Choosing Outcome Measures a Workplace Commissioner Will Actually Accept

Somebody with budget authority is going to ask how you’ll know the peer-support group, the lunch scheme, or the mentoring pairing actually worked. If your answer is a satisfaction survey handed out at the last session, you have already lost the argument. Satisfaction data tells you people who showed up liked it. It tells a commissioner nothing about whether the thing was worth paying for again.

This is a procedure for choosing measures before you launch, not after, so the numbers you produce in month six are ones a finance director or external commissioner will treat as evidence rather than as marketing.

Step 1: Decide whether you’re measuring isolation, loneliness, or both

These are not the same construct, and commissioners increasingly know it. Isolation is structural — how many people someone has contact with, how often. Loneliness is subjective — the gap between the contact someone wants and what they get. Research on remote healthcare workers found the two track differently and respond to different interventions; a study of pandemic-era isolation and age found the relationship between them shifts across the life course. If your programme increases contact (more team lunches, more cross-department pairing) but doesn’t touch how connected people feel, measuring only loneliness will make a working intervention look like a failure. Pick your construct first, then pick the instrument that measures it.

Step 2: Use a validated instrument, not a homemade one

A four-question pulse survey your HR system generated is not evidence. The UCLA Loneliness Scale is the instrument the academic literature uses, which means your results are comparable to published benchmarks — a large AARP survey of adults 45 and older used it and found the difference between people who talk to their neighbours and those who don’t (33% lonely versus 61%). If you use the same scale, you can say your post-programme score moved toward the “connected” end of a distribution other people recognise. If you invent your own five-point “how connected do you feel” question, you can say nothing comparable to anyone.

Step 3: Pick the business-facing outcome the commissioner actually cares about, separately

A commissioner rarely funds a connection programme because loneliness is bad in the abstract. They fund it because someone told them it costs money. Cigna’s 2020 workforce data found lonely workers missed work twice as often for illness and five times as often for stress-related reasons — that is the kind of number that gets a programme renewed. Gallup’s 2024 workplace data adds a structural angle: fully remote employees report loneliness at 25% against 16% for fully on-site staff, which matters if your commissioner is deciding between remote-first policy and space for in-person contact. Track absenteeism, turnover intent, or utilisation of EAP services alongside the connection score, not instead of it. The connection score explains the mechanism; the business metric is what gets you the next budget cycle.

Step 4: Take a baseline before anyone hears about the programme

Measure the week before launch, not the week people sign up. People who volunteer for a connection programme are often already more socially motivated than the workforce as a whole, and a baseline taken after recruitment will overstate how lonely your starting population actually was. This is the same self-selection problem that limits every voluntary wellbeing intervention, and it does not disappear because you measured carefully afterward.

Step 5: Be honest about what a control group would cost, and what happens without one

The intervention evidence for loneliness is thinner than the correlational evidence. The 2024 HEAL-HOA trial is notable precisely because it is one of the few randomised trials of a loneliness intervention rather than an uncontrolled pilot report — most of the field runs on before-and-after numbers with no comparison arm. If you can run even an informal staggered rollout — one team starts in month one, another in month three, giving you a natural comparison — do it. If you can’t, say so plainly in your report rather than implying causation you can’t support.

Step 6: Don’t borrow mortality-risk numbers to describe a six-week programme

Holt-Lunstad’s meta-analysis linking social isolation and loneliness to elevated mortality risk, and the American Heart Association’s statement on cardiovascular outcomes, are population-level findings built on decades of exposure. They are the right citation for why workplace connection matters as a category. They are the wrong citation for what your team lunch achieved by September. The AHA statement itself names the absence of intervention evidence as the field’s central gap — use that candour rather than papering over it.

What this means in practice: report three numbers, not one — a validated loneliness or isolation score, a business metric the commissioner already tracks, and a plain statement of what you can’t yet prove because you didn’t have a comparison group. That combination reads as competent. A single satisfaction score with a mortality-risk citation attached to it reads as a pitch deck.

The failure mode: the attendance-as-outcome trap

The most common mistake is reporting attendance or session count as the outcome — “40 people joined the walking group” — because it’s the easiest number to produce. Attendance is an input. A qualitative synthesis of social prescribing found the benefit people describe is restored participation and purpose, not contact for its own contact’s sake, and that structured, purposeful activity outperforms unstructured contact. Measure whether the activity did what it claimed to do, not whether people turned up.

What this does not solve

Measurement discipline will not fix the fact that voluntary workplace programmes mostly reach people who were already inclined to join something. It will not shorten the timeline commissioners want results on — most credible connection effects take longer to show than a single budget cycle allows. And no combination of validated scales will substitute for the randomised evidence the field still largely lacks. What good measurement buys you is a report a commissioner can trust the next time you ask for money, which, in this line of work, is most of the point.