Designing Groups That Survive Week Four

Recruitment is the part of a connection programme that gets planned. It is rarely the part that fails. Groups fill, the first session runs, the second is well attended, and then somewhere between the third and the fifth session the numbers fall to a level from which they do not recover.

This note is about that window, and about the design decisions that determine what happens in it.

Why week four is the failure point

The first two sessions run on novelty and on the commitment someone made when they signed up. By the third or fourth, both have been spent. What has to carry the group from that point is whether anyone in it has formed a specific reason to come back — a person they expect to see, a role they hold, or an unfinished thing they are part of.

If the programme has not manufactured at least one of those by session three, attendance becomes a matter of whether the session is more appealing than the alternative on a given evening. It usually is not.

What this means in practice

Treat sessions one to three as a distinct phase with its own objective. The objective is not enjoyment or content delivery. It is that every attendee leaves session three able to name at least one other attendee and holding one small commitment for session four.

The three structures that hold a group

The qualitative evidence on social prescribing is consistent on one point: where programmes reduce loneliness, the mechanism runs beyond increased social contact. A 2022 meta-synthesis found that participants described benefit in terms of restored meaningful participation and purpose rather than contact volume. Contact alone was not what people valued, and it is not what makes them return.

Three structures reliably produce that sense of participation.

Named roles. Somebody sets out chairs. Somebody makes the tea. Somebody messages people who missed a week. A role is a reason to be missed if you are absent, which is a materially different thing from being welcome if you attend.

Unfinished work. A group that completes something each week has no thread pulling it forward. A group building, growing, rehearsing, or planning something has one. The content matters less than whether it spans sessions.

Fixed, unremarkable logistics. Same day, same time, same room, indefinitely. Every variation is a decision the attendee has to make again, and each decision is an opportunity to not come.

What the evidence does and does not support

It is worth being precise about the state of this literature, because guidance in this field is often written with more confidence than the underlying studies carry.

Claim	Evidence status
Group-based activity shows more promise than one-to-one contact for isolation	Supported across systematic reviews, though effect sizes are inconsistent
Social prescribing improves self-esteem and confidence	Reported consistently, mostly in small or uncontrolled studies
Social prescribing reduces service use	Reported in a minority of studies; not established
Any specific session structure outperforms another	Not established; no comparative trials

The 2021 systematic review of social prescribing and loneliness found all nine included studies reporting positive individual impacts, and three reporting reductions in GP, emergency, social worker or inpatient use. That is encouraging and it is also a very small evidence base with heterogeneous designs. A parallel review the same year found gains in self-esteem and self-confidence while noting limited trial evidence throughout.

The honest position is that the direction of effect is reasonably well supported and the magnitude is not. Design accordingly: assume your programme probably helps, and do not promise a funder a figure the literature cannot support.

Measuring the right window

Most programme evaluation measures outcomes at completion, which systematically misses the failure this note is about. A programme that loses 60% of participants by week four and reports strong outcomes among those who finished has measured the people it did not fail.

Three additions to a standard evaluation:

1. Session-level attendance by individual, not aggregate headcount. Aggregate numbers hide replacement — new joiners masking the departure of early ones.
2. A week-four retention rate reported as a headline figure alongside outcomes.
3. A reason recorded for every drop-out, collected by someone who is not the facilitator, within two weeks of the last attendance.

The third is the one that gets skipped and the one that carries the most information. People who stop attending rarely announce a reason, and the reasons they give a facilitator differ from the ones they give anybody else.

Where this fits with the wider picture

None of this addresses who is not in the room. The prevalence data is a reminder that programme design reaches a self-selected group: the National Academies concluded in 2020 that roughly a quarter of adults over 65 are socially isolated, and Harvard’s 2021 survey put serious loneliness at 36% of Americans and 61% of young adults. The people at highest risk are, by definition, the least likely to sign up for a group.

Retention work makes a programme worth running for the people who arrive. It does not solve reach, and it should not be reported as though it does.