The Week Four Problem: Why Attendance Collapses and What to Do About It

Programme managers tend to describe attrition as a slope. It isn’t. In most recurring group programmes, attendance holds roughly steady for the first two or three sessions and then drops sharply at a specific point, usually the fourth session, before settling into a smaller, stable core. Call it the week four problem. If you are only tracking overall attendance as a percentage across a twelve-week run, you will miss this entirely — the average looks like gradual decline, but the underlying pattern is a cliff followed by a plateau.

This matters because the two patterns call for completely different fixes. Gradual decline is a dosage problem: people are drifting away steadily, so you’d look at content quality, scheduling friction, or whether the activity itself has staying power. A cliff at a specific week is a threshold problem: something specific happens around that point that a chunk of the group cannot get past. Fixing the wrong problem — polishing content when the real issue is a threshold — wastes a redesign cycle.

Why week four, specifically

There is no single trial that isolates “week four” as a universal constant, and this note is not claiming one exists. What the evidence does support is the mechanism, and the mechanism is consistent enough across social prescribing research to take seriously.

Qualitative work on social prescribing has found that participants describe the value of programmes as extending well beyond social contact itself — restored purpose, a sense of meaningful participation, a reason to structure the week. That value doesn’t arrive in session one. A systematic review of social prescribing’s impact on loneliness found positive outcomes across nine studies, but the mechanism in most of them depends on sustained contact, not a single encounter. The first two or three sessions are spent on orientation: learning names, working out where you sit, deciding whether this is a room you want to be in. The benefit — the thing that would actually motivate someone to keep coming — typically hasn’t landed yet. Meanwhile the initial motivation that got someone through the door (a referral, curiosity, obligation) has usually faded by then. There’s a gap between the old motivation running out and the new motivation kicking in, and that gap tends to fall around the third or fourth session for weekly programmes. People who don’t feel any pull yet, and whose original reason for showing up has expired, don’t come back.

This lines up with something broader in the loneliness literature. Cacioppo’s account of loneliness describes it as a state that makes people simultaneously want connection and defensively wary of it — hypervigilant to social threat, quicker to read rejection into ambiguous signals. A new group is exactly the setting where that wariness gets tested repeatedly in the first few weeks. If nothing has happened by week three or four to override the wariness — no moment of being remembered, included, or missed when absent — the defensive read wins.

What actually happens at the cliff

Three things tend to be true of programmes that lose people at this point, based on how social prescribing evaluations describe both successful and unsuccessful models:

1. No one has been individually missed yet. Structured, purposeful activity outperforms unstructured social contact in participant accounts of what helped — but structure alone doesn’t create the feeling of being tracked as a person. If a participant misses session three and nobody follows up, the programme has told them, correctly, that their presence wasn’t specifically noticed.
2. The activity has not yet produced a reason to return beyond the activity itself. Early sessions are usually still explaining the format. If the format is still being explained by week four, there’s been no room for the actual payoff — a shared task, a running joke, a piece of unfinished business — that gives people a specific reason to come back next week rather than a general one.
3. The group has not yet sorted into any kind of sub-structure. Larger, unstructured groups rarely produce a felt sense of belonging in the first few sessions. Putnam’s distinction between bonding and bridging social capital is relevant here: bridging contact (being in the same room as strangers) is what most first sessions offer, and it is weakly motivating on its own. Bonding — a subset of people who start to recognise each other specifically — takes a few repeated sessions to form, and if it hasn’t started forming by the cliff point, there’s nothing pulling anyone back.

What to build instead

What this means in practice: don’t treat weeks one to four as identical sessions with identical content. Treat them as a distinct onboarding phase with one explicit job — get each participant a reason, specific to them, to come back — and build a deliberate intervention at the point just before the historical drop-off, rather than reacting to it after the numbers are already down.

Concretely:

• Track attendance by session number, not by week-on-week percentage. You need to see the shape of the curve, not just the average. If you don’t already have this broken out, get it before you change anything else — you may not have a week four problem at all; some programmes cliff at session two (a mismatch problem) or session eight (a fatigue problem), and the fix is different in each case.
• Build a check-in for any missed session in the first month. A message or a call from a named person, not an automated one, after the first absence. This is the single most concrete lever available and it is cheap. The point is not the content of the message — it’s that someone noticed.
• Give the first three sessions a visible throughline, something incomplete at the end of session two that pulls people back for session three: a shared activity with a next step, not three self-contained sessions that could be attended in any order.
• Deliberately seed smaller sub-groupings early, rather than letting them form or fail to form on their own. Pairs or small clusters within a larger group, assigned in the first or second session, give bonding capital a head start rather than leaving it to chance over four weeks.
• Do not schedule the biggest content investment for week one. Programmes routinely front-load their best material into the first session to make a strong first impression, which is backwards: the risk isn’t week one attendance, it’s week four attendance. Save something worth returning for.

Evidence status

Claim	Evidence status
Recurring group programmes commonly show a sharp early drop rather than gradual decline	Consistent with programme-design experience and mechanism-level evidence; no source in this literature isolates a universal “week four” constant
Structured, purposeful activity produces more perceived benefit than unstructured contact	Reasonably well supported — qualitative meta-synthesis of social prescribing
Individual follow-up after a missed session improves retention	Plausible and cheap, but not directly tested in the sources reviewed here — treat as a low-cost hypothesis worth piloting, not a proven fix
Bonding social capital takes several sessions to form and is what sustains attendance	Supported at the concept level by Putnam; not measured session-by-session in this evidence base
Loneliness produces defensive wariness that a new group must specifically override	Supported by Cacioppo’s account of loneliness as a threat-detection state

What this does not solve

None of this addresses the people who never show up to session one. Social prescribing evaluations, and the loneliness literature more broadly, are near-unanimous that programme design of any kind reaches the people who already accepted a referral or walked through a door — a self-selected group. The AARP Foundation’s national survey found that loneliness correlates strongly with physical isolation and small, undiverse networks; the people most isolated by that definition are also the least likely to enrol in a twelve-week group programme in the first place. Fixing the week four cliff will improve outcomes for people already in your programme. It will not touch the larger population who never attempt the on-ramp at all, and no attendance-curve fix changes that.