Theory of change for NPOs: from programme logic to evidence

What is a theory of change?

A theory of change is a structured explanation of how and why a programme is expected to lead to the change it aims for. It names the problem, the long-term change, the outcomes that must happen first, the activities meant to produce them, and the assumptions that must hold along the way.

The World Health Organization describes it as an approach to developing a systematic and visual representation of how a programme or intervention is expected to achieve its intended outcomes (WHO, 2025). WHO lists its uses as designing and implementing interventions, monitoring and evaluation, transparency, policy transfer and scale-up, and institutional strategic plans.

Many NPOs build one for a proposal, a strategy session or a board workshop, and then file it. The diagram stops shaping what the team collects, which assumptions it tests, or what goes into reports. This guide is about keeping it in use. It is one part of our wider guide to monitoring and evaluation reporting for South African NPOs, which covers the full reporting workflow.

A theory of change does not prove impact. It gives the team a logic it can test with evidence.

How is a theory of change different from a logic model?

A theory of change explains the why and how: causal links, assumptions and context. A logic model summarises the what, usually as a linear chain from inputs to impact. Most programmes need both.

A logic model typically follows this chain:

Inputs -> Activities -> Outputs -> Outcomes -> Impact

DPME's outcome evaluation guideline puts it this way in practice: the theory of change and the logical framework together structure the programme's theory into measurable parts, which is what makes monitoring and evaluation possible (DPME, Guideline 2.2.28, 2025).

Where should a theory of change start?

Start with the problem, not the programme. A theory of change built around existing activities tends to justify them rather than test them.

A weak problem statement: "Youth unemployment is high." It is true but too broad to guide a programme.

A stronger one: "Young people in this community are leaving school without the support, networks, information and pathways they need to enter further study, training or work." It names the group, the context, the barriers and the kind of change the programme might influence.

Questions to settle first:

• Who is affected, and where?
• What immediate problem are we addressing, and what root causes sit behind it?
• What system barriers make it harder to solve?
• What can we realistically influence, and what sits outside our control?

WHO's technical guidance sets out six stages: define the problem, define expected outcomes, define the interventions, define change mechanisms and build the model, validate the theory of change, and revise it (WHO, 2025). A vague problem leads to vague outcomes, vague outcomes lead to weak indicators, and weak indicators make reporting harder.

How do you map the pathway backwards?

Begin with the long-term change, then ask what must be true just before it, and keep working back until you reach the activities. DPME recommends this backward mapping for government programmes because it puts outcomes before activities.

DPME's outcome evaluation guideline gives a five-step protocol (DPME, Guideline 2.2.28, 2025):

1. Define the long-term change or societal goal.
2. Identify the intermediate and short-term outcomes that must occur to reach it.
3. Map the causal pathways showing how interventions contribute to those outcomes.
4. Test assumptions and identify external factors through stakeholder consultation.
5. Refine and document the theory of change visually, aligned with monitoring indicators.

An illustrative pathway for a youth development NPO:

This does not guarantee change. It gives the team a working theory it can check against evidence.

How do you make assumptions testable?

Write each assumption next to the causal link it supports, then decide what evidence would show whether it holds and when you will check. Assumptions left in workshop notes are where most programme logic quietly fails.

UNEP's evaluation office separates two kinds of external factor. Drivers are external conditions needed for one result to lead to the next that the project can influence to some degree. Assumptions are similar conditions that the project cannot control (UNEP Evaluation Office). The distinction matters: you manage drivers, and you monitor assumptions.

Typical NPO assumptions: participants can attend regularly, families support participation, schools cooperate, referral partners have capacity, staff have time to collect data.

How do you turn a theory of change into indicators?

Give every major outcome and every key assumption at least one way to observe progress. That does not mean a complex measurement system for each; it means the team can say what evidence would show whether that part of the pathway is working.

Four kinds of measure usually cover it:

• Output indicators measure delivery: sessions held, participants reached, home visits completed, referrals made.
• Outcome indicators measure change: knowledge, confidence or behaviour, improved access to services, completion rates.
• Assumption indicators test whether the logic still holds: attendance rate, partner response time, referral completion rate, staff data submission rate.
• Qualitative evidence explains what the figures do not: interviews, case studies, focus groups, fieldworker observations, open-ended responses, partner feedback.

A theory of change without indicators is hard to test. Indicators without a theory of change are easy to collect but hard to interpret. Our guide to SMART indicators and indicator reference sheets covers how to define each one so it is collected the same way every time.

Large agencies work the same way. UNICEF's theory of change for child protection in humanitarian action is organised around outcome pillars, with monitoring frameworks for several of its workstreams, and is intended to make programmes easier to evaluate (UNICEF, undated).

This is also what evaluators look for. DPME's National Evaluation Policy Framework describes design evaluations as assessing the strength of the theory of change, the logframe, the indicators and the implementation plan (DPME, NEPF 2025). DPME's outcome evaluation guideline goes further and advises checking readiness before commissioning an outcome evaluation: a clearly articulated theory of change, baseline data, and outcomes that can be measured (DPME, Guideline 2.2.28, 2025).

How do you link a theory of change to evidence?

For a small programme, a matrix linking each outcome to its indicator, evidence source, data owner and report section is enough. For larger programmes with interviews, case studies and partner feedback, the theory of change works better as a structured evidence database, where each evidence point carries IDs that link it to its source and to an outcome.

The purpose is to keep the logic connected to source material while protecting confidentiality. Reports draw on reviewed evidence statements, not raw interview text.

In large qualitative studies, for example a situation analysis drawing on interviews with service users and providers, this structure is usually built near the end: raw material is anonymised, each source gets an ID, evidence points are extracted and coded to theory of change domains, and the coded set then supports findings and recommendations. It does not make qualitative data statistically representative, but it lets the team see where evidence is concentrated and where gaps remain.

The table-level mechanics are covered in our guide to a source-linked evidence table for a report. For the same method at programme scale, with indicators linked to their sources, see the Social Employment Fund evidence portal.

Where teams want this structure in place and owned internally, we offer a source-linked evidence system, built and handed over.

How should a theory of change shape reporting?

Each report section should map back to the pathway, so the report reads as a test of the logic rather than a list of activities.

1. What we planned to change
2. What we delivered
3. What changed
4. What evidence supports each claim
5. Which assumptions held or failed
6. What we learned
7. What we will adapt next

A donor report built this way starts from the evidence base, not a blank page: output figures from the delivery tracker, outcome signals from surveys and interviews, coded quotes with source IDs, assumption notes from review meetings, and risks and adaptations from programme logs.

Once a theory of change evidence base is reviewed and anonymised, some teams add AI-supported retrieval over it to find past findings or compare themes. That only works on structured, source-linked material, and people check every output against its source. AI does not decide what the evidence shows.

How often should a theory of change be reviewed?

Review it at least once a year, and whenever the evidence or the context shifts. WHO's guidance makes revision a formal stage of the process, not an afterthought (WHO, 2025).

Triggers for review: the programme expands or enters a new phase, the context changes, monitoring shows weak progress, an assumption proves wrong, funder requirements change, or an evaluation challenges the original logic.

A workable rhythm:

• Monthly: check outputs and activity data.
• Quarterly: review outcome signals and assumption indicators.
• Annually: revise the theory of change based on evidence.
• Before an evaluation: check that the logic is testable and has evidence behind it.
• Before a funding renewal: update the theory of change and its evidence matrix.

What templates help keep a theory of change in use?

Two templates do most of the work: a one-page canvas that keeps the logic visible, and an assumption register with review dates. Copy and adapt them.

One-page theory of change canvas
Problem:                 What specific problem are we addressing?
Target group:            Who is affected, and where?
Long-term change:        What change are we working towards?
Intermediate outcomes:   What must change before that?
Short-term outcomes:     What early changes should happen first?
Outputs:                 What will be delivered?
Activities:              What will we do?
Assumptions:             What must hold true at each link?
Risks:                   What could weaken the pathway?
Indicators:              How will progress be observed?
Evidence sources:        Where will the evidence come from?
Data owner:              Who collects and reviews it?
Reporting use:           Which report section uses it?

Assumption register
Causal link | Assumption | Confidence (low/med/high) | Evidence available | Evidence needed | How we will test it | Owner | Review date

The evidence sources, data owner and reporting use lines are the ones most often missing. Without them, the theory of change stays a strategy document.

Common mistakes

• Building around activities, not outcomes. Start from the change you are trying to support.
• Vague impact statements. Name the group, the context and the type of change, or nobody can test it.
• Hidden assumptions. Assumptions explain where the logic may fail; monitor them.
• Indicators added after implementation starts. You miss baselines and early evidence.
• Treating qualitative evidence as informal. Stories, interviews and field notes are evidence once captured, coded, reviewed and linked to outcomes.
• A diagram only a specialist can read. If staff cannot explain it, they will not use it.
• No data owners. Every indicator and evidence source needs a named person.
• Used in the proposal, absent from the report. If reports do not map back to the theory of change, the organisation cannot show its logic was tested.