Five Structural Tests for Whether You Built a Service or a Community

Your walking group has fourteen regulars and a short waiting list. Attendance data looks good, the wellbeing scores have moved, and the funder is pleased. Now answer a different question: if you stopped paying someone to be there on Thursday morning, would anyone walk?

That question separates a service from a community, and most programmes cannot answer it because they were never designed to. A service delivers a defined activity to people who arrive as recipients. A community is a set of relationships among the people who arrive, in which the activity is partly a pretext. Both are legitimate. The failure is claiming one in a bid and building the other in practice — then being surprised when the group evaporates the week the contract ends.

This framework gives you five tests, the design consequences of each answer, and an honest account of how thin the evidence is underneath all of it.

The five tests

Run these on a group you already operate. Answer from what happened last month, not from what you intend.

1. The absence test. A regular stops coming for three weeks. Who notices, and who makes contact? If the answer is your staff member checking a register, you have a service. If two members texted them in week one, you have something else. This is the single most diagnostic question, because it tells you where the relational load sits.

2. The side-channel test. Does contact between members happen outside the scheduled hour — a lift shared, a WhatsApp thread, a coffee afterwards that nobody scheduled? Ray Oldenburg’s 1989 account of third places rests on the informality: the value of a hangout comes from unstructured contact around the ostensible purpose, not from the purpose itself. A group with no side channel is a service with good attendance.

3. The role test. Do members hold roles with real consequence — holding the key, ordering the tea, deciding the route, welcoming a newcomer? A role only counts if the group would notice its absence. Handing out a job title with no consequence attached produces resentment, not ownership.

4. The boundary test. Who decides who joins? Services take referrals from outside; communities have opinions about their own membership. This one cuts both ways, and it is where the framework gets uncomfortable — a group with real say over its boundary will sometimes exclude the person your referral pathway most needed it to take.

5. The empty chair test. Cancel the facilitator for one session, with notice, and see what happens. Not a thought experiment. Actually do it. Whether anyone turns up tells you more than a year of satisfaction surveys.

What each answer costs to run

Design decision	If you are running a service	If you are building a community
Intake	Continuous referral; churn is expected	Cohort or slow trickle; every new arrival resets the group
Group size	Whatever the room and the staffing ratio allow	Small enough that everyone tracks everyone — Robin Dunbar’s layers put close-friend capacity at roughly 15
Staff role	Deliverer of the activity	Convener who withdraws, deliberately and on a schedule
Duration	Time-limited by design, with a defined discharge	Open-ended; a hard end date is a design flaw
Primary metric	Attendance, completion, individual outcome change	Contact between members, roles held, survival after funding
Failure mode	Attendance drops and you close it	It survives but closes to newcomers

The staffing implication is the one funders resist. A community requires someone whose job is to become progressively less necessary, which is very hard to write into a job description and harder to evidence quarterly.

What this means in practice Pick one, write it down, and staff for it. If you are commissioned for eight-week blocks with continuous referral and individual outcome reporting, you are commissioned to run a service — build a good one and stop promising community as a by-product. If you want the group to outlive the grant, you need to reduce facilitator presence on a stated timetable from week one, not attempt a handover in the final month.

The facilitator-shaped hole

Here is the failure mode to name in your team meeting. A group runs well for eighteen months. The facilitator is warm, reliable, and knows everyone. Relationships in the room are real but they all route through her — she remembers the birthdays, she chases the absentees, she smooths the friction between two members who irritate each other. She leaves. Within six weeks the group has stopped.

Nothing went wrong with the relationships. The problem is that the network was a star, not a mesh, and the facilitator was the hub. The facilitator-shaped hole is created by good practice, which is why it is so hard to spot: the better someone is at holding a group together, the more thoroughly they become the thing holding it together.

The counter is structural, not attitudinal. Split the hub’s functions across at least three members early — absence-chasing, logistics, welcome — and rotate them. Then run the empty chair test twice a year.

How much of this is actually evidenced

Be careful here, because the community claim is where practice runs furthest ahead of the literature.

Claim	Evidence status
Structured, purposeful group activity outperforms contact alone	Reasonable. A 2022 qualitative meta-synthesis in BMC Health Services Research found participants described benefit extending past social contact into restored participation and purpose
Social prescribing reduces loneliness for people who engage	Consistent but weak in design. All nine studies in the 2021 Perspectives in Public Health review reported positive individual impacts; three reported reduced GP, emergency or inpatient use
Programme effects are comparable across sites	Not supportable. The 2021 systematic review in IJERPH reported gains in self-esteem and confidence alongside limited trial evidence and substantial heterogeneity
Shared physical space raises rates of social contact	Argued, not trialled. Eric Klinenberg’s case for social infrastructure, including the 1995 Chicago heat wave, is persuasive but observational
Member-held roles help groups survive the end of funding	No direct evidence found. Practice inference. Treat as a hypothesis you are testing
Population-level connection interventions work	Largely untested. The American Heart Association’s 2022 scientific statement named the absence of intervention evidence as the central research gap, and the 2023 difference-in-differences evaluation of the UK Campaign to End Loneliness is unusual precisely because most campaign evaluation in this field is uncontrolled

The Surgeon General’s 2023 advisory, reporting that around half of US adults experience loneliness, makes strengthening social infrastructure and local institutions the first of six pillars in the accompanying national strategy. That is a mandate to build the mesh. It is not evidence that you know how.

One suggestive data point for why informal contact is worth designing for: AARP’s 2018 survey of 3,020 adults aged 45 and over found 33% of those who had spoken to their neighbours were lonely, against 61% of those who never had. Cross-sectional, direction unclear, but consistent with everything Robert Putnam documented about the decline of informal socialising rather than only of formal membership.

What this does not solve

Nothing here fixes reach. Every test above assumes people already in the room. A group with a strong boundary, real member roles and a live side channel is by construction harder for a stranger to enter — you have traded accessibility for durability, and you should say so out loud when you report it.

Putnam’s distinction between bonding and bridging capital is the warning. The five tests measure bonding. A group that passes all five may have become excellent at holding its own members and useless to anyone new, which is a worse outcome for a commissioner buying population reach than a mediocre service with an open door.