sees averages for large groups only. Vitality+ shows a value only for groups of 30 or more people, with at least 10 people with data on the day and a sufficient gender mix; below the threshold the value stays hidden.
Separation of roles. Individual data is visible only to the person, in their app. HR sees adoption, participation and averages of health data by team, site and country.
Which KPIs belong on the dashboard?
Six numbers are enough, and they are the same ones you take to the board.
Adoption: share of eligible employees who activated the programme.
Active participation: share with data synced in the last 7 days. The most honest number, because it separates people who downloaded the app from people who use it.
Average group sleep.
Average group stress and recovery.
Average group activity.
Average group BMI.
Each is read as a change against baseline, by team, site and country: the trend, not the absolute value, tells you whether the programme is working. Everything else (completed sessions, challenges, points) measures engagement with the programme, not wellbeing.
For a practical guide to calculating and presenting the six core KPIs, see Employee wellbeing KPIs: the 6 numbers to take to the board. To see how these metrics are presented in practice, explore the Vitality+ HR dashboard.
How long before a trend appears?
Longer than it seems, and less than a year.
Weeks 1-2: only adoption is measurable. Physiological data is still thin and novelty distorts behaviour.
Weeks 2-4: the baseline forms. It needs at least one full cycle of working week and weekend, ideally two.
Day 90: the first defensible comparison. Three months cover a peak period and a normal one, and absorb the noise of one bad week.
Month 12: year-on-year comparison, the only one that truly neutralises seasonality. July sleep is not January sleep.
Two mistakes to avoid: reading the data day by day, and comparing groups with different composition, such as a shift-working unit against an office.
Where to start: a pilot on one site
You do not need to start with the whole organisation. You need a group large enough to clear the aggregation threshold and homogeneous enough to be readable: one site or function of 100 to 300 people.
A pilot works when two things are clear before it starts.
What HR will and will not see, written on one page and sent to everyone before launch, together with the privacy notice.
What happens to people without a wearable. Either you work only with those who own one, accepting a partial sample, or you provide devices.
Vitality+ offers a free 30-day demo on one site or team, with devices on loan for anyone who does not have one. It is the fastest way to get adoption and a baseline before signing a contract.
In short
Employee wellbeing is measured with objective, daily wearable data, not with an annual snapshot. The organisation sees only averages for large groups, above a threshold written into the software. Six KPIs, a four-week baseline and a reading at 30-60 days are enough to decide.
Related questions
Is a wearable mandatory to measure employee wellbeing? No. Enrolment is voluntary and people who opt out are not counted. A wearable is, however, the only way to get daily objective data; without one you are left with surveys and benefit usage rates, which measure something else.
Is smartwatch data reliable enough for an organisation? For one person there is a margin of error, because these are consumer devices. For a group of dozens of people observed over weeks that error averages out and the trend is readable. That is why a serious programme works with group averages and baselines, not with a single day's value.
How many people do you need for a pilot? Enough to clear the aggregation threshold with room to spare: with a threshold of 30 people per group and realistic adoption, a site or function of 50+ people is the right starting point. Below that, groups stay hidden and the pilot produces no number.