Start with unique people
In a hypothetical month, 100 people are referred, 80 are contacted, 60 enroll, and 45 attend. Reporting only 100 referrals overstates completed access. Reporting only 45 attendees hides where the losses occurred.
Repeated referrals for the same person should not silently inflate the first stage.
Use both overall and stage-specific proportions
Attendance is 45% of referrals in this example, and 75% of enrollees. Both are correct; they answer different questions.
| Transition | Question to investigate |
|---|---|
| Referred to contacted | Were contact details usable and outreach feasible? |
| Contacted to enrolled | Did eligibility, cost, or preference affect the decision? |
| Enrolled to attended | Did scheduling or other barriers arise? |
Do not label every loss a failure
Someone may decline because the program does not fit their goals or they found another service. Record reasons respectfully where feasible, without pressuring participation.
Align the time window
Recent referrals may not yet have had time to attend. Mixing them with older completed referrals can make the cascade appear worse. Use defined cohorts or appropriate follow-up periods.
The cascade is useful when it leads to a specific process improvement, not merely a funnel chart that assigns blame to people at the bottom.
