What is a theory of change?
A theory of change explains how and why a program’s activities are expected to lead to a long-term change for specific people, and names what must be true at each step for that change to happen. It is built backward from the change you want, and it tells you what evidence would show whether the explanation holds.
In a job-training program, training sessions are an activity; graduates in living-wage jobs a year later is the change. The theory of change spells out the steps between, such as certification and placement.
THE SHORT VERSION
- Start from the long-term change and the people who should experience it, then work backward to the earlier changes it depends on.
- Keep what you deliver (outputs) apart from what changes for people (outcomes), and write the assumption under every arrow.
- Give each output and outcome an indicator; those indicators become the reporting agreement, and the same model can be laid out as a logic model, logframe or results framework.
The Center for Theory of Change calls this backward mapping. If the idea is new to you, start with the seven-minute introduction below, and watch for the moment the example moves from what the program delivers to what changes for people.
What are the components of a theory of change?
Most theories of change have six parts: the problem, inputs, activities, outputs, outcomes and impact, with assumptions and outside factors written across the connections between them. Treat that as a starting structure; some models branch into several pathways. Name the factors you do not control, such as transport or employer demand, beside the arrows they affect.
The table fills each part with the running example in this guide: a fictional regional workforce fund and one of the four job-training partners it funds. It shows the plan, not reported results.
Scroll the table sideways to read all columns →
| Component | Question it answers | Workforce example |
|---|---|---|
| Problem | What needs to change, and for whom? | Adults seeking work lack the skills local employers hire for, and a way to show them |
| Inputs | What resources and relationships exist? | Trainers, mentors, employer contacts |
| Activities | What will the program do? | Training and mentoring |
| Outputs | What will it deliver or complete? | Participants enrolled; completed training; certified |
| Outcomes | What changes for people? | Placed in a job within 90 days of exit; retained in the same job at 12 months |
| Impact | What longer-term change is intended? | Graduates in living-wage jobs |
The line that matters most runs between outputs and outcomes. Enrolled and completed training are counts the program controls. Placement and retention are changes that also depend on employers and on each graduate’s life, so they are measured per person, with the people you could not reach named.
How do you build a theory of change, step by step?
Name the long-term change and who should experience it, work backward to the earlier changes it needs, connect your activities, then write the assumption under each arrow and plan the evidence. Starting from the change, not from your workshops, forces the question every funder asks: why would this work lead there?
The first draft often comes from a conversation. On the fund’s onboarding call, the partner said it wants “graduates in jobs that pay a living wage, and still there a year later.” That sentence, recorded with consent, became the chain of steps and the five metrics both sides agreed to report.
Define the problem with the people affected
Ask participants and frontline staff what a useful outcome means to them; graduates may value a steady schedule as much as the wage.
Name the long-term change
“Improve lives” guides no decision; “graduates in living-wage jobs” gives the team something to examine.
Work backward through the outcomes
For each outcome, ask what must happen first. Living-wage jobs need people still employed at 12 months, which needs placement within 90 days of exit, which needs certification.
Connect the activities
Link training and mentoring to the first changes they should produce. An activity that connects to no outcome is a question for the budget.
Write the assumptions
Under each arrow, write what must be true for the next step to follow, and mark which still need testing.
Plan the evidence and the review
Give each output and outcome an indicator with a definition, a source and a due date, and agree when you will review the model.
The chapter Turn the onboarding call into a reporting agreement drafts the model and the metrics from a call transcript; the theory of change template gives you boxes to fill.
What does a theory of change look like? A worked example
A theory of change diagram is a chain of boxes from activities to the long-term change, with a short assumption beside every arrow. This is the model shown at the top of this page, now with the evidence that would test each step. Read each arrow as a question: why should this step lead to the next?
- ActivitiesTraining and mentoringContent matches what local employers hire for.
- OutputsEnrolled, completed, certifiedEmployers recognize the certification.
- OutcomePlaced within 90 days of exitThe job fits the person, so they stay.
- OutcomeRetained at 12 monthsThe job, or the next one, pays a living wage.
- ImpactGraduates in living-wage jobsThe change the partner named on the call.
Evidence to follow: enrollment form and completion record → placement survey → 12-month follow-up. A step with no evidence behind it is still an assumption.
One assumption usually matters more than the rest. Here it is that suitable jobs are within reach: if that is wrong, placements stall however good the training is. Name who will check it, with which records, and by when.
Keep each earlier model with the reports that used it, noting what changed and who approved it; BetterEvaluation’s guidance shows how research, experience and new evidence feed those revisions. The walkthrough below builds a model with an AI tool; watch how each generated step is checked against the source material.
Is a theory of change the same as a logic model, logframe or results framework?
They are four layouts of one change model: the same five levels, from inputs to impact, under different labels and with different room for assumptions and indicators. Build the model once, then lay it out in whichever format a funder asks for.
Scroll the table sideways to read all columns →
| Format | What it emphasizes | Where you meet it |
|---|---|---|
| Theory of change | Why each step should lead to the next; assumptions and context | Strategy, program design, learning reviews |
| Logic model | A compact delivery plan from inputs to outcomes | Program planning and grant applications |
| Logframe | A matrix with indicators, means of verification and assumptions per row | International development and government grants |
| Results framework | A tree read down from one strategic objective | Large public donors |
In the workforce model, a logframe calls the two outcomes “purpose” and the living-wage aim “goal”; a results framework puts living-wage jobs at the top as the strategic objective. The 90-day and 12-month windows do not move.
Re-lay rather than rebuild, because two separately written models soon report two different placement numbers. See the logic model guide, the theory of change vs logic model comparison, the logframe guide and the results framework guide.
Where do the indicators in a theory of change come from?
From the model itself: each output and outcome gets an indicator, and those indicators become the rows of the reporting agreement and the entries in a shared data dictionary. If a metric fits no level, ask why you collect it; if a level has no indicator, it is an assumption you have not yet tested.
In the example, enrolled (unique people, not visits) and completed training are outputs; placed within 90 days of exit and retained at 12 months are outcomes; starting wage, hourly and by track, speaks to the living-wage aim.
Each indicator then needs one definition both sides count the same way, a dimension from the Five Dimensions of Impact and, where one fits, an IRIS+ code. The chapter A shared data dictionary builds those entries.
PROMPT · PASTE INTO CLAUDE, CHATGPT OR YOUR AI TOOL
Below are our theory of change and the metrics we report. Check the model against the metrics. Do not rewrite the model. Rules: 1. List our levels in order: inputs, activities, outputs, outcomes, impact. 2. For each output and outcome, name the metric from our list that measures it. If none does, write "not in our data". 3. List every metric that fits no level. 4. For each arrow, quote the assumption we wrote. If we wrote none, write "no assumption stated". Do not invent one. 5. Do not add outcomes, targets, baselines or numbers that are not in the text below. 6. End with the assumption that would most change our plan if wrong. Our theory of change: [PASTE] Our metrics and their definitions: [PASTE]
What can a theory of change prove, and what can’t it?
A theory of change states how you expect change to happen; it proves nothing until evidence tests each arrow, and even then a change in results is not proof that the program caused it. Keep three questions apart: did the activity happen, did the outcome occur, and how much did the program contribute?
Delivery records answer the first. Follow-up answers the second for the people you reached, within the limits of self-report and non-response: graduates who answer at 12 months may differ from those who do not. The third needs a comparison or a method such as contribution analysis.
Outputs are not outcomes: a count of people enrolled says nothing yet about who was placed. An AI tool can draft a model from your notes but cannot supply views nobody collected; people who know the setting decide what is plausible and check every draft against the record.
In Sopact Sense each participant keeps one ID from the first form, so enrollment, placement and the 12-month follow-up sit on the same person, and every line of an AI Assistant answer links to a record you can open. The method itself works with any tools.
Start with one program’s theory of change
Pick one program with a report due this year and build its model around the metrics you already send.
- Write the long-term change in one line and name who experiences it.
- Work back to three or four earlier changes, then add your main activities.
- Under each arrow, write one assumption, and mark the one that would most change your plan.
- Put each metric you already report on a level; list the levels with no metric and the metrics with no level.
- Ask two participants and one frontline staff member whether the chain matches their experience, and revise.
- Lay the model out in the format your main funder asks for, changing only the labels.
After the first cycle you have one model, named assumptions, an indicator for each output and outcome, and the question your next report should answer.
Frequently asked questions
What does theory of change mean?
It means a reasoned explanation of how a particular change is expected to happen for particular people. It links activities to outcomes and a long-term change, and makes the assumptions behind each link explicit. It stays open to testing: as evidence arrives, you keep the links that hold and revise the ones that do not.
How do you write a theory of change statement?
Use an “if, then, because” sentence that names the people, the activity, the expected outcome and the reason the link should hold, then add the main condition. For the fictional workforce program: if adults receive training and mentoring, then more will be placed within 90 days and retained at 12 months, because they can show the skills employers hire for, provided jobs are within reach.
How long does it take to build a theory of change?
It depends on the program and the evidence you already hold. A team that knows a small program well can draft a first model in one working session, especially from an onboarding call or an existing logic model. A partnership of several organizations may need rounds of consultation. Agree on a draft and a review date rather than waiting for a finished model.
What is the role of a theory of change in monitoring and evaluation?
It tells you what to monitor and which evaluation questions matter. Each output and outcome gets an indicator, which is monitored all year against the reporting agreement, and each assumption becomes a question an evaluation can test. See theory of change in monitoring and evaluation for how the model becomes the M&E plan.
Can AI build a theory of change?
An AI tool can turn a call transcript or program notes into a draft model, suggest assumptions and flag levels with no indicator. It cannot supply views of people it never heard or judge which links are plausible. Tell it to use only your material and to write “not in our data” where something is missing, then check every line.
Where can I find theory of change examples and a template?
The theory of change examples page compares models across workforce, education, health and environmental programs, with outputs and outcomes separated in each. The theory of change template gives a copyable structure filled with the workforce model. Adapt either to your own participants and partners rather than copying another program’s assumptions.
