Distinction
Three Different Jobs in an Evaluation Report
A finding states what the evidence shows. A conclusion interprets findings to answer an evaluation question or make a judgment against stated criteria. A recommendation proposes a response to a conclusion. Each can be concise, but they are not interchangeable labels for the same sentence.
The distinction matters because readers need to challenge different things. They may accept the observed pattern but disagree about its importance, or accept the judgment but prefer a different response. A report that blends all three leaves those disagreements hidden.
Start from the agreed evaluation questions and the evaluation matrix. Decide what evidence would inform each question before drafting a strong conclusion. The report should expose the argument, not make a confident paragraph stand in for it.
Compare the Three Statements
| Statement | Main Question | What It Needs | Typical Error |
|---|---|---|---|
| Finding | What does the evidence show? | Identifiable sources, analysis and limitations | Presenting an impression as a measured pattern |
| Conclusion | What does that mean for the evaluation question? | Relevant findings and an explicit basis for judgment | Introducing a claim not established in the findings |
| Recommendation | What response is warranted? | A supported conclusion, intended user and feasible action | Asking someone to solve a problem outside their authority |
Findings can include analysis. A pattern derived from several interviews, records and observations is more than a data extract. Show how the sources support it, including important exceptions. Conclusions move from that account to its evaluative meaning.
A conclusion about effectiveness, for example, should explain which intended result is assessed and how the evidence relates to it. A count of completed activities alone cannot establish that participants experienced the intended change.
Evidence Chain
A Worked Evidence Chain
The following example is hypothetical. Its numbers and circumstances are illustrative, not findings from a real service.
An evaluation asks whether a community advice service is accessible to the people it intends to reach.
Finding. In the reviewed booking records, 18 of 30 canceled appointments were marked as conflicts with paid work. Interviews with attending and nonattending users identified limited evening availability. The cancellation reasons were entered by staff and were incomplete for some bookings, so they do not establish the reasons for every missed visit.
Conclusion. Current opening hours create an access barrier for some users with inflexible working hours. The evidence supports that specific access concern; it does not establish the share of all eligible residents affected or prove that opening hours explain the service's overall uptake.
Recommendation. The service manager should assess an evening-session trial with user representatives and staff before the next scheduling decision, document its staffing and safety implications, and define how attendance and user feedback would inform a continuation decision.
The recommendation does not assert that an evening session will solve the problem. It proposes a proportionate response to the supported conclusion and identifies what must be assessed before resources are committed.
Make the Chain Traceable
Give findings and conclusions stable identifiers when the report has several linked arguments. An evidence table can record the question, source references, finding, confidence or limitations, conclusion and associated recommendation. A simple reference is enough when the report is short.
Look both backward and forward. For each recommendation, ask which conclusion warrants it. For that conclusion, identify the findings that support the judgment. Then check whether any important finding contradicts the conclusion or suggests a different response.
Use triangulation to examine agreement and disagreement between sources. Repeating the same claim across documents does not create independent confirmation if all documents use one underlying source. Explain why a source deserves weight rather than counting how many times the claim appears.
Keep limitations attached to the claims they restrict. A methodological caveat in an annex does not repair an unqualified headline saying that the program caused change. If evidence cannot resolve a question, an explicit uncertain conclusion is more useful than a decisive answer without support.
Recommendations
Write Recommendations Someone Can Assess
A usable recommendation identifies the intended decision-maker, the change sought, the reason for it and a sensible decision point. Clarify any dependencies, resource implications and evidence needed to judge completion. Details should be specific enough to assess without prescribing operational choices that the evaluation cannot justify.
Prioritize by the seriousness of the supported issue and the opportunity to act. Explain sequencing where one action depends on another. “Improve coordination” gives little direction; “agree which team is responsible for following up unresolved referrals before revising the referral process” names a decision.
Discuss feasibility with responsible people while preserving the evaluator's independent judgment. Consultation can reveal a legal constraint, a conflicting obligation or an option the team missed. It should not let management delete an inconvenient evidence-based conclusion.
A recommendation is not management's commitment. The management response should state whether it is accepted, partly accepted or rejected, with reasons and agreed actions. Assigned owners, dates and resources become organizational commitments through the appropriate management process.
Common Mistakes
- Repeating a Finding as a Conclusion. Explain its meaning for the question and criterion rather than restating the number.
- Adding Evidence Late. Put the supporting analysis in the findings so readers can assess it.
- Treating Every Problem as a Training Need. Consider whether authority, resources, incentives or the process itself explains the problem.
- Recommending Beyond the Evidence. Reduce the scope of the action or propose a bounded test when confidence is limited.
- Confusing Acceptance With Completion. Track the action agreed in the management response and the evidence of change, not merely the date a recommendation was accepted.
FAQs
Frequently Asked Questions
Can a Finding Include Analysis?
Yes. A finding can be an analytical statement derived from several sources, not just a raw number. Make its evidence and limitations visible, and distinguish it from the evaluative judgment made in the conclusion.
Does Every Conclusion Need a Recommendation?
No. A conclusion may establish that something works, identify uncertainty, or describe a constraint outside the intended user's authority. Recommend action only where the evidence supports a useful response.