A logic model is a visual diagram showing how a program's inputs, activities, outputs, outcomes, and impact connect in an if-then chain. It is the standard tool program designers, evaluators, and funders use to articulate what a program does, what it produces, and what change it is meant to create. The W.K. Kellogg Foundation's 2004 Logic Model Development Guide is the most widely cited version, and the five-component structure — inputs, activities, outputs, outcomes, impact — has been the canonical shape for two decades.

A logic model has five canonical components in left-to-right order. Inputs are the resources committed to the program — funding, staff, materials, partnerships, time. Activities are what the program does with those inputs — workshops delivered, services provided, curriculum taught. Outputs are the direct, countable products of activities — number of participants served, hours of training delivered, documents produced. Outcomes are the changes that result for participants — short-term changes in knowledge or attitude, medium-term changes in behavior, long-term changes in condition. Impact is the broader, longer-term societal or systemic change the program contributes to.

Inputs are what goes into the program — resources committed before any work happens. Outputs are what comes out of the activities — countable products of the work done. Inputs are about resourcing the program. Outputs are about quantifying the work the program produced. A grant of $200,000 is an input. The 150 participants who completed the training funded by that grant is an output.

Outputs are the countable products of activities — how much was delivered. Outcomes are the changes that resulted — what changed for participants because of the delivery. 150 participants trained is an output. 90 of those participants placed in jobs within 90 days at a wage above their previous earnings is an outcome. The output question is "what did we do?" The outcome question is "what changed because we did it?"

A theory of change is the narrative reasoning about why a program should work — the assumptions, the causal pathways, the contextual factors. A logic model is the operational diagram of how the program will work — inputs flowing through activities into outputs, outcomes, and impact. Theory of change answers why and under what conditions. Logic model answers what and in what sequence. Strong programs build both: theory of change first to articulate the reasoning, logic model second to operationalize the delivery. See theory of change for the upstream framework.

The W.K. Kellogg Foundation logic model is the version published in the foundation's 2004 Logic Model Development Guide, which became the most widely adopted logic model framework for nonprofits, public health programs, and education initiatives. It uses the five-component structure (inputs, activities, outputs, outcomes, impact) and introduces the if-then chain — the explicit logic that if these inputs are committed, then these activities can happen; if these activities happen, then these outputs result; and so on through outcomes and impact.

The if-then chain is the causal logic that connects the five components of a logic model. Reading left to right: if these inputs are committed, then these activities can be delivered. If these activities are delivered, then these outputs will be produced. If these outputs are produced, then these outcomes will follow for participants. If these outcomes occur at sufficient scale, then this broader impact will be contributed to. The if-then chain forces program designers to articulate the assumptions linking each step — and to identify which links are evidence-based and which are aspirational.

Build a logic model in five steps. 1) Start with the impact and work backwards — name the long-term societal or systemic change the program is contributing to. 2) Identify the outcomes that would have to occur for that impact to be plausible. 3) Specify the outputs that would have to be produced to drive those outcomes. 4) List the activities that would produce those outputs. 5) Enumerate the inputs required to run those activities. The reverse-order build prevents the common failure mode of starting with available inputs and reasoning forward into whatever fits.

A workforce-training nonprofit logic model: Inputs include $400K grant funding, 4 instructors, partnerships with 12 employers, a curriculum, and a community space. Activities are recruitment, 12-week training cohorts, employer matching events, and post-placement coaching. Outputs are 200 participants enrolled, 160 completing the program, 12 employer matching events held, and 6 months of post-placement coaching delivered. Outcomes are 120 participants placed in jobs within 90 days, average wage of $24/hour (vs $16 pre-program), 80% retention at 6 months. Impact is reduced regional underemployment, increased family economic stability for participants' households, and a replicable model for peer nonprofits. Four full worked examples appear in the Examples section above.

Logic model software falls into four eras. PowerPoint and Word templates from foundations like Kellogg and CDC are still used for one-off models. Diagram tools like Lucidchart, Miro, and SmartDraw handle the visualization layer with logic-model templates built in. Evaluation platforms like DevResults, Salesforce Nonprofit Cloud, and Tableau dashboards include logic-model fields but do not automate construction. AI-native impact platforms — Sopact Sense in this category — replace the six-month logic-model project with an outcome-first design that generates the full inputs-to-impact chain from a short program goal and keeps it updated as program data flows in.

AI can do meaningful work on logic models — and cannot do all of it. Generative AI tools like ChatGPT, Claude, and Perplexity can draft a logic model from a program description, suggest outcomes from impact statements, and identify gaps in the if-then chain. They cannot maintain that logic model as live program data flows in, validate outputs against actual program delivery, or join the model to participant-level outcomes. The strong pattern is partnership: an AI-native impact platform manages the persistent program record and the data joins; the generative model drafts and refines the logic-model content.

Traditional logic-model projects ran four to six months — staff workshops, consultant engagement, multiple draft rounds, board approval cycles. The classical W.K. Kellogg Foundation development guide assumed this timeline because the manual work of articulating assumptions, drafting components, and securing stakeholder alignment took months. AI-native platforms compress the initial build to days. The shift is not "AI replaces the team's thinking" — it is "AI handles the drafting and document mechanics so the team's thinking goes into the parts that matter, the assumptions and the outcome priorities."

In grant writing, a logic model is the diagram that demonstrates to funders how the proposed program will work — how the grant funding (inputs) will be converted into activities, outputs, outcomes, and impact. Most major foundations and federal funders require a logic model as part of proposal submissions. A strong grant-writing logic model is specific, evidence-based on the if-then links, and clearly distinguishes outputs from outcomes — the most common reviewer critique.

In program evaluation, a logic model is the foundation document that defines what the evaluation will measure. The model identifies which outputs the evaluation will count, which outcomes it will assess change against, and which impact-level indicators it will track over time. Evaluation plans that lack a logic model tend to drift toward measuring whatever data is convenient to collect rather than whatever data answers the question "did the program produce its intended changes."

Inputs in a logic model are the resources committed to make the program possible. Categories include financial inputs (grants, contracts, donations), human inputs (staff, volunteers, consultants), material inputs (curriculum, facilities, equipment, technology), and relational inputs (partnerships, community ties, regulatory permissions). Strong logic models list inputs at sufficient specificity that a reader can answer "is this program adequately resourced?" Vague inputs ("funding, staff") signal a logic model that has not yet been costed against the work it proposes.

Outputs in a logic model are the direct, countable products of program activities — quantifying what the program produced. Examples: number of participants served, hours of training delivered, services provided, materials distributed, events held, reports published. Outputs answer "how much did we do?" They are necessary but not sufficient — strong programs distinguish outputs from outcomes because output achievement does not by itself demonstrate impact. A program that delivered 500 training hours achieved high output. Whether those hours produced learning, behavior change, or placement is an outcome question.

Outcomes in a logic model are the changes that result for participants because of program activities. Outcomes are typically grouped by time horizon. Short-term outcomes are changes in knowledge, attitude, or skill (a participant learned the concept). Medium-term outcomes are changes in behavior or practice (the participant applied the concept). Long-term outcomes are changes in condition or status (the behavior change produced a sustained life change). Outcomes are where impact measurement actually lives. Programs that confuse outcomes with outputs systematically over-report success.

Impact in a logic model is the broader, longer-term societal or systemic change the program contributes to — the fundamental shift the program is part of producing alongside other actors and forces. Impact is typically beyond the reach of any single program in isolation. A workforce program contributes to regional economic mobility. A literacy program contributes to community educational attainment. A health intervention contributes to disease-burden reduction. Impact is contribution rather than attribution — the program is one of many factors moving the larger system.