Topic 19 of 64· Social value and impact measurement

Logic Model

A logic model is a linear diagram connecting inputs, activities, outputs, outcomes, and impact for a programme, read left to right as an accountability chain: resources go in, activities happen, outputs get produced, outcomes change for beneficiaries, and impact accrues at a wider or longer timescale. It is the standard structure funders and auditors expect a programme to be reportable against, and the forward-facing counterpart to a backward-mapped theory of change.

Why it matters

HM Treasury's Magenta Book specifies the logic model as a required element of programme evaluation design, and funders such as the National Lottery Community Fund build their application and reporting templates around exactly this five-column chain. Its value is that it forces a programme to state, in one diagram, what it will spend, what it will do with it, what it will produce, and — critically — what should change as a result, at a level of specificity that a paragraph of prose tends to obscure. A logic model with a populated inputs and activities column but an empty or vague outcomes column is diagnosable at a glance, which is precisely why funders ask for one.

The maths

The logic model is a structural chain rather than a formula:

Inputs          Activities        Outputs             Outcomes              Impact
(resources      (what is done     (direct, countable   (change for           (long-term,
 committed)      with them)        products)            beneficiaries)        population-level
                                                                               or systemic change)

Each column should be more specific than the last: inputs are what you spend, activities are what you do, outputs are what gets delivered regardless of effect, outcomes are what changes as a result — the distinction covered in full in outcomes vs outputs — and impact is the durable, often attributable-only- in-part, long-run change.

Worked example

Local authority (digital debt advice service):

  • Inputs: £180,000 annual budget, 4.0 FTE advisers, a case management system.
  • Activities: outreach sessions, one-to-one debt advice appointments.
  • Outputs: 900 appointments delivered; 750 debt and benefit plans issued.
  • Outcomes: of clients reaching a 6-month follow-up, 60% (450 of 750) report reduced arrears, averaging a £1,200 reduction per client — £540,000 in aggregate arrears reduction.
  • Impact: a measurable fall in homelessness applications from the service's client base over two years, only partly attributable to this service alongside other interventions (see counterfactual analysis).

Charity (food bank referral partnership):

  • Inputs: £45,000, 1.5 FTE coordinator, partnership agreements with 12 referral agencies.
  • Activities: referral triage, parcel packing and distribution.
  • Outputs: 5,000 food parcels distributed to 1,100 households.
  • Outcomes: 68% of surveyed households (748 of 1,100) report improved food security at a 4-week follow-up call.
  • Impact: contribution to reduced local crisis-service demand, evidenced only in aggregate area statistics, not attributable to this charity alone.

Software engineering connection

The logic model is close to a literal data model for an outcomes system: inputs and activities are operational data you already hold (spend, staffing, session logs); outputs are easy to instrument because they are counted at the point of delivery; outcomes require deliberately designed follow-up data collection (surveys, admin data linkage) that will not exist unless someone builds it; impact usually requires linked, longitudinal, or population-level data beyond any one programme's systems. Engineers building reporting tools should push commissioners to define outcome and impact indicators at design time, rather than defaulting to an outputs-only dashboard because that is what the transactional data already supports. See social return on investment for a method that values the outcomes and impact columns specifically, and benefits realisation for tracking whether the impact column was actually delivered.

Pitfalls

  • Stopping at outputs. A dashboard that reports appointments delivered or parcels distributed and implies benefit is reporting activity, not results — see outcomes vs outputs.
  • No stated causal link between columns. A logic model states the chain but not why activities should produce outputs which should produce outcomes; that reasoning belongs in a theory of change, and a logic model without one behind it is untested.
  • Treating it as a one-off bid document. Logic models produced only to satisfy a funding application and never updated stop reflecting what the programme actually does.
  • Attribution creep at the impact column. Claiming population-level change as caused solely by one programme, without a counterfactual, overstates what the evidence supports.

Sources