NSF Logic Models: Where They Go in the Project Description and What They Drive
GrantCopilot Team
November 1, 2026
12 min read
TL;DR
The PAPPG does not require a logic model; individual solicitations decide. As of November 2026, S-STEM requires a logic model or theory of change, ATE makes one optional, and EPSCoR RII-FEC accepts one as a way to lay out its plan. Unless your solicitation says otherwise (S-STEM wants it as a Supplementary Document), it goes inside the Project Description, where visual materials count toward the 15-page limit. Then use it as your outline: the goals section walks through the model, the timetable covers every activity, the evaluation measures the short-term outcomes, and the long-term outcomes tie back to your rationale and Broader Impacts.
In most NSF proposals that include one, a logic model is not a separate document. It is a figure inside the Project Description, it counts toward the Project Description's page limit, and the rest of the Project Description should be built around it. Used well, the model becomes the outline of the proposal: the goals section walks through it, the timetable schedules every activity in it, and the evaluation plan measures its short-term outcomes. This guide covers which NSF solicitations ask for a logic model, where it goes, what it costs you in pages, and how to make it drive the goals, objectives, activities, and timetable of a standard Project Description. Writing for a different funder? See our nonprofit logic model guide and NIH logic model guide.
Does NSF Require a Logic Model?
Not as a general rule. The phrase "logic model" does not appear in the PAPPG, NSF's Proposal and Award Policies and Procedures Guide. NSF 24-1 is still the version in effect; NSF deferred the replacement it had planned for 2026. Requirements come from individual program solicitations instead.
Even where no solicitation asks for one, NSF's merit review criteria ask whether the plan for the proposed activities is well-reasoned, well-organized, and based on a sound rationale, and whether it includes a way to assess success. A logic model answers both questions on one page.
Three current solicitations show the range. Solicitations change between competitions, so check the current one for your program, and ask the program officer when it is unclear.
- S-STEM (NSF 25-514) requires a logic model or theory of change that informs the project's evaluation plan, and asks for it as a Supplementary Document.
- Advanced Technological Education (NSF 24-584) makes a logic model optional in the evaluation plan and says proposals are not penalized for leaving it out. It does require a timeline for the evaluation activities.
- EPSCoR RII Focused EPSCoR Collaborations (NSF 24-573) asks for a timetable or strategic plan for reaching the project's goals, a logic model with a clearly stated theory of change, or both.
Where It Goes: Inside the Project Description
The PAPPG limits the Project Description to 15 pages unless a solicitation sets a different limit, and Results from Prior NSF Support, when you have it, counts within those 15 pages (up to five of them). It also says that visual materials, including charts, graphs, maps, photographs, and other pictorial presentations, are included in the 15-page limit. A logic model is a visual, so it counts.
There is a second reason it belongs there. The PAPPG says special information and supplementary documentation must be included as part of the Project Description, except for the specific documents it lists separately. A logic model is not on that list, so unless your solicitation asks for it elsewhere, as S-STEM does, moving it to Supplementary Documents to save space is not an option the PAPPG gives you. S-STEM, for its part, warns that supplementary documents beyond the ones it specifies may be returned without review.
Place the figure where it helps the reader, such as next to the goals discussion or at the end of the Project Description, give it a caption, and refer to it by that caption where the goals section begins.
Budget the Page Before You Write
A readable logic model with five or six columns usually needs most of a page, and it often reads best on a landscape page. Either way, that page comes out of the same 15 you have for the motivating rationale, methods, management plan, evaluation, and Broader Impacts. Plan for it in your page budget at the outline stage, not the night before submission when the Project Description runs a page long.
Two habits keep the cost down without making the figure unreadable:
- Write entries as short phrases, not sentences. The narrative explains each activity; the model only has to name it.
- Do not repeat the model in prose. The goals section should walk through the model and add what the figure cannot show (the reasoning and the literature behind each link), not restate every box.
A Theory of Change Format That Fits NSF Proposals
NSF does not publish a required logic model template, so follow your solicitation's format if it gives one. When it does not, a theory of change layout works well for NSF proposals because it connects the project's purpose to its research questions and shows how far each outcome is expected to reach:
- Goal: one sentence across the top of the model stating the purpose of the project. It should connect to your research questions.
- Research questions: up to three, shown with the model rather than inside it. The activities are the work needed to answer them.
- Inputs: institutional and leadership support, faculty expertise, students, partners, prior research.
- Activities: five to seven high-level, action-oriented statements that start with a verb such as create, develop, or establish. Not tasks and not methods detail.
- Outputs: split into things you count (students served, partners engaged, meetings held) and things you create (a curriculum, a manual, a support program).
- Short-term outcomes: tangible results during the project period, tied to your measurable objectives. This is what the evaluation measures.
- Mid-term outcomes: results near or just after the end of the project. These can be aspirational and need not be measurable within the award.
- Long-term outcomes: the best-case change after the award. They link back to the motivating rationale and Broader Impacts, and you do not propose to measure them.
How the Model Drives Goals, Objectives, Deliverables, and Activities
The PAPPG asks the Project Description to outline the general plan of work, including the broad design of the activities. One clear way to organize that plan is a section on goals, objectives, deliverables, and activities. However you label it, it should be the logic model in prose: walk through the model in the order a reviewer reads the figure, and add what the figure cannot show, such as why each activity leads to its outputs and what the literature says about why those outputs should produce the outcomes.
A few rules keep the narrative and the figure in step:
- Each activity becomes an objective or a cluster of objectives. Use the activity's own wording so the reviewer can match them at a glance.
- Things you create are your deliverables. If the model lists a curriculum and a mentoring handbook, the narrative says what each contains and when it is finished.
- Things you count become measurable objectives. "120 students complete the redesigned courses" is both an output and a target your evaluation reports against.
- Every research question is answered by at least one activity. If a question has no activity behind it, either add the work or drop the question.
- Nothing appears in the narrative that the model omits. An activity described in prose but missing from the figure tells the reviewer the model is decoration.
How the Model Drives the Timetable
The PAPPG does not require a timetable, but some solicitations do: EPSCoR RII-FEC asks for a timetable or strategic plan, and ATE requires a timeline for its evaluation activities. Even when it is not required, a schedule is the most direct way to show reviewers that the plan is well-organized, one of the questions the merit review criteria ask.
When you include one, list every logic model activity in it with an estimated timeframe, along with the high-level tasks needed to complete each one. Timeframes can be terms, months, or recurring intervals, for example Fall 2027, Months 3 to 6, Quarterly, or Ongoing.
Run two checks before you submit. First, every activity in the model appears in the timetable. Second, every major task in the timetable traces back to an activity in the model. A timetable full of work the model never mentions suggests the model was drawn after the plan, not used to make it. If your solicitation also asks for a management plan, apply the same crosswalk there.
How the Model Connects to Evaluation and Broader Impacts
The evaluation plan should read like the right half of the model. Write evaluation questions tied to the short-term outcomes, name the data you collect for each, and say when. Track mid-term outcomes where the project timeline allows. Leave long-term outcomes unmeasured; they are your motivating rationale, not a promise. S-STEM makes this link explicit by asking for a logic model or theory of change that informs the evaluation plan.
Broader Impacts usually live in the outcome columns too. The PAPPG requires the Project Description to include a separate section labeled Broader Impacts. If the model's long-term outcome is a stronger technician workforce in your region, that section should say so in the same words and explain the path from this project's outputs to that change. Our NSF Broader Impacts guide covers how to make that section specific.
Worked Example: A Technician Education Project
A two-year college proposes a three-year project to grow its semiconductor technician certificate. Here is a compact version of its logic model.
- Goal: increase the number of students who complete the semiconductor technician certificate and enter technician jobs in the region.
- Research question: which supports raise first-year retention for part-time technician students?
- Inputs: faculty with industry experience, the college's cleanroom training lab, an industry advisory board, institutional research data.
- Activities: (1) redesign the two gateway courses around skills the advisory board validates; (2) establish paid summer internships with regional employers; (3) develop a peer mentoring program for first-year students; (4) create faculty externships with industry partners; (5) evaluate and refine the supports each year.
- Outputs counted: 120 students in redesigned courses, 30 internships, 6 faculty externships. Outputs created: two redesigned courses and a mentoring handbook.
- Short-term outcomes: higher gateway course pass rates; first-year retention data that answers the research question.
- Mid-term outcomes: more certificate completions; graduates hired by regional employers.
- Long-term outcome: a sustained regional pipeline of technicians.
Worked Example: The Timetable It Produces
Each line of the timetable names the activity it serves, so a reviewer can check coverage in seconds.
- Year 1, Fall: advisory board validates the skills list; course redesign begins (Activity 1).
- Year 1, Spring: redesigned courses piloted; first mentoring cohort recruited and paired (Activities 1 and 3).
- Year 1, Summer: first internship cohort and first faculty externships (Activities 2 and 4).
- Years 2 and 3: redesigned courses run at full enrollment; internships and externships repeat each summer (Activities 1, 2, and 4).
- Ongoing: peer mentoring each term (Activity 3).
- Annually, May: evaluation report to the advisory board, followed by refinements to the supports (Activity 5).
Common Mistakes in NSF Logic Models
Each of these is easy to catch before submission and costly to leave in:
- Moving the logic model to Supplementary Documents to save Project Description space when the solicitation does not ask for it there.
- Activities in the model that never appear in the timetable, or timetable work the model never mentions.
- Twenty activities that read like a task list instead of five to seven high-level statements.
- An evaluation plan that promises to measure long-term outcomes within a three-year award.
- Text shrunk until the figure is hard to read. The PAPPG allows type smaller than 10 points in figures and tables, but it also says readability comes first.
- A figure that the narrative never refers to. Point to it by caption where the goals section begins.
- Inputs and activities that do not match the budget and budget justification.
How GrantCopilot Helps with NSF Logic Models
GrantCopilot's logic model builder has an NSF Theory of Change format laid out like the one above, and its standard NSF Project Description template is organized around it: the goals part walks through the logic model, and the timetable covers every logic model activity. Here is what it does at each step.
- Build it from your goal and research questions. Enter your goal and up to three research questions, or ask for suggested questions based on your goal and keep the ones that fit.
- Draft the columns from your own text. GrantCopilot drafts inputs, five to seven activities, outputs you count and create, and short, mid, and long-term outcomes from your goal, your questions, and the text of a linked proposal. Your goal and questions stay as you wrote them. Where your text says nothing about a column, entries are described by type, such as "Faculty with expertise in the study area"; GrantCopilot does not name an institution, partner, or person, or add a number, that your text does not give. You can edit every entry.
- Back it with sources. The annotated bibliography finds recent peer-reviewed sources and checks each one's DOI against Crossref, or takes citations you paste or upload. You tag which parts of the model each source supports. From three or more sources, GrantCopilot writes a literature review in your citation style for the background and rationale, notes gaps in the literature, and suggests research questions that address them.
- Link it to your proposal. In an NSF proposal, the Logic Model row sits under the Project Description in the section list. Choose one of your saved logic models there or create a new one. One model can serve several proposals.
- Your drafts follow it. When you write or revise a section with AI, GrantCopilot describes activities, outputs, and outcomes the way your model words them and does not promise ones the model leaves out. Analysis flags sections whose activities or outcomes do not match the model, and model items a section should cover but misses.
- It exports with the Project Description. PDF and Word exports print the model as "Figure: Project Logic Model" on a landscape page right after the Project Description text, so it stays with the Project Description when you upload and its page counts toward your 15. If your solicitation wants the logic model as a Supplementary Document, as S-STEM does, upload that page separately instead.
Which GrantCopilot Plans Include NSF Logic Models
Three GrantCopilot plans include NSF logic models. Building a model is a planning tool; linking it to an NSF proposal and exporting it with the Project Description needs a plan that writes NSF proposals. Every plan starts with a 7-day free trial, and the pricing page has annual prices and full plan details.
- Research, $99/month: the NSF and NIH logic model formats with drafting, research question suggestions, the annotated bibliography, and literature reviews, plus NSF and NIH proposal writing, so you can link the model to your proposal and export it with the Project Description.
- Professional, $125/month: everything in the Research plan, plus every other logic model format and proposal type, for grant writers who work for many clients.
- Planning, $25/month: NSF logic models and the same planning tools (drafting, research question suggestions, bibliography, and literature review), but no proposal writing. To link the model to an NSF proposal and export it, you need the Research or Professional plan.
Frequently Asked Questions
Requirements come from solicitations, not the PAPPG. As of November 2026, S-STEM (NSF 25-514) requires a logic model or theory of change, ATE (NSF 24-584) makes one optional, and EPSCoR RII-FEC (NSF 24-573) accepts one alongside or in place of a timetable or strategic plan. Check the current solicitation for your program.
Yes, when it is part of the Project Description. The PAPPG counts visual materials, including charts, graphs, maps, photographs, and other pictorial presentations, toward the Project Description's 15-page limit, unless a solicitation sets a different limit.
Only when your solicitation asks for it there, as S-STEM (NSF 25-514) does. Otherwise, the PAPPG says special information and supplementary documentation belong in the Project Description, except for the specific documents it lists separately, and a logic model is not one of them.
A useful target is five to seven high-level, action-oriented statements, such as "Develop a peer mentoring program for first-year students." More than that usually means the model has slipped into task-level detail that belongs in the timetable.
No. Measure the short-term outcomes, which happen during the award, and track mid-term outcomes where the timeline allows. Long-term outcomes usually arrive after the award ends; present them as the motivating rationale rather than something the evaluation measures.
An NSF logic model earns its page when the rest of the Project Description is built from it. Check what your solicitation asks for, put the model inside the Project Description as a figure unless the solicitation says otherwise, budget its page at the outline stage, walk through it in the goals section, schedule every activity in the timetable, measure the short-term outcomes, and tie the long-term outcomes to your rationale and Broader Impacts. For the rest of the proposal, see our NSF grant proposal guide, the standard NSF proposal overview, and our advice on writing NSF proposals with clarity and precision. For the vocabulary, see the logic model glossary entry. Writing for other funders too? The nonprofit logic model guide and NIH logic model guide cover how they use the model.
Topics
logic model
logic model template
program evaluation
NSF logic model
theory of change
Project Description
NSF grants
NSF proposal writing