CANS Analysis Report & Methodology

Sharing a walkthrough of our CANS Analysis Report and the methodology behind it.

The short version: traditional p-value testing was not built for the realities of human services caseloads — small samples, structural attrition, and the need to answer "did this client improve?" rather than "was the group average difference unlikely to be chance?" The report uses effect sizes (Cohen's d) for magnitude and the Reliable Change Index for individual-level change against measurement error. It scales from items to domains to composites, with a companion module (CANS-EC) for younger clients.

For agencies: defensible outcomes for funders and boards.

For analytics peers: every threshold and calculation is inspectable in the model.

We are curious to hear how others are moving past p-value defaults in evaluation work. What has worked for you?