Counting Who Showed Up Is Not the Same as Counting Who Got Better

A link worker refers 40 people to a walking group over a quarter. Thirty-two attend at least once. Of those, 24 complete a pre/post wellbeing measure showing improvement. The end-of-quarter report says: “24 people improved.” That sentence is true and it is also the wrong headline. It tells a funder about the 24 people who engaged, completed the paperwork, and got better. It says nothing about the other 60% of the original referral list, and nothing at all about the people in the practice population who were never referred.

Two different questions are being asked here, and most programme reporting answers only one of them.

Question one: for the people who reached us, did anything change? That is an outcome question. Question two: who did we actually reach, relative to who needed this? That is a reach question. They require different data, they fail for different reasons, and a programme can score well on one while quietly failing the other.

Why the two numbers get collapsed

Outcome data is easier to produce than reach data, and reach data is more embarrassing. A pre/post wellbeing score on your attenders is a clean number you already have from delivery. Establishing your denominator — who was eligible, who was referred, who dropped before ever attending, and who was never referred at all because they don’t come to the GP or don’t speak the language your leaflet is printed in — takes work that isn’t part of running the group. So programmes report what they measured, not what they need to know, and the report reads as more successful than the programme is.

The systematic review evidence supports the outcome half cautiously: social prescribing studies do report gains in self-esteem and self-confidence among participants, but the evidence base is thin and heterogeneous, and a qualitative synthesis found the perceived benefit depends on participants experiencing structured, purposeful activity rather than contact alone. None of that literature can tell a commissioner what share of the eligible population those gains apply to, because the studies were built around people who already attended.

What reach actually requires you to track

Reach is not attendance. Attendance is a number you get once someone has already engaged. Reach is the funnel before that:

Stage	What to record	Why it matters
Eligible population	Estimated count with the presenting need (isolation, low mood, post-diagnosis adjustment)	Your denominator. Without it, every later number is unanchored.
Identified	How many of the eligible population were flagged by a referrer at all	Tests whether your referral routes see the people who need it, or only the people already in the system
Referred	How many identified people were actually offered a referral	Gap here usually means referrer time, not client refusal
Attended once	How many referred people showed up at all	The number most reports lead with — and the weakest one
Sustained engagement	How many attended at a threshold you define in advance (three sessions, six weeks)	Where outcome measurement should start

Most programme reports start at row four. The Campaign to End Loneliness’s long push for consistent loneliness measurement exists precisely because without a shared denominator, no two programmes’ reach numbers mean the same thing, and cross-programme comparison — which is what funders actually want when they ask “does this work” — becomes impossible.

The failure mode: the volunteer-shaped funnel

Call it the volunteer-shaped funnel. It happens when a programme’s outcome data looks excellent and its reach data, if anyone bothered to collect it, would look poor — because the people who complete outcome measures, attend consistently, and report improvement are disproportionately the people who were already socially resourced enough to self-refer, fill in a form, and turn up on a wet Tuesday. The AARP Foundation’s national survey of adults 45 and older found that predictors of loneliness cluster around network size and physical isolation — precisely the people least likely to make it through a multi-step referral pathway with paperwork attached. A programme optimising only for the outcome number will, without noticing, optimise for recruiting people who needed it least.

This is not a hypothetical failure. It is the default shape of voluntary uptake in almost every community programme, and it is why outcome-only reporting systematically overstates programme value to exactly the population the funding was meant to serve.

What this means in practice: report reach and outcome as two separate lines, never as one blended success rate. State your denominator explicitly, even when it is an estimate. If you cannot produce a reach number this quarter, say so in the report rather than defaulting to an attendance figure dressed up as reach — a funder who later asks “reach compared to what” and gets no answer will trust the next report less, not more.

What to do with the two numbers once you have them

They point to different fixes.

A weak outcome number, strong reach number, means people are getting into the programme but not getting anything from it. Look at session content, facilitation quality, and whether the group actually gives people a role rather than passive attendance — the qualitative evidence on social prescribing consistently finds that structured, purposeful engagement outperforms unstructured contact.

A strong outcome number, weak reach number, means the programme works for the people it reaches but isn’t reaching enough of the right people. Look at referral routes, not delivery. This usually means fixing who refers, not what happens after referral. It also means being honest with a funder that the 24-person success story sits on top of an 80% drop-off between eligible and engaged, because a renewal decision made on the outcome number alone will fund more of the same narrow reach.

A weak number on both axes is a programme decision, not a measurement decision — the National Academies’ consensus report on isolation in older adults calls for routine assessment precisely because without it, health systems cannot tell which of these two failures they are looking at, and end up cutting programmes that reach well but measure badly, or renewing programmes that measure well but reach badly.

What this does not solve

Reach data corrects for one kind of bias, not all of them. Even a well-constructed reach funnel still depends on someone identifying eligible people in the first place, and the routes into most social prescribing schemes run through primary care contact, which itself is unevenly distributed. The UK’s 2018 loneliness strategy embedded loneliness measurement into national statistics partly because programme-level reach data, however carefully built, cannot see the people who never come near a referrer at all — the isolated person who does not visit a GP, is not connected to a community organisation, and therefore never enters any funnel this framework can measure. Fixing that is a population-level surveillance and outreach problem, not a reporting fix, and no programme evaluation, however honest about its denominator, closes that gap on its own.