cassionData Analysis

Project · Advanced

Protection referral pathway performance

Finding where a referral pathway loses people, using only the fields a performance analysis actually needs — and treating the restraint about what not to collect as part of the deliverable.

PythonRCore Humanitarian Standard (CHS)Sphere Standards

Context

Organisation
Protection and GBV programme across six admin2 areas
Problem
Fewer than half of the people who consented to a referral reached a service, but the pathway was reported as a single completion rate. A single number cannot say whether the failure is at intake, at issuing the referral, or at the receiving provider, so every corrective action taken in the previous year had been aimed at the wrong node.
Audience
Protection cluster coordinator and the GBV sub-cluster
Decision it informs
Which service line and which area receive additional capacity, and whether the disability gap needs a separate response.

Datasets used

Deliverables

What you do not collect is part of the design

This project’s first deliverable is the data protection note, and it is not paperwork. The dataset holds no names, no free text, no incident date, no location below admin2, no exact age, and no incident type — and the analysis still answers every question the cluster asked. Demonstrating that is the point.

Under GBV information management principles, incident-level detail is never shared outside the case management agency. A pathway performance analysis does not need it. Holding it anyway creates risk with no analytical return, and that trade is the one this sector gets wrong most often.

A completion rate is three rates

Consent, referral issued, referral accepted. Collapsing them hides which step fails, and consent is not a step that fails at all — a person who declines referral has exercised a choice, and counting them as a pathway failure both misstates performance and misrepresents that decision. The completion denominator is consenting cases.

Decomposed, the pathway is not uniformly weak: it runs from about 23% for livelihood support to 62% for health, and from 31% to 57% across areas. Those are different problems with different fixes.

The finding that matters most is the one about equity

Cases where a disability was reported complete at roughly 31% against 48% where none was. Reaching that finding requires handling a field that one area coded differently and that is blank for sixty-two cases — which is exactly how equity gaps stay invisible in real reporting.