How to Write a Logic Model for a Grant Proposal

What a Logic Model Is (and Why Reviewers Read It First)

A logic model is a one-page diagram that answers five questions in order: What are you investing? What are you doing? What will it produce? What will change? What is the long-term point? It is not the program description in picture form. It is a causal argument: this money plus this work produces these deliverables, which cause this change, which contributes to this larger difference.

Reviewers read it first because it is the fastest way to test whether a proposal thinks clearly. A narrative can hide a muddle in twenty pages of good writing. A logic model cannot hide anything. If the activities do not plausibly produce the outcomes, if the outcomes are really outputs, if the impact is disconnected from everything before it, the reviewer sees it in thirty seconds. Build the model before you finalize the narrative, not after. The narrative should be the model, written out in sentences.

The Five Components, With a Nonprofit Example

Every funder-ready logic model has the same five columns. Here is what each one holds, with a running example: a youth mentoring program applying for a $250,000 federal grant.

1. Inputs: the resources you invest. Funding, staff, volunteers, partners, facilities, curriculum, equipment. Be concrete and complete. In the example: the $250,000 grant, 12 trained volunteer mentors, a partnership with the school district, a 40-hour mentor training curriculum, 2 program coordinators.
2. Activities: what the program does. The work itself, stated as verbs. Weekly 1:1 mentoring sessions, monthly life-skills workshops, quarterly family engagement nights, attendance check-ins with school counselors. If an activity cannot be traced back to an input, you are missing an input. If an input does not feed any activity, you are over-resourced or under-planned.
3. Outputs: the countable deliverables. What the activities produce, in numbers. Train 24 volunteer mentors. Deliver 960 mentoring sessions per year. Enroll 120 youth. Host 12 workshops. Distribute 120 school-supply kits. Every output answers "how much of what." If a line has no number, it does not belong here.
4. Outcomes: the measurable change. What is different because the outputs happened. Increase school attendance from 82% to 92% among participants. Reduce disciplinary referrals by 40% within 12 months. Improve self-reported confidence scores by 25%. Every outcome answers "so what changed, and by how much."
5. Impact: the long-term difference. The systemic or generational change the program contributes to, stated in one or two lines. Break the cycle of youth disengagement in the county. Impact is aspirational by design; it is the only column allowed to be.

Read the chain left to right and then right to left. Left to right: "with these inputs we do these activities, which produce these outputs, which cause these outcomes, contributing to this impact." Right to left: "to get this impact we need these outcomes, which require these outputs, produced by these activities, funded by these inputs." If either reading breaks at a link, fix the link.

Outputs vs Outcomes: The Distinction Reviewers Test

This is where most logic models fail, so learn the rule cold: outputs count what you did; outcomes measure what changed. "Held 12 workshops" is an output. "15-point gain in participants' financial literacy scores" is an outcome. "Served 500 meals" is an output. "30% fewer households reporting missed meals" is an outcome.

Three quick tests for any line you are unsure about:

• The number test. If the line's number counts things your staff did (sessions, meals, kits, people enrolled), it is an output. If the number measures a difference in the people you serve (rates, scores, percentages changing), it is an outcome.
• The "so what" test. Read the line, then ask "so what?" If the answer is another fact about your program, you wrote an output. If the answer is a change in someone's life, you wrote an outcome. "We trained 24 mentors" (so what?) "who delivered 960 sessions" is still output. "so participants' attendance rose" is the outcome.
• The funder's-eye test. A reviewer scanning your outcomes column should see change language: increase, decrease, improve, reduce, from X to Y. If they see activity language (held, delivered, trained, distributed), they will conclude you do not know what you are measuring, and your evaluation plan loses credibility with it.

The most common mix-ups: putting "increased community awareness" under outputs (awareness is a change, not a deliverable), and putting "conducted 40 home visits" under outcomes (visits are activity). Both are instant credibility hits with experienced reviewers.

How to Build One, Step by Step

1. Start with the outcomes. Counterintuitive, but right. Decide what change you are accountable for first, because everything else is justified by it. Write each outcome as a change statement with a target: direction, amount, population, timeframe.
2. Work backward to outputs. For each outcome, ask what deliverables would plausibly cause it. Outcomes without outputs are wishes; outputs without outcomes are busywork. Every outcome should have at least one output that feeds it, and every output should feed an outcome.
3. Work back to activities and inputs. Name the work that produces each output, then the resources each activity needs. This is where budget gaps surface: an activity with no funding attached is a promise you cannot keep.
4. Write the impact last, in one or two lines. It should feel like the natural conclusion of the outcomes, not a new idea. If the impact surprises you given the outcomes, the outcomes are too small or the impact is too grand. Bring them into line.
5. Draft one SMART indicator per outcome. Not the outcome restated; the measure you will actually track, with a data source. "School attendance rate of enrolled participants, from district records, measured quarterly" is an indicator. "Increase attendance" is an outcome.
6. Stress-test the chain. Read it both directions (see above). Then hand it to someone outside the program and ask them to explain it back. Where they stumble is where your logic is weakest.
7. Align the narrative. The objectives, activities, and evaluation sections of your proposal must use the same language and numbers as the model. Reviewers cross-check. A model that says 40% and a narrative that says "significant improvement" reads as two different programs.

SMART Indicators: The Five Rules Applied

Every outcome needs at least one indicator that passes all five SMART tests. Here is what each rule actually demands, with a failing and passing example for a food-security program:

• Specific. Names exactly who and what. Fails: "Improve food security." Passes: "Reduce the share of enrolled households reporting very low food security."
• Measurable. Has a number a real data source can produce. Fails: "Families feel better about food." Passes: "from 34% to 20% on the USDA 6-item food security module."
• Achievable. Realistic given the timeframe and resources. Fails: "Eliminate all hunger in the county in 12 months." Passes: "a 14-point reduction over 24 months, in line with the pilot's 11-point result."
• Relevant. Measures the outcome, not something adjacent. If the outcome is food security, the indicator cannot be "number of cooking classes held" (that is an output). It must be a food-security measure.
• Time-bound. Has a deadline or measurement period. Fails: "eventually." Passes: "measured at baseline, 12 months, and 24 months; target reached by month 24."

Put together: "Reduce the share of enrolled households reporting very low food security from 34% to 20% on the USDA 6-item module, measured at baseline, 12, and 24 months, target by month 24." One sentence, five rules, zero ambiguity about what success looks like. For ten worked before-and-after examples across real nonprofit programs, see SMART objectives examples for nonprofit grants.

Where It Goes in the Proposal

The model itself usually goes in the appendix as a one-page figure, referenced by name in the program description ("see Logic Model, Appendix B") and again in the evaluation plan. Some applications ask for it in the narrative; follow the funder's instructions over this guide. Either way, the model's language and numbers must match the narrative exactly. The most embarrassing logic model mistake is not a bad model; it is a good model that contradicts its own proposal.

The evaluation plan is where the model pays for itself. Each indicator becomes a row in your evaluation table: indicator, data source, collection schedule, target, responsible party. Reviewers who see the same outcomes in the model, the objectives, and the evaluation plan conclude the program is well designed. That conclusion is half the scoring rubric.

Mistakes That Sink Proposals

• Outputs in the outcomes column. The most common and most damaging. "Conducted 40 home visits" is not an outcome no matter how important the visits were. Reviewers read miscategorized columns as "this team cannot evaluate itself."
• Indicators with no data source. An indicator you cannot actually measure is decoration. Every indicator needs a named source: district records, the USDA module, intake surveys, case management data. If you cannot name the source, change the indicator.
• Impact as a wish list. "End poverty in America" as the impact of a 50-person job training program is not ambition, it is a broken chain. Impact should be the believable long-term consequence of your actual outcomes, scaled honestly.
• More than one page. A logic model is a summary diagram. If it does not fit on one page, the program is described too broadly or the outcomes are really objectives in disguise. Cut until it fits.
• Numbers that disagree with the narrative. The model says 40%, the objectives say "significant improvement," the budget serves 200 people the model says 500. Reviewers cross-check. One set of numbers, everywhere.
• Building it after the narrative. A model drawn to justify an already-written proposal is backwards and it shows: the chain reads like rationalization. Build the model first, then write the narrative from it.
• No baseline. "Increase X by 20%" means nothing without a starting point. Baselines come from prior program data, community statistics, or a baseline data collection period stated in the evaluation plan. No baseline, no credible target.

Frequently Asked Questions

What is a logic model in a grant proposal?

A one-page map of how your program works: inputs (resources), activities (what you do), outputs (countable deliverables), outcomes (measurable change), and impact (long-term difference). Reviewers read it before the narrative to check whether the program's logic holds together.

What is the difference between outputs and outcomes?

Outputs are countable deliverables: 200 meals served, 12 workshops held, 50 volunteers trained. Outcomes are the measurable changes those deliverables cause: 30% fewer missed meals, 15-point gains in food-security scores, 40% higher volunteer retention. Outputs count what you did; outcomes measure what changed.

Do funders require a logic model?

Federal funders (USDA, HHS, Department of Education) almost always require one, and most large foundations expect one or reward applications that include one. Even when it is not required, a tight logic model strengthens the narrative because it forces your objectives, activities, and evaluation plan to line up.

What is a SMART indicator?

A measure that is Specific (names who and what), Measurable (has a number), Achievable (realistic), Relevant (tied to the outcome), and Time-bound (has a deadline). "Increase program enrollment" is not SMART. "Enroll 120 youth ages 12-17 by December 2027" is.

How long should a logic model be?

One page. A logic model is a summary diagram, not a narrative. If it does not fit on one page, the program is described too broadly. The proposal narrative carries the detail; the model carries the structure.

Where does a logic model go in a grant proposal?

Usually in the appendix as a one-page figure, referenced from the program description and the evaluation plan. Some applications put it in the narrative itself when the funder explicitly asks. Either way, the objectives and evaluation sections must use the same language and numbers as the model.