Who Holds a Workplace Group Together: A Role Audit for Peer Groups and ERGs

Most evaluations of workplace connection groups — peer support circles, employee resource groups, informal lunch clubs — ask whether the group met, and whether people liked it. That tells you almost nothing about whether it will still exist in a year. The variable that predicts survival is usually structural, not experiential: who is doing what inside the group, and what happens when that person is out sick, promoted, or gone.

This is an evaluation problem before it is a design problem. If you are commissioning or running workplace groups and reporting on them to HR, a wellbeing budget holder, or an external funder, you need a way to measure role structure, not just attendance and satisfaction. This guide sets out what to track and how to present it.

Why roles, not just attendance, are the right unit of measurement

Cigna’s 2020 workplace survey found that 61% of U.S. adults report loneliness sometimes or always, and that lonely workers miss work roughly twice as often for illness and five times as often for stress-related absence. That is the case for running something. It is not evidence about what to run.

The more useful evidence comes from social prescribing research, where structured, purposeful group activity has repeatedly outperformed unstructured social contact on the outcomes people actually report caring about — a 2022 qualitative meta-synthesis found that participants describe benefit extending beyond contact itself to restored participation and purpose, and this depended on the activity having a shape: a task, a rotation, a reason to show up beyond showing up. A 2021 systematic review of social prescribing found consistent gains in self-esteem and confidence, but was explicit that trial evidence remains thin and heterogeneous. Read together, the signal is: structure helps, but do not expect a clean effect size.

Workplace isolation and loneliness are also not the same construct. A 2024 study of healthcare workers found that workplace isolation — feeling disconnected from colleagues and the organisation — has different correlates than general loneliness, and that perceived social support changes how remote or hybrid work affects wellbeing. This matters for evaluation design: a group can reduce workplace isolation (people feel embedded in a team) without touching someone’s loneliness outside work, and vice versa. Decide which one you are actually trying to move before you pick outcome measures.

The role audit: five positions to check for

Before measuring outcomes, audit the group’s structure. Robert Putnam’s distinction between bonding social capital (ties within a group) and bridging social capital (ties across groups) is useful here: most workplace peer groups are built for bonding, and bonding groups are especially prone to collapsing when their internal structure is thin.

Score each group against these five roles. A role can be filled informally, but it has to be filled by someone identifiable, not by “whoever feels like it.”

1. Convenor — sets the date, sends the invite, decides the group happens this month. This is the role that collapses fastest under workload pressure.
2. Continuity holder — remembers who came last time, notices who has dropped off, and follows up. Distinct from the convenor; a group can have a convenor with no continuity holder and lose members without anyone noticing.
3. Sponsor — a manager or senior figure who gives the group legitimacy to exist on work time. Without this, attendance is the first thing cut when workload spikes.
4. Content or agenda owner — decides what the group actually does when it meets. Groups that exist purely as “catch up” tend to decay faster than groups with a recurring task, topic, or rotation.
5. Successor — someone identified, in advance, to take over the convenor role. Almost no workplace group has this, which is precisely why so many end when one person leaves.

Score each role as present and named, present but undefined, or absent. A group with three or more roles absent is at high risk regardless of how positive last month’s feedback was.

What to measure, and on what schedule

What to track	How	When
Role coverage score (0–5)	The audit above, self-reported by the convenor	At launch, then every quarter
Attendance decay	Attendance at session 1, 4, 8, 12 as a percentage of session 1	Every session, reviewed quarterly
Convenor turnover	Number of convenor changes and whether a successor was in place at the time	Continuous, logged as an event
Workplace isolation	A short validated isolation measure, not a bespoke satisfaction question	Baseline and 3 months
General loneliness (optional, sensitive)	UCLA-type item only if you have consent processes for health-adjacent data	Baseline and 6 months, opt-in

The attendance decay curve is the single most informative number here, and most programmes do not compute it. A group that starts at twelve people and holds nine by session eight is structurally sound. A group that starts at twelve and holds four is telling you something about roles, not about interest.

Evidence-status table

Claim	Evidence status
Lonely and isolated employees cost more in absence	Supported — Cigna 2020 survey data, self-reported
Structured group activity outperforms unstructured contact on participant-reported benefit	Supported, moderate confidence — qualitative meta-synthesis, not a trial
Named roles (convenor, successor, sponsor) predict group survival	Plausible, not directly tested — inferred from social capital theory and programme attrition patterns, not a controlled comparison
Workplace connection groups reduce loneliness outside work	Weak — most evidence measures workplace isolation, a related but distinct construct
A defined successor role prevents group collapse	Not tested — a design recommendation, not a finding

Be honest with a funder about which row you are standing on. The claim most programmes present (“this group reduces loneliness”) is the weakest-supported one on the table. The claim you can actually defend is closer to “this group reduces workplace isolation for people who attend, and group survival depends heavily on role structure we can now measure.”

What this means in practice: Before you fund or renew a workplace connection group, ask for its role audit score, not its satisfaction score. A group with a happy founder and no successor is a group you are about to lose. Fund the transition plan, not just the sessions.

The convenor cliff

Name the failure mode so people recognise it when it is happening. The convenor cliff is what occurs when a group’s entire operating structure lives in one person’s calendar and goodwill. It looks fine for months — attendance is steady, feedback is warm — and then the convenor changes teams, goes on leave, or simply gets too busy, and the group does not decline gradually. It stops. There is rarely a final session; there is just no next one.

The fix tested nowhere formally but implied by the pattern: identify a successor before you need one, and treat the sponsor role as separate from the convenor role, so legitimacy does not disappear with the person.

The American Enterprise Institute’s 2021 survey on friendship found that 12% of American adults report no close friends at all, up from 3% in 1990 — a broader erosion of the informal structures that used to make workplace bonding easier to sustain without deliberate design. That is the environment workplace groups now operate in: less ambient social infrastructure to fall back on, which makes deliberate role structure more load-bearing than it used to be, not less.

What this does not solve

This is a structural diagnostic, not a loneliness intervention. It will tell you whether a group is built to survive, not whether the people in it were lonely to begin with or whether the group reaches anyone outside its existing membership. Workplace connection groups are, almost by definition, opt-in — they reach people who already chose to show up to a meeting about connection, which is a self-selecting slice of any workforce. The employees most likely to be isolated at work are often the ones least likely to join a group designed to address it. No amount of role clarity fixes that; it is a separate, harder design problem.