Topic 20 of 64· Social value and impact measurement
Outcomes vs Outputs
An output is the direct, countable product of an activity — it exists the moment delivery happens, regardless of what effect it has. An outcome is the change that follows for the people, place, or system involved. "500 people attended a job-search workshop" is an output: it is true even if none of them find work. "500 people's employment prospects improved" is an outcome claim, and it requires evidence of change, not just evidence of attendance — the confusion that produces more misleading grant reports than almost any other measurement mistake in the sector.
Why it matters
HM Treasury's Magenta Book and funders such as the National Lottery Community Fund both require outcome reporting specifically because outputs are what programs report by default: they are cheap to count, always available, and always look positive. An output count can literally never go down as a result of the program failing — more sessions delivered is always "more", whereas an outcome can reveal a program is not working. The National Audit Office has repeatedly criticized government programs for reporting activity levels as if they were evidence of success; a software system that only makes outputs easy to report reinforces this by default, because outputs require no follow-up data collection and outcomes do.
The math
There is no formula, but there is a reliable test to classify a metric:
Output test: is it countable at the point of delivery, true even if the recipient is unaffected?
Outcome test: does it require a before/after or with/without comparison to be meaningful?
If a number can be true with zero benefit to anyone, it is an output.
This sits inside the wider logic model chain and depends on the outcome links defined in a theory of change; turning an outcome into money uses the methods in social return on investment.
Worked example
Local authority (employment support): output — 500 people attended job-search workshops. Outcome — at 12-month follow-up, 140 of those 500 (28%) are in sustained employment (6+ months). A comparison group with similar characteristics but no program access has a 15% baseline employment rate over the same period. Net outcome uplift: 28% − 15% = 13 percentage points, so an estimated 500 × 0.13 = 65 additional people are in work who would not otherwise have been — the attributable outcome, distinct from either the 500 attendance figure or the 140 raw employment count.
Charity (literacy charity): output — 1,200 reading sessions delivered to 300 children. Outcome — average reading age improved by 8 months over a 6-month period, against an expected 6-month natural progression baseline of 6 months. Net outcome gain: 8 − 6 = 2 months' additional reading age improvement per child attributable to the program, not the full 8-month figure.
Software engineering connection
Event logs and transactional systems instrument outputs almost automatically — page views, sessions, tickets closed, appointments booked — because they are generated by the system doing its job. Outcomes require a data model that captures the same individual at a later point in time against a baseline or comparison, which has to be designed in deliberately: follow-up surveys, linked administrative records, or a comparison cohort. A reporting tool that only supports the former will quietly steer an organization toward output-only reporting no matter what the funder asked for. See logic model for where outcomes sit in the accountability chain, cost-per-outcome for turning this distinction into a unit-cost metric, and public sector KPIs for the wider pattern of metric selection.
Pitfalls
- Reporting outputs as if they were outcomes. "500 people attended" implies benefit without demonstrating it; label attendance as an output explicitly.
- No baseline or comparison group. An outcome figure with no counterfactual — see counterfactual analysis — cannot separate program effect from what would have happened anyway.
- Optimizing for the funded metric. When funding is tied to output volume, delivery teams rationally maximize attendance over durable change, a Goodhart's law failure mode.
- Outcome washing. Relabeling an output metric with outcome-sounding language ("engagement outcomes: 500 attendees") without any follow-up measurement behind it.
Sources
- HM Treasury, Magenta Book (2020). https://www.gov.uk/government/publications/the-magenta-book
- National Lottery Community Fund, outcomes reporting guidance. https://www.tnlcommunityfund.org.uk/
- National Audit Office, value-for-money report methodology. https://www.nao.org.uk/