When feasible, we apply rigorous impact designs such as randomized controlled trials (RCTs), difference-in-differences, and propensity score matching to isolate program effects.
These methods help us separate causality from correlation and offer the highest level of confidence in attributing results.
When randomization isn’t possible, we use quasi-experimental methods with careful matching and baseline controls. These designs ensure reliability while remaining ethical and context-sensitive. We always balance statistical rigor with practical constraints to deliver credible and actionable results.
