Pathform AI Browse packs

Grants & Proposals

What Is a Grant Proposal Logic Model? Structure and Stages

Logic models are required by many funders and misunderstood by most applicants. The confusion is rarely about the boxes. It is about the difference between an output and an outcome, which decides whether a proposal reads as a plan for results or as a plan for effort.

Guide clusterGrants & Proposals

Direct answer

A logic model is a one-page chain that connects what a project uses to what it changes: inputs, activities, outputs, outcomes and impact. Inputs are resources, activities are what the team does, outputs are countable products, outcomes are changes in knowledge or behavior, and impact is the longer-term condition the work contributes to.

ClusterGrants & Proposals
Reading time6 min
FormatConcept walk-through
Sources5 listed

01The five columns and what belongs in each

A logic model is read left to right as a causal chain. Inputs cover the resources the project draws on: staff time, equipment, access to data or sites, and partner commitments. Activities describe what the team will actually do. Outputs are the countable products of those activities, such as samples analyzed, workshops delivered or software released. Outcomes describe changes that follow, and impact describes the longer-term condition the work contributes to.

The chain is deliberately simple, which is why it is so easy to fill in badly. Each column should contain only the kind of thing it names. When an output column starts listing benefits, or an impact column starts listing deliverables, the model has collapsed two stages and the reader loses the ability to judge whether the project is realistic.

ColumnTest questionExample entry
InputsWhat do we draw on?Two field seasons of instrument access and a named technician
ActivitiesWhat will we do?Calibrate sensors and run monthly surveys at four sites
OutputsWhat can be counted?A quality-controlled eighteen-month time series
OutcomesWhat changes?Managers use the series when setting seasonal plans
ImpactWhat improves over time?Reduced exposure of coastal infrastructure to erosion

02Outputs versus outcomes, the distinction that matters

An output is something your team produces and controls. An outcome is a change in someone else's knowledge, behavior or condition, and it depends partly on factors outside your control. That difference is not pedantic, because reviewers use it to judge whether a project overpromises. Claiming an outcome for something you merely produced is the most common way proposals inflate their own case.

A simple test is to ask who has to change for the statement to be true. If the answer is nobody outside the team, the statement is an output. If the answer is farmers, clinicians, teachers or policymakers adopting something, it is an outcome, and the proposal should show the mechanism that makes adoption plausible rather than assumed.

01

Write the statement you want to make about the project's effect.

02

Ask who outside the team must change behavior or knowledge for it to be true.

03

If nobody must change, move the statement into the outputs column.

04

If someone must change, name the mechanism that makes that change likely and check it is described in the plan.

Outputs are countable by your team; outcomes require someone else to change.

03Where assumptions and external factors belong

Every logic model rests on assumptions, and mature models state them. A model that assumes continued site access, stable staffing, or a partner's ongoing participation should say so, because a reviewer who spots an unstated assumption will discount the whole chain. External factors that shape outcomes, such as policy change or market conditions, also belong in the model rather than being quietly omitted.

There are two workable places to put them. Assumptions can sit in a narrow band beneath the chain, aligned with the stage they support, or they can be attached as short annotations to the specific arrow they affect. The band is easier to scan; the annotations are more precise. Either is better than a paragraph at the end that a reader may never reach.

Name any assumption that would change the design if it turned out to be false.
Attach assumptions to the stage they support rather than listing them separately.
Distinguish assumptions you can monitor from external factors you cannot influence.
If an assumption is load-bearing, show what happens if it fails.

04How a logic model differs from a research roadmap

The two diagrams overlap but answer different questions. A logic model is written for a funder, so it emphasizes resources, products and change, and it usually stays at the level of the program rather than the study. A research roadmap is written for a scientific reader, so it emphasizes method, validation and uncertainty, and it goes deeper into how a claim is established.

In practice, the roadmap often fits inside the activities and outputs columns of the logic model, which is why the two documents can contradict each other if they are drafted separately. Write the roadmap first, then derive the logic model from it, so the outputs you list are genuinely produced by the activities you described.

05Keeping the model honest over the life of the grant

A logic model submitted with a proposal should be revisited at reporting time, because it already defines the outputs and outcomes you will be asked to report against. Teams that treat it as a one-time submission document find themselves reconstructing evidence later, often from imperfect records. Teams that treat it as a monitoring plan collect data as they go.

The simplest practice is to keep the model next to the project plan and update the status of each output as work proceeds. When a stage slips or changes shape, the model shows immediately which downstream outcomes are affected, which is exactly the information a program officer wants in a progress report.

Worked example

Worked example: a logic model for a public health screening pilot

A university team proposes to test whether a community pharmacy can deliver a brief cardiovascular risk check to adults who rarely visit a physician. Read as a raw list of activities, the project looks busy but unfocused. Read as a logic model, the chain becomes a set of claims that can be examined one at a time.

The inputs column names pharmacy staff time, a training package developed by the team, a risk-scoring protocol, and a data-sharing agreement with two participating chains. The activities column describes the training, the recruitment of walk-in adults, the delivery of the check, and the referral of high-risk participants to a physician.

Outputs are countable: pharmacists trained, checks delivered, and referrals made with documented follow-up. Outcomes are changes: pharmacists report confidence in delivering the check, and a measurable share of referred participants attend a follow-up appointment. Impact is stated cautiously, as a contribution to earlier detection of cardiovascular risk in a group that is otherwise under-served, with an explicit note that the pilot is not powered to measure health events.

Two assumptions sit beneath the chain and are aligned to the stages they affect. The first is that walk-in adults will accept a check offered by a pharmacist, which the pilot tests directly. The second is that referral capacity in local practices will remain available, which the team cannot control and has therefore listed as an external factor with a stated contingency.

Common mistakes

Outputs written as outcomesAsk who must change for the statement to be true; if only your team is involved, it is an output.
Impact stated as a promiseDescribe impact as a contribution the project supports, and state honestly what the design can and cannot measure.
Assumptions left unstatedAttach each load-bearing assumption to the stage it supports, and mark which ones you can monitor.
The model never revisited after submissionKeep it beside the project plan and update each output's status so reporting is a matter of transcription rather than reconstruction.

Frequently asked questions

Is a logic model the same as a theory of change?

They are related but not identical. A logic model shows the sequence of inputs, activities, outputs and outcomes. A theory of change usually goes further and explains why each link is expected to hold, including the evidence or reasoning behind it. Some funders ask for both.

How detailed should the activities column be?

Detailed enough that a reader can tell the project is feasible, and general enough that it still fits on one page. Group related tasks into a single activity and let the narrative carry the protocol detail. Six to ten activities is a practical range for most projects.

Can I use AI to build a logic model draft?

You can use an assistant to sort content you have written into the five columns, which is useful when a draft has grown messy. It cannot judge whether your outcomes are plausible or whether an assumption is load-bearing, so check every entry against what your project can actually deliver.

What if the funder does not require a logic model?

Build one anyway as an internal check. The exercise usually exposes at least one activity with no output, or one outcome with no activity that plausibly produces it. Both findings are cheaper to fix before submission than after a review panel notices them.

Sources and further reading

  1. Grants and funding process, including application structureU.S. National Institutes of Health
  2. Grant application guidance and reporting expectationsBill & Melinda Gates Foundation
  3. Funding opportunities and proposal preparationU.S. National Science Foundation
  4. Recommendations for the Conduct, Reporting, Editing and Publication of Scholarly WorkInternational Committee of Medical Journal Editors
  5. ORCID - persistent identifiers for researchersORCID

These are the organisations whose published guidance the recommendations on this page follow. Listing them does not imply that they endorse Pathform AI, and this page is not affiliated with them.

Put this guide into practice

Get the editable layouts this guide describes, ready to retype and submit.

Browse template packs