Logic Model for Grants: How to Build One and Use It While You Write
A logic model for grants connects inputs, activities, outputs, outcomes, and impact on one page. Here is how to build one backwards and use each column as you draft.
Last updated: September 2026
Most grant writers meet this diagram under pressure. A funder's checklist asks for one, the deadline is close, and the instructions never say what it is for. Built early, a logic model for grants tells you whether the proposal holds together before you have written a paragraph of narrative. It shows which promises your budget can actually pay for. Built last, it usually exposes a gap you no longer have time to close.
Quick Answer: What Is a Logic Model for Grants?
A logic model for grants is a one-page diagram that connects the resources a program uses to the results it promises. It has five columns: inputs, activities, outputs, outcomes, and impact. Funders read it as a consistency check, since the budget, the workplan, and the promised results should all describe the same project.
Free download: Logic model builder workbook as Excel .xlsx. No email address required.
What Is a Logic Model?
A logic model has five columns, and each column answers one question about the program. The columns run left to right, and the answers have to agree with each other.
| Column | The question it answers | What belongs in it |
|---|---|---|
| Inputs | What does the program need to run? | Staff and volunteer time, space, materials, partner capacity, existing curriculum, the grant funds |
| Activities | What does the program do with those resources? | Tutoring sessions, home visits, a training series, case management, equipment installation |
| Outputs | What do those activities produce, counted? | Sessions delivered, people served, workshops held, units distributed |
| Outcomes | What changes for participants in the short and medium term? | Skills gained, behavior changed, conditions improved, status advanced |
| Impact | What long-term condition does the work contribute to? | Community-level change the program influences but cannot claim alone |
An output is a count of what the program produced. An outcome is a change in the people or conditions the program touched. Confusing the two turns the outcomes column into a second count of the same work, which is the failure mode this format exists to prevent.
The five-column structure comes out of program evaluation practice; no single funder owns it. The W.K. Kellogg Foundation's *Logic Model Development Guide*, updated January 2004, treats the model as a planning tool first and a reporting artifact second. The framing matters for grant writing, because the model does most of its work during planning. Our guide to how to write a grant proposal puts that planning ahead of drafting, and the logic model is where the planning becomes visible on a page.
Logic Model for Grants: What Reviewers Cross-Check
A reviewer reads a logic model against the rest of the application and looks for places where the documents disagree. The diagram is short enough to hold in view while checking three or four other pages, which is exactly what makes it useful to a reviewer and risky for an applicant.
The budget is the first cross-check. If the model lists a full-time family navigator and the grant proposal budget funds a quarter-time coordinator, the reviewer trusts the budget and discounts the model.
Then the workplan. Every activity in the model should appear on the timeline with a start date and an owner. An activity that appears only in the diagram is aspiration.
The evaluation plan gets read against the outcomes column. Each outcome needs a measure, a data source, and a collection schedule. That is the whole job of the evaluation plan, and the outcomes column supplies its raw material.
Last, the statement of need against the impact column. If your need statement documents a shortage of bilingual caseworkers, the impact column should point at that shortage. General community wellbeing is a different claim, and the two do not substitute for each other.
When these four disagree, the model works against the application. A reviewer who spots one contradiction has a reason to go looking for others.
How to Build a Logic Model for Grants
Build a logic model backwards. Start with the change you are promising and work leftward to the resources you actually have, because that direction surfaces the gaps while there is still time to fix them.
1. Name the impact. Write the long-term condition the project contributes to, in plain language, in one sentence. Keep it honest about scale. A two-year, one-site program contributes to a community condition; it does not eliminate it.
2. Write the outcomes. For each outcome, answer four questions: what changes, for whom, by when, and by how much. Split them into short-term (knowledge, skills, attitude) and medium-term (behavior, practice, status). This column and the goals and objectives section of your application should say the same thing in different formats.
3. Set the outputs. For every outcome, name the countable evidence that the work happened. Outputs are attendance, completion, delivery, and distribution figures. They prove effort, not effect.
4. List the activities. Include only the activities that produce the outputs you just listed. An activity with no output attached is either missing a measure or does not belong in this proposal.
5. Price the inputs. Walk each activity back to the staff time, space, materials, and partner capacity it requires. This step catches an overreaching workplan, because the inputs column has to match the budget line by line.
6. Add assumptions and external factors. Underneath the columns, write what has to be true for the chain to hold, and what could break it. Assumptions might include that partner schools will release student data. External factors might include a funding cut at a referral agency. Naming them signals that you have thought past the best case.
7. Read it back as a sentence. Left to right, the model should read: if we have these resources, we can run these activities, which produce this much service, which leads to these changes, which contribute to this condition. If any link needs a paragraph of explanation to make sense, the chain has a hole.
Steps two through five usually need a working session with program staff. A solo drafting hour will not surface the same gaps, because the people who have to deliver the program are the ones who can tell you which links are thin. Expect the session to change at least one outcome, and expect at least one activity to fall out of the model entirely once someone asks who is running it on a Tuesday afternoon.
Logic Model Grant Writing: Turning the Columns Into Narrative
Each column of a finished logic model feeds a specific section of the application, so the model doubles as your drafting outline.
| Logic model column | Proposal section it feeds | What carries across |
|---|---|---|
| Impact | Statement of need | The condition you are naming and the scale you claim |
| Outcomes | Goals and objectives | The measurable change, the population, and the timeframe |
| Outcomes plus indicators | Evaluation plan | Measures, data sources, collection schedule, targets |
| Activities | Project description and workplan | The service model, sequence, and staffing |
| Outputs | Reporting commitments | The numbers you will report at each interim deadline |
| Inputs | Budget and budget narrative | Personnel, space, materials, and the justification for each line |
Write the outcome text once and reuse it. The objectives section and the evaluation plan then carry identical wording, which is harder to achieve than it sounds when three people draft three sections in the same week. The grant proposal walkthrough covers the drafting order for the rest of the sections.
Funders differ on whether the model belongs in the narrative page count or in the appendix, and some ask for a narrative description with no diagram at all. Read the instructions before you design the page. Our logic model template covers the layout options if you want a blank grid to start from.
When the Funder's Template Does Not Use Five Columns
Funder templates rename the columns and sometimes split them, but the chain underneath stays the same. Map the headings on the form onto the five standard columns before you type anything into it, and the model you already built transfers without rework.
Nothing standardizes the labels, so read the column definitions in the instructions instead of trusting the column names.
| What the template calls it | Where it belongs in the five columns | What to watch |
|---|---|---|
| Resources | Inputs | Same content, different label |
| Strategies or interventions | Activities | Keep participant-facing work only |
| Participation, reach, or people served | Outputs | Headcounts belong here, not beside the outcomes |
| Short-term, intermediate, and long-term outcomes as three columns | Outcomes, split by time | The long-term column usually holds what the five-column version calls impact |
| Goal | Impact, or a single top-line outcome | Ask the program officer which one they mean |
| Objectives | Outcomes | Each one still needs a number and a date |
| Indicators or measures | Attached to each outcome | The wording has to match the evaluation plan |
| Rationale or assumptions | The notes under the columns | Write them even when the form has no box for them |
Two variants cause most of the trouble. The first is a template with three outcome columns, which invites padding, and a medium-term outcome that restates the short-term one in different words is visible to anyone reading down the page. If the program has two real outcomes, leave the third box empty and say why in a line of text. The second is a template that opens with a problem statement box at the far left. That text is your statement of need in compressed form, so lift it from the narrative rather than writing it fresh, because a reviewer will read the two versions side by side.
Fill in the funder's form, then keep your own five-column version as the working copy. The form is an export. The model is what you edit when the program changes.
A Logic Model Example You Can Adapt
A grant logic model for a $25,000 after-school tutoring request fits on one page, with one row for each chain that runs from an input through to an impact. The version here is a composite illustration, not a real organization: a small literacy nonprofit serving one elementary school.
| Inputs | Activities | Outputs | Outcomes | Impact |
|---|---|---|---|---|
| Part-time program coordinator, 12 volunteer tutors, donated classroom space two afternoons a week, existing phonics curriculum, $25,000 grant | Recruit and train volunteers, run 90-minute tutoring sessions twice weekly, administer reading assessments in fall and spring, hold two family literacy nights | 30 students enrolled, 48 tutoring sessions delivered, 12 volunteers trained, 60 assessments completed | Short term: students finish the program with improved decoding skills. Medium term: participating students advance a reading level by the spring assessment | More third graders in the school reach grade level in reading |
Every output in this literacy-tutoring model is a number the coordinator can produce from a sign-in sheet, and every outcome has a named measure attached. Swap the population and the assessment and the same structure holds for a health clinic or a workforce program.
Notice what the impact column does not say. It does not claim the program will close the district's reading gap. A tutoring project reaching 30 students contributes to a district goal. Claiming ownership of that goal invites the reviewer to check the arithmetic, and the arithmetic will not support it.
What the First Draft of This Example Got Wrong
Building backwards catches problems that a finished table hides. Three things changed between the first pass at this composite example and the version above.
An outcome had no instrument. The first version promised that students would gain confidence in reading. Nobody could say where that number would come from at report time, and confidence appears on none of the assessments the program already gives. It became a decoding-skills result measured by the fall and spring assessments, which the activities column was paying for anyway.
An activity lost its place. The first list included monthly check-ins with classroom teachers. The idea sounded useful and produced nothing countable. Asked who would schedule the meetings and what record would come out of them, the coordinator answered nobody and none, so the activity came out of the model.
The inputs column forced a cut. The first draft assumed a full-time coordinator. At a $25,000 request, with volunteer tutors and donated space absorbing most of the delivery cost, that line pays for part-time hours and nothing more. The program dropped from three tutoring afternoons a week to two, and the output count moved with it. Outputs have to survive arithmetic: two sessions a week across a 24-week term is 48, which is the figure in the table. A session count carried over from the larger schedule would have described a program the budget does not fund.
Two of those inputs also belong on the assumptions line, because the organization controls neither one. The donated classroom depends on a principal who can reassign the room, and twelve volunteers depend on a recruitment season that has come up short before. Write both underneath the columns with the fallback in a sentence, so a reviewer can see that the plan survives losing the room.
Common Mistakes in a Logic Model for Grants
A frequent error is an outputs column dressed up as outcomes. Counting what you delivered does not show what changed, and a reviewer reading the two columns side by side ends up seeing the same information twice.
| Written as an outcome | Actually an output | Rewrite it as |
|---|---|---|
| 200 people attended workshops | Attendance count | Target: 70 percent of attendees pass the certification exam |
| We distributed 500 food boxes | Distribution count | Target: participating households report fewer days of skipped meals |
| Staff completed 40 hours of training | Training count | Target: case managers apply the new screening protocol at intake |
Four more errors show up repeatedly.
The activities column turns into a to-do list. Grant writing, hiring, and reporting are project management tasks, not program activities. The activities column holds what you do for participants.
The impact claim outruns the award. Scale the impact statement to the size and length of the grant. A one-year, single-site project that claims a county-wide condition change is promising more than the money buys.
The model runs to too many rows. It is a one-page format, and a dozen outcome chains crowded onto a page stop being readable at print size. Fewer chains, each one traced all the way through the five columns, does more for the application than a full page of half-finished ones.
You build the model after the narrative is finished. Reverse-engineering a diagram from finished text reproduces whatever gaps the narrative already has. Build it during planning and edit both together.
Using the Model After You Submit
A logic model keeps working after submission because the outputs and outcomes columns are the reporting schedule for the award. Interim reports ask for the numbers already listed in the outputs column and the measures already named beside each outcome.
Teams that handle this well set up the collection dates at award time, months before the first report is due. In GrantCue you can put each collection date on the calendar as a task and assign it to whoever owns the number, so the reporting view is already populated when the deadline arrives.
The harder case is a target you are going to miss. Report the variance rather than quietly restating the target as though it had always been lower. Put the original number, the actual number, and the reason in one paragraph, then say what changes for the next period. Edit the model itself when an activity changes and date the edit, so the final report starts from a document that matches what the program did.
The model also gives you a head start on the renewal. A funder deciding whether to continue support wants the same chain, with the outcomes column showing what actually happened.
FAQ
What is a logic model?
A logic model is a one-page diagram that connects the resources a program uses to the results it promises. It has five columns: inputs, activities, outputs, outcomes, and impact. Grant applications usually ask for it as a one-page attachment, and program evaluators use the same format outside of fundraising.
What is the difference between a logic model and a theory of change?
A logic model shows what a program does and what results from it. A theory of change explains why those steps should produce that result, including the assumptions and conditions behind each link. A theory of change is usually longer and written in prose. An application that asks for a logic model often covers the reasoning in the project description.
What is the difference between outputs and outcomes?
Outputs count what the program delivered, such as sessions held or people served. Outcomes describe what changed for those people, such as skills gained or behavior adopted. A number that could be pulled from a sign-in sheet is an output. A finding that requires a before-and-after measure is an outcome.
Do grant applications require a logic model?
Some funders ask for one by name in the application instructions, often as a one-page appendix. Others fold the same information into the project description or evaluation section without using the term. Read the notice of funding opportunity before deciding, and build the model regardless, because it tests the proposal whether or not you attach it.
Where does the logic model go in a grant proposal?
Placement varies by funder. Many put it in the attachments or appendix outside the narrative page limit; some count it against the limit. When the instructions are silent, attach it as a labeled one-page appendix and reference it in the project description so the reviewer knows to look.
How long should a logic model be?
One page. The format exists to make a program legible at a glance, and a model that runs to three pages has stopped doing that. If the chains do not fit, the project has too many outcome claims for the size of the request.
Next Step
The same five columns work for a federal notice and for a family foundation, so build the model before you have picked a funder. Then search open grant opportunities for funders whose priorities match the outcomes you can defend. When you are ready to draft, the guide to writing a full grant proposal covers the sections each column feeds.