Why Reach Numbers and Outcome Numbers Answer Different Questions

A social prescribing service reports 400 referrals last quarter, 350 completed a first session, and average loneliness scores dropped two points on a validated scale for those who finished the programme. That looks like success. It answers almost none of the question a commissioner should actually be asking, which is: who in the eligible population did this reach, and who did it miss?

Those are two different measurement problems, and most programme reporting collapses them into one.

Two questions, not one

Outcome measurement asks: for the people who went through the programme, did they get better? Did loneliness scores fall, did GP contacts drop, did wellbeing improve? This is what most evaluation frameworks default to, because it’s what funders ask for and what validated instruments are built to capture.

Reach measurement asks: of everyone who could plausibly benefit, what share did the programme actually touch, and does that share look like the eligible population or a narrow slice of it? This is almost never reported with the same rigor, partly because it’s harder — it requires a denominator, and most services don’t know their denominator.

A programme can have excellent outcomes and terrible reach. It can also have modest outcomes and strong reach. Reporting only one number lets a service look effective while serving a self-selected fraction of the people it exists for.

The self-selection problem, not disguised as anything else

Social prescribing evidence is instructive here precisely because it is thin on the reach side. A 2021 systematic review of social prescribing and wellbeing found consistent gains in self-esteem and self-confidence among participants, but noted heterogeneous programme design and limited trial evidence. A separate systematic review focused specifically on loneliness found all nine included studies reported positive individual impacts, with three showing reduced use of GP, emergency, social worker, or inpatient services.

Read those findings straight and social prescribing looks like it works. Read them as a measurement scientist would, and the more interesting fact is what’s absent: none of that evidence tells you who didn’t take up the referral, who dropped out before a first session, or who was never referred in the first place because a clinician didn’t think to ask. The people captured in these outcome studies are, by construction, people who engaged. That’s not a flaw in the studies — it’s what outcome measurement is designed to do. It’s a flaw in treating outcome measurement as if it answered the reach question too.

A qualitative meta-synthesis on social prescribing found that participants describe benefits extending beyond social contact — restored meaningful participation, a sense of purpose — and that structured, purposeful activity appears to work better than contact alone. Useful for programme design. Still silent on reach.

Naming the failure mode: the good-outcomes trap

Call this the good-outcomes trap: a programme reports strong results for its participants and uses that as evidence it is working well overall, when in fact its participants are the easiest 15% of the eligible population to serve — the ones with transport, flexibility, trust in institutions, and enough social confidence to show up to a group activity in the first place.

The National Academies’ 2020 consensus report on older adults is a useful anchor here, because it puts a hard number on the population at risk: roughly a quarter of adults 65 and older are considered socially isolated. If a programme aimed at older adults is serving a caseload where fewer than a quarter show markers of isolation on intake, that’s a reach signal worth investigating regardless of how good the outcome scores look. The companion clinical commentary on that report makes a related point for the health system context: routine assessment of isolation requires systems to actually ask, at scale, not just to measure well among those who arrive.

The AARP Foundation’s 2018 national survey of adults 45 and older adds a specific, actionable predictor: 33% of people who have spoken to a neighbour in the past week report loneliness, against 61% of those who never have. That gap between socially embedded and socially isolated respondents is exactly the gap that outcome-only reporting tends to erase, because the people never spoken to a neighbour are also the people least likely to walk into a group session unprompted.

What to actually track

You do not need a research team to track reach. You need three things most services don’t currently record.

1. A defined eligible population, even a rough one — GP list size, a ward’s over-65 population, an employer’s headcount — so referrals have a denominator, not just a numerator.
2. Referral source and route, broken down by who referred (self, clinician, family, other agency) and by demographic markers relevant to your population — deprivation index, ethnicity, living-alone status, first language.
3. Drop-off points, recorded separately: referred but not contacted, contacted but declined, attended once and never returned. Each of these is a different problem requiring a different fix, and lumping them into a single “non-completion” rate hides which lever to pull.

None of this requires a validated instrument. It requires a spreadsheet column that currently doesn’t exist in most services.

What this means in practice: Report reach and outcomes as two separate lines to your funder, not one blended narrative. If you can only measure one this quarter, measure reach — you can improve outcomes for people you’re already serving through programme tweaks, but you cannot know if you’re serving the right people at all without a denominator. A funder who accepts outcome data alone as proof of population-level impact is being sold half a measurement.

The evidence-status table

Claim	Evidence status
Social prescribing improves self-reported wellbeing for participants who complete it	Reasonably well supported across multiple systematic reviews
Social prescribing reduces downstream service use (GP, A&E, inpatient)	Suggestive — reported in some included studies, not consistently measured
Structured, purposeful activity outperforms unstructured social contact	Supported by qualitative synthesis, not by controlled comparison
Social prescribing reaches a representative slice of the eligible isolated population	Largely unmeasured — this is the actual gap

The UK’s 2018 national loneliness strategy was notable for building loneliness measurement into the Office for National Statistics rather than leaving it to individual programme evaluations — an implicit acknowledgment that population-level reach data has to be built at a level above any single service, because no single service can generate its own denominator credibly.

What this does not solve

Tracking reach tells you who you’re missing. It does not tell you why, and it does not fix the structural reasons — transport, trust, language, shift patterns, disability access — that keep the same groups underrepresented year after year. A reach dashboard is a diagnostic, not a remedy. And reach measurement itself has a bias: it’s easiest to build for populations already known to services — people with a GP, an employer, a case file. The people least reached by any formal system are also the hardest to count as missing from one. Improving your denominator is real progress. It is not the same as reaching everyone the denominator can’t see.