When to use it
Use it before implementation when a later impact claim will matter.
When to be cautious
A before/after difference or observational regression alone does not establish causality.
Apply it
- Define the evaluative question and plausible counterfactual.
- Select an ethical, feasible design and record its identification assumptions.
- Examine data quality, implementation and alternative explanations alongside outcomes.
An example in practice
A processing-time improvement coincides with a demand decline. Use a suitable comparison and process evidence before attributing the whole difference to the project.
Source & credit
Evaluation practice has multiple research traditions. This is an authoritative later guide, not a claim of sole invention.
This explanation and example are independently written. The source is a reading reference; linking it does not grant permission to reuse its text, figures or assessments.