Evaluation Types Compared

One Family, Many Questions

"Evaluation" is not a single method. It is a family of approaches, and each one answers a different question at a different moment in a program's life. Picking the wrong type is a common and costly mistake. The evaluation ends up unable to answer the questions it was meant to answer, because the right data was not collected at the right time.

This page compares seven evaluation types, then compares three options for who conducts the evaluation. Process and outcome evaluation are the two most programs need. The others are worth knowing so you can tell when they apply and when they do not.

The Seven Types Side by Side

Terms used on this page:

• Fidelity: whether a program is delivered the way it was designed.
• Attribution: showing that the program caused a change, rather than something else.
• Quasi-experimental: a design that builds a comparison group without random assignment.

Costs are relative ratings, not prices, and vary by setting. The methods and best uses differ for each type:

• Process: document review, interviews, observation and fidelity checklists. Best for new programs, complex interventions and checking readiness to scale.
• Outcome: before and after comparison, surveys and mixed methods. Best for most programs; it answers "did it work?"
• Impact: randomized or quasi-experimental designs, or contribution analysis. Best for large programs that need causal evidence for a scale-up decision.
• Formative: rapid assessments, participatory methods and real-time feedback. Best for pilots, adaptive programs and testing innovations.
• Summative: mixed methods, meta-analysis and cost-effectiveness analysis. Best for closing programs, accountability and future funding decisions.
• Developmental: an embedded evaluator and methods suited to complexity. Best for innovation and systems change. It needs a long-term embedded evaluator, so it is not a standard option for most grant-funded programs.
• Real-time: rapid surveys, remote sensing and key informant networks. Best for humanitarian response and emergency programs.

Four Questions Before You Choose

Answer these first. The answers usually point to one or two types.

1. What decision will the evaluation inform: scale-up, funding renewal, learning or accountability?
2. When will it happen: midway, at the end, or after activities conclude?
3. Do you need to prove attribution, or is a plausible contribution enough?
4. What is the budget, and what level of independence does the donor require?

Whatever type you choose, it should still cover the questions your donor frames. Many donors use the OECD DAC criteria (revised 2019): relevance, coherence, effectiveness, efficiency, impact and sustainability.

If the answers are unclear, start with an outcome evaluation. It fits most programs, most stages and most donor requirements.

The Three Most Confused Pairs

Process and outcome. A process evaluation asks whether the program delivers what it was designed to deliver, in the way it was designed. An outcome evaluation asks whether the intended changes happened among the people the program serves. A program can have excellent fidelity and still see no change. It can also see change while the delivery drifts from the design. Fix the process first, then measure the outcome.

Outcome and impact. An outcome evaluation says that these changes occurred in the target population. An impact evaluation says the program caused them, and not other factors. That claim needs a credible comparison: a control group, a comparison group or a rigorous analytical approach. Most programs lack the sample size, the budget or the comparison group for it. A sound outcome evaluation with a plausible contribution argument is usually the right answer.

Formative and summative. These terms describe purpose, not method. A formative evaluation feeds back into improvement while the program runs. A summative evaluation judges overall worth at the end. One evaluation can do both. A midterm often reports progress to the donor (summative) and informs the next year's adjustments (formative).

Most small and mid-size organizations do not need an expensive impact evaluation. A solid process evaluation in Year 1 and an outcome evaluation at the end cover most needs. Save impact evaluation for when the budget, the sample size and a real need for proof of causation all exist.

Who Conducts It: Internal, External or Hybrid

Every evaluation also involves a choice about who does the work. The options run from fully internal to fully external, with a hybrid model in between.

Use each option where it fits:

• Internal: small pilots, formative evaluation and learning reviews where the accountability stakes are low.
• External: donor-required evaluations, high-stakes summative evaluations and sensitive programs.
• Hybrid: midterm and endline evaluations where the team has capacity and cost is a constraint.

When findings carry high-stakes decisions, or the donor requires it, use an external evaluator with documented independence, meaning no prior relationship with the program or its donor. Put the cost in the M&E budget from the start. No universal percentage of program cost fits every evaluation, so base the line on a costed plan for design, fieldwork, analysis and reporting. Without a budget line, the bill arrives in Year 2 with nothing to pay it.

Sector note, education. Evaluations often use test scores as outcome indicators. A score gain can reflect teaching to the test rather than learning. Add a process component that checks whether teachers actually changed their practice.

Sector note, governance. Policy change and institutional reform have many causes beyond one program. Contribution analysis, which assesses the program's plausible contribution rather than sole attribution, often suits governance work better than an experimental design. Ask "what did this program contribute to the change?" and not "did it cause the change?"

Worked Example: Healthy Futures

Healthy Futures is a fictional child nutrition program in Kasenda District. It runs 36 months and is funded by the Global Health Partnership Fund (GHPF). GHPF expects a midterm evaluation, with the final report due in Month 24.

The M&E officer starts from the questions GHPF wants answered, not from a type. Is the program delivered as designed? That is a process question. Are early outcomes appearing? That is an outcome question. What should change in Year 3? That is a formative question. An impact evaluation is ruled out: a midterm point in a 36-month program has too little follow-up time to answer an attribution question.

The recommendation is a combined process and outcome evaluation with a formative lens. It is not an impact evaluation, because the aim is to assess progress and surface lessons, not to prove causation.

For independence, the team chooses an external evaluator because the midterm findings will shape Year 3 decisions and report to the donor. The external evaluator leads and writes the report, and the team gives access to monitoring data and field logistics. The cost ceiling is $35,000, and the evaluator's contract is $28,000.

Working back from Month 24: the terms of reference are drafted in Month 16 and circulated in Month 17, when the contract is signed. Inception is Month 18, fieldwork Months 20 to 21, the draft report Month 22 and the final report Month 24.

The type followed the questions, and the evaluator followed the stakes and the budget.

Where to Go Next

Write the chosen type and evaluator into the evaluation terms of reference. For the individual types, see process evaluation, impact evaluation, formative vs summative evaluation and developmental evaluation.