Scaling a Connection Programme Without Losing the Thing That Worked

A funder likes your twelve-week group programme. Retention is strong, participants report feeling less isolated, and the referral pipeline works. Now they want it in six more sites next year, at three times the caseload. This is the point where most programmes quietly stop being the programme that got funded, and start being a diluted version of it that happens to share a name.

The failure mode has a name worth using: the photocopier problem. You can photocopy a manual, a referral form, and a logo. You cannot photocopy the facilitator who remembered everyone’s name in week two, the small room that made six strangers feel like a group rather than an audience, or the slack in the schedule that let a session run long because someone needed it to. Scale-up plans that treat the programme as the manual will look identical to the original on paper and feel nothing like it in the room.

What actually drove the outcome — and what you’re guessing at

Before you scale anything, separate what you know caused the result from what you’re assuming caused it. Most small programmes cannot tell the difference, because they were never designed to.

A 2022 qualitative meta-synthesis on social prescribing found that participants describe the benefit as extending well beyond social contact itself — toward restored purpose and meaningful participation, and that structured, purposeful group activity appears to work better than contact alone. That’s a real and useful finding. But it also tells you exactly where your risk sits when you scale: the “structure” and “purpose” are the active ingredients, and those are precisely the things a facilitator improvises session to session, not the things written down in a delivery manual.

The wider evidence base on social prescribing is honest about its own limits. A 2021 systematic review of social prescribing and loneliness found that all nine included studies reported positive individual impacts, and three reported reductions in service use — but the review base is small and heterogeneous, and a separate systematic review the same year flagged the same problem: real gains in self-esteem and confidence, but limited trial evidence and wide variation across programmes. This is not evidence that social prescribing doesn’t work. It’s evidence that nobody has isolated which part of most programmes is doing the work, which is exactly the question a scale-up forces you to answer.

If you don’t know which part is load-bearing, you’ll cut it first, because it’s usually the expensive, unscalable, human part.

The four things that are hardest to preserve, in order

1. Group size and composition. The mechanism in almost every credible model of informal connection — going back to Ray Oldenburg’s account of the third place and Eric Klinenberg’s argument that shared physical spaces shape rates of contact — depends on a group small enough that people become known to each other, not just co-present. Scaling by doubling group size to hit throughput targets is the single most common way a programme stops working while its attendance numbers go up.
2. Facilitator judgment. If your facilitator adapts pacing, reads the room, and knows when to let a session run over, that judgment is a skill, not a script. Scaling by hiring at the same rate but training only on the manual will produce facilitators who run the schedule, not the room.
3. Local referral relationships. If your pipeline depends on a GP, link worker, or community organisation that trusts the programme because they’ve seen it work, that trust doesn’t transfer with a new site number. New sites start with cold referral relationships, which changes who shows up and why.
4. The unmeasured slack. Time to arrive early, stay late, decompress. Budgets for scale-up almost always compress the schedule to fit more cohorts through the same space. This is usually the first thing cut and the last thing anyone thinks to ask about.

What this means in practice: before you write a scale-up budget, write down the four things above and rate your confidence, high or low, that each will survive the move to six sites at three times the volume. Anything rated low needs its own line item and its own person accountable for protecting it — not a mention in the fidelity checklist.

What to measure differently at scale

At small scale, you can get away with measuring only the outcome — loneliness scores, self-reported wellbeing, attendance. At scale, you need to measure fidelity as a distinct thing from outcome, because if fidelity drops before outcomes do, you have a warning you can act on. If you wait for outcomes to drop, you’ve already lost a cohort’s worth of participants to a programme that stopped working without anyone noticing.

Practical fidelity measures that are cheap to collect and hard to fake:

• Group size at each session, not just at enrolment (attrition changes composition mid-cohort).
• Facilitator tenure and hours of shadowing before solo delivery.
• Time from referral to first session, by site (a proxy for whether local relationships are working or cold).
• Session length actually delivered versus scheduled.

None of these are outcome measures. That’s the point. Outcome measures tell you whether something went wrong; fidelity measures tell you what went wrong, in time to fix it before the next cohort.

Evidence status: what you can and can’t claim to a funder

Claim	Evidence status
Small, structured group programmes are associated with reduced loneliness and improved self-esteem	Reasonably supported by qualitative synthesis and multiple systematic reviews, though trial quality is weak
A specific programme’s outcomes will replicate at 3x scale with the same fidelity	Not established — this is a management assumption, not a research finding
Group size and facilitator quality are the active ingredients (rather than contact per se)	Suggested by qualitative work, not proven by controlled comparison
Social connection interventions can be tested with proper randomised designs, not just uncontrolled evaluation	Demonstrated — the 2024 HEAL-HOA trial randomised lonely older adults to a volunteering intervention against a control, one of the few loneliness interventions tested this way
Isolation and loneliness carry real physiological risk, which is the reason scale matters at all	Well established — the American Heart Association’s 2022 scientific statement links isolation to a roughly 30% increased risk of heart attack, stroke, or death from either

The honest sentence for a funder proposal is something like: the evidence supports investing in structured, small-group connection work; it does not yet support a guarantee that this specific programme will produce the same result at higher volume, and the reason is that almost no study in this field has tested fidelity loss directly. Say that. Funders who have seen scale-ups fail before will trust you more for saying it than for promising outcomes you can’t back.

A staging approach that protects the mechanism

Rather than scaling to six sites simultaneously, stage it:

1. One new site, same facilitator model, deliberately slow ramp. Test whether the four load-bearing elements survive a single replication before multiplying the replication.
2. Measure fidelity weekly for the first two cohorts at the new site, not just at the end.
3. Only add sites two through six once the first replication’s fidelity measures match the original within a defined tolerance. If group size crept up or referral-to-session time doubled, fix that before adding volume, not after.
4. Budget for the unscalable parts explicitly — facilitator shadowing hours, smaller group caps, local relationship-building time — rather than assuming efficiencies will appear at higher volume. They usually don’t; they get demanded.

This is slower than a funder wants and cheaper than the alternative, which is discovering at site four that the thing being delivered no longer resembles the thing that was funded.

What this does not solve

None of this fixes the underlying evidence gap: the field still lacks controlled trials that isolate which components of small-group social connection programmes actually drive outcomes, as opposed to correlate with them. Staging your scale-up protects fidelity to a model whose active ingredients are still partly guessed at. It’s also worth being clear that a well-staged scale-up still reaches the people who could get to a session, self-referred or were referred by someone paying attention — it does nothing for the harder-to-reach cases that a bigger budget is usually, quietly, expected to solve as well.