The Logframe Doesn't Lie: How to Check Your Intervention Logic Before an Evaluator Does

4 August 2026

Intervention logic is the causal argument at the core of your proposal: the claim that evidenced needs, addressed through specific objectives and activities, will produce outputs that lead to outcomes and, over time, impact. A logframe is that argument laid out as a grid, which means it can be checked mechanically, cell by cell, before anyone with a scoring sheet does it for you. Three checks catch most of the damage: the traceability test, the thirty-second indicator test, and an honest read of the assumptions column. What follows are those checks, one worked example, and a half-hour self-check protocol per work package.

The vocabulary evaluators expect you to get right

Terminology errors cost credibility before the logic is even examined, because the five core terms look interchangeable and are not.

  • Output is what the project itself produces and controls: a curriculum, a platform, a delivered training course, a report. If your team can put it on a shelf or a server, it is an output.
  • Outcome is the change in the target group that outputs enable: teachers apply the method, members adopt the tool. The project cannot fully control outcomes, only make them likely.
  • Result is the loosest of the five. Some programmes use it as an umbrella for outputs and outcomes; the Erasmus+ Programme Guide has its own conventions. Choose one meaning, state it once, and hold it steady.
  • Impact is the longer-term, wider change your outcomes contribute to, beyond your direct participants, at sector, system, or policy level. Projects contribute to impact; they do not deliver it.
  • Indicator is a measurement, not a goal. A usable one names what is measured, in whom or what, by when, and how the data will be collected.

Mixing these is more than a style problem. When an objective is written as an output ("develop a platform") or an outcome as an activity ("conduct workshops"), the causal chain has a hole, and holes are what evaluators are trained to find.

How an evaluator reads a logframe

Evaluators rarely read the grid top to bottom. A typical read picks one objective and traces downward: which activities serve it, which outputs they produce, which indicators will show the intended change. Then it reverses direction from a random activity: which objective does this serve, and which evidenced need sits behind it? The evaluator also cross-reads the needs analysis and the budget, and inconsistencies between the three are where scores drop. Prose can blur a weak link with good writing; a grid cannot.

The traceability test

Run the same two traces yourself, systematically. Downward: every need leads to an objective, every objective to at least one activity, every activity to an output, every output to an indicator. Upward: every indicator measures something an output produces, every activity serves a named objective, every objective answers an evidenced need.

Anything that fails a trace is an orphan, and every orphan is a finding:

  • An activity with no objective above it is usually a partner's wish item that survived the negotiation. Cut it or connect it.
  • An objective with no activities under it is an unfunded promise; evaluators price these instantly.
  • A need described in the analysis but addressed by nothing tells the reader the needs chapter was written for atmosphere.
  • An output with no indicator will be treated as unverifiable, which for scoring purposes means optional.

Do this on the printed document, pen in hand.

The thirty-second test for indicators

For each indicator, ask one question: could the project hit this indicator while the situation of the target group stays exactly the same? If yes, it measures effort, not change. "Number of workshops held" and "platform launched" pass while nothing changes for anyone. They are legitimate at output level, where you prove delivery, and damaging at objective and outcome level, where the reader wants change: the share of participants who apply what they learned, or a measured shift in competence against a baseline.

The repair is mechanical. Take the effort indicator, ask "so that what?", and measure the answer: workshops held so that teachers use a new method becomes measured use, by follow-up survey or observation, at a stated point. Every objective carries at least one change indicator; effort indicators stay next to outputs.

Assumptions and risks: filled last, read first

The assumptions column is where most teams paste generic text at midnight before the deadline, and it is one of the first places an experienced evaluator looks, because it shows whether the authors understand their own causal mechanism. Every arrow in the chain depends on something outside your control, and this column proves you know what.

Apply two tests to each assumption. First, is it specific to a causal step? "Partners will cooperate effectively" is a template line; "school directors allocate teaching time for the pilot" is an assumption someone thought about. Second, is it genuinely external? If the condition is within the project's control, it is not an assumption but a task: move it into the activities and the mitigation plan. A well-filled column lists the exact points where the project could fail for reasons you cannot fix, each paired with the mitigation you can.

A worked example: one weak chain, repaired

The example is generic and illustrative.

Before:

  • Need: "Youth unemployment in the region is high."
  • Objective: "Empower young people through digital skills trainings."
  • Activity: "Deliver trainings and a summer school."
  • Output: "Trainings delivered."
  • Outcome: "Young people are empowered."
  • Indicator: "Number of young people trained."
  • Assumption: "Participants will be motivated."

Every cell fails a check. The need is neither evidenced nor specific. The objective restates the activity with the word empower added. "Empowered" is not observable, so the outcome can never be verified. The indicator is pure effort. The assumption is a hope.

After:

  • Need: employers in the partner regions report unfilled entry-level digital roles, while young jobseekers lack the competences those roles require (with sources in the needs analysis).
  • Objective: increase the employability of unemployed young adults in the partner regions for entry-level digital roles.
  • Activities: co-design the curriculum with local employers, deliver the training cycle, place participants in short internships.
  • Outputs: an employer-validated curriculum, a trained cohort, completed internships.
  • Outcome: participants demonstrate the competences employers named and move into work, further training, or extended placements.
  • Indicators: share of participants reaching the defined competence level; share in employment or further training at a stated follow-up point, collected by survey.
  • Assumption: employers stay engaged through co-design and internship hosting, mitigated by commitments signed at application stage.

Same project, same budget, same partners; now every cell traces in both directions, the indicators measure change, and the assumption names the real external dependency.

The thirty-minute self-check protocol, per work package

Run this once per work package, starting with the one you trust least.

  • Step 1, five minutes: down-trace. Follow each objective the work package serves to activities, outputs, and indicators. Mark every orphan.
  • Step 2, five minutes: up-trace. For each activity, write one sentence naming the objective it serves and the need behind it. Any activity that resists the sentence is a finding.
  • Step 3, five minutes: run the thirty-second test on every indicator, labelling each E for effort or C for change. Each objective needs at least one C.
  • Step 4, five minutes: rewrite the assumptions so each is specific to a causal step and genuinely external. Move anything controllable into activities or mitigation.
  • Step 5, five minutes: cross-check against the budget table. An activity funded but absent from the logic, or argued but unfunded, is the inconsistency evaluators are paid to notice.
  • Step 6, five minutes: fix what you can on the spot and log what you cannot, with an owner and a date before submission.

If the protocol keeps surfacing the same class of problem across work packages, the issue is structural and worth an external pass; an intervention-logic check on a live draft is one of the things we do for coordinators. Weak intervention logic is only one of the recurring reasons EU applications get rejected, but it is the one you can fix mechanically before you submit.

Where this comes from

StrandsUnited checks intervention logic on live drafts for EU consortia, alongside budget cross-checks and dissemination infrastructure that outlives the funding. The patterns above come from that work and from our grant intelligence engine, a consortium graph built on CORDIS and Erasmus+ open data covering 114,731 organisations and 83,490 projects across the 2014-2027 programming periods.