Connect the problem, the activity, and the outcome
Evaluation is clearer when the team can show why an activity should change a condition and what else may explain the result.
Baseline
Describe the starting condition, data limits, period, and groups affected before treating a change as improvement.
Delivery
Check whether the program reached the intended people and whether its core work happened with enough quality.
Outcome
Compare the result with the baseline and plausible alternatives. Report null findings and uncertainty honestly.
Start with a simple logic chain: resources support activities, activities reach participants or places, and that exposure may produce short and longer-term outcomes. If the chain is not explicit, a disappointing result cannot show whether the idea was weak, implementation failed, or the intended group never received the intervention.
Choose measures before reviewing results. Include delivery measures, user experience, intended outcomes, unintended effects, and relevant context. Compare with a baseline when feasible and document missing data or changes in collection. A decline after launch is not proof of causation, just as a null result is not proof that every component had no value.
- Define the decision the evaluation must support.
- Measure whether the program was delivered before judging outcomes.
- Report uncertainty, alternative explanations, and unequal effects.
Useful measure set
For a lighting repair project, implementation measures include fixtures repaired and outage time. Experience measures include whether users report better visibility. Outcome measures may examine incidents over an appropriate period, alongside nearby patterns and other changes.
Why a logic model and baseline matter
- A reduction after launch does not by itself prove the program caused it.
- Resident experience can reveal changes that administrative data misses, but collection must protect privacy and avoid coercion.
- Measures should not reward displacement of a problem to a nearby block or another group.
Evidence basis: The cited sources support the checks above. They do not replace current provider, property, workplace, authority, or local requirements. CDC: Prevention Resources for Action.
Crime prevention evaluation checklist
Write the logic and measures before reading the final result. That reduces the temptation to choose only favorable evidence later.
Causation, privacy, and small samples
A change after launch does not by itself prove causation. Use proportionate methods, report missing data and null findings, protect participant privacy, and avoid claims the design cannot support.