Hoverview

logframe · results frameworks

The difference between an output and an outcome, settled

The question is asked in every logframe workshop. The answer has been in the source text of the method since it was written.

In every logframe training, somebody asks: is this an output or an outcome? The facilitator gives a rule. The rule is usually something about control, or about what the programme delivers versus what changes because of what it delivers, or about direct results versus indirect ones. The trainees write the rule down. Then the first real submission comes back from the donor with a note saying the outcome should be an output or the output should be an outcome, and the rule stops working.

The reason the rule stops working is that the question is misconceived. "Is this an output or an outcome?" treats the level as a property of the statement, as if a sentence carries its altitude with it. The source text of the method says the opposite. A level is the position a statement occupies in a particular frame. The same sentence can be an output under one donor and an outcome under another, not because one donor is wrong, but because the two donors have placed different frames over the same programme, and the frame determines the level.

That was settled in 1979, in the document that introduced the method. It has been relitigated in every training programme since, because the answer is inconvenient: it means there is no universal test for distinguishing an output from an outcome, which is the one thing a trainee most wants to leave the room with.

What the original text says

The logframe was designed by Practical Concepts Incorporated for USAID. The founding document, published in November 1979, is Rosenberg and Posner's "The Logical Framework: A Manager's Guide to a Scientific Approach to Design and Evaluation." PCI no longer exists and no agency hosts an official copy, but a third-party mirror survives.

On page 37, the document describes what happens when a programme has both a programme-level logframe and a project-level logframe underneath it. Each of the projects needed to achieve the purpose would be an output on the programme logframe, and that output would become the purpose on the project manager's own logframe.

The implication is precise. A result that is an output in one frame is a purpose in another. The result did not change. The frame changed. The level is not a property of the result. It is a property of the relationship between the result and the frame placed over it.

This is not an ambiguity in the text. It is the design. The logframe was built for a programme architecture where multiple levels of management each had their own frame, and the cascade between levels was the mechanism by which a programme's outputs became a project's purpose. The method assumes from the start that the same result will sit at different levels in different frames, because that is how hierarchical management works.

What Gasper confirmed

Des Gasper, writing at ISS The Hague in 2000, reviewed the method's history and its subsequent interpretation in "Logical Frameworks": Problems and Potentials. His observation is the one that should have ended the debate: most practitioners continue to act as if distinctions between levels are contextual, not inherent.

The word "contextual" does the work. A statement is not inherently an output or inherently an outcome. It is an output in the context of one frame and an outcome in the context of another. The context is the frame, and the frame is determined by the donor, the funding instrument, or the management level at which the logframe is drawn.

Gasper also identifies the deeper problem with vertical logic. The method conflates sufficient and necessary conditions. A bare link between two levels cannot say whether the lower level is sufficient for the higher one, merely necessary for it, or one of several contributing causes. That conflation is part of why the output-outcome boundary feels unstable: the relationship between an output and an outcome can be sufficiency ("if we deliver this, the outcome follows"), necessity ("unless we deliver this, the outcome cannot follow"), or contribution ("this makes the outcome more likely, alongside other things"), and the logframe cannot distinguish between them.

What the donor templates prove

If levels were inherent properties of statements, every donor would place the same statement at the same level. They do not.

A vocabulary analysis across 19 donor documents in the source corpus counted nine distinct terms for the top row of a logframe and ten for the second row. "Outcome" names the top row in one source and the second row in another. The terms include Impact, Goal, Overall Objective, Long-term impact, Principal Objective, and Development Objective at the top, and Purpose, Outcome, Specific Objective, and Intermediate Result at the second level, among others. No two donors use identical terminology, and several use the same word to mean different things.

The UK guidance from 2011 places the outcome at the second level, below an impact statement. DG ECHO's Single Form Guidelines (2017) add a Principal Objective level above the Specific Objective, making the Specific Objective the third level rather than the second. The Asian Development Bank's DMF guidelines (2024) allow one outcome statement and up to three impact statements, with no indicators at the impact level. The World Bank IEG guidance (2012) starts at intermediate results and ends at the Project Development Objective, with no activities row and no impact row.

Consider a conservation programme that trains community rangers. Under the FCDO frame, "120 community rangers trained and deployed" is an output: something the programme delivered. Under a bilateral grant where the project-level logframe sits below a programme-level one, the same statement could be the purpose of the training project, because the programme frame treats it as an output and the project frame treats it as the top-level result the project manager is accountable for. The rangers are the same. The training is the same. The level changed because the frame changed.

This is exactly the cascade Rosenberg and Posner described. It is not an edge case. It is the normal operation of multi-level programme management, and it is the reason the output-outcome question cannot be answered without first asking: in which frame?

The tests that fail

Three common tests for distinguishing outputs from outcomes are taught in training programmes. Each one works for a single donor and breaks across donors.

The control test says: an output is within the programme's control; an outcome is not. This works for the FCDO frame, where the outcome is the change the programme contributes to and the outputs are the deliverables the programme controls. It does not work for the ADB, which places the outcome at a level the programme is accountable for and excludes impact-level indicators entirely because the ADB does not consider project-level impact measurable through a project's own monitoring. Under the ADB frame, the outcome is within control, which contradicts the control test.

The deliverable test says: an output is a thing you deliver; an outcome is a change that results from it. This works until you encounter a programme whose "outcome" is a deliverable at a higher level. GFDRR's banded framework (2017) has eighteen outcomes and no indicator columns, and some of those outcomes are things a programme delivers (a risk assessment completed, a plan adopted). The deliverable test fails because deliverability is not a property of a level.

The verb test says: outputs use delivery verbs (trained, distributed, constructed); outcomes use change verbs (increased, improved, reduced). This is a stylistic convention, not a structural rule. Defra's Biodiversity Challenge Funds require outcome statements to begin with a present-tense description of the changed state, while DFID's examples include both forms at both levels. A statement phrased as "improved maternal health outcomes in target districts" could sit at impact, outcome, or output level depending on the frame, because the verb describes the change, not the altitude.

Each of these tests is correct within a single frame and incorrect as a universal rule. Trainees learn one test, apply it to a different donor's template, and discover it does not hold.

The two tests that work

Two tests survive across frames, and they work because they ask about the frame rather than about the statement.

The first is the accountability test. At which level does this donor hold the programme accountable for performance? The level below accountability is the output level: the things the programme must deliver. The accountability level is the outcome level: the result the programme must plausibly contribute to. Anything above accountability is the impact or goal level: the long-term change the programme is positioned within but not held responsible for.

This test works because it is frame-relative. The accountability boundary sits in a different place for a bilateral donor than for a development bank, and the output-outcome split moves with it. Under FCDO, the programme is accountable for contributing to the outcome but not for achieving the impact. Under ADB, the programme is accountable for achieving the outcome directly. The statement at the accountability level is the outcome in both cases, even though the scope of what "outcome" means differs.

The second is the position test. Where does this statement sit in this donor's hierarchy? If it is in the second row, it is at the level this donor calls the outcome, regardless of what other donors call the same statement. This test is trivially correct and entirely practical. It says: stop asking what the statement inherently is and start asking where the template puts it.

The position test is unsatisfying because it is circular: an outcome is whatever sits at the outcome level. But the circularity is the point. The method was designed so that levels are positional, and a test that acknowledges that is more useful than one that pretends levels are inherent and fails when the frame changes.

Why the question keeps coming back

If the answer was settled in 1979 and confirmed in 2000, why does every training programme reteach it as an open question?

Three reasons, all structural.

The first is that the answer is hard to teach. "It depends on the frame" is a correct answer and an unsatisfying one. A facilitator who says "an output is what you deliver and an outcome is what changes" gives the room a rule they can apply immediately. A facilitator who says "an output is whatever sits at the output level in your donor's template" gives the room a tautology, even though the tautology is correct and the rule is not. Training materials optimise for actionability, and the inherent-level model is more actionable than the contextual-level model, so the inherent-level model is what gets taught.

The second is that most programmes report to one donor. For a team that only ever fills in FCDO logframes, the control test works every time, because the FCDO frame places the accountability boundary in a consistent place. The test fails only when the team encounters a different frame, and many teams never do. The problem is invisible to anyone operating within a single frame, which is most people.

The third is that the contextual answer has an uncomfortable implication for programme design. If a level is not inherent, then the programme does not have "outcomes" in any absolute sense. It has results, and different donors assign those results to different levels. A programme that believes its outcome is inherently an outcome will resist when a donor template places it at the output level, because it feels like a demotion. A programme that understands levels as positional will accept the reassignment as a rendering decision, not a judgement about the significance of the work.

That last point is worth pausing on. A great deal of friction between programmes and donors around logframe structure comes from the belief that level assignment is a valuation: that being at the outcome level means the result is more important than being at the output level. The original method contains no such implication. A level is a management layer, not a ranking. But the belief is widespread, and it makes the contextual answer emotionally harder to accept than the inherent one.

What "settled" means here

The title of this post claims the question is settled. That claim rests on two sources: the method's own founding document and its most cited academic critique. Both say the same thing. Levels are contextual. The same result occupies different levels in different frames. The output-outcome distinction is a property of the frame, not of the statement.

What the post does not claim is that the sector has accepted this. It has not. Training materials, guidance notes, and working definitions continue to treat levels as inherent properties, and the control test, the deliverable test, and the verb test continue to be taught as universal rules. The question is settled in the literature and unsettled in practice, which means the cost of the misconception is paid every time a team redesigns its programme logic to satisfy a template rather than re-rendering the same logic under a different frame.

Whether any practical mechanism exists that would make the contextual model usable for a programme team filling in a logframe under deadline, when the facilitator taught the inherent model and the donor's reviewer has their own implicit model, is the question this post does not attempt to answer.

Sources

The method's own text

Academic critique

  • "Logical Frameworks": Problems and Potentials, Des Gasper, Institute of Social Studies, The Hague (2000). Source for levels being contextual rather than inherent, and for the conflation of sufficient and necessary conditions in vertical logic

Donor frameworks cited