A screen-failure waterfall only helps when the categories match how your medical monitor thinks about eligibility—not how your EDC happened to label fields six months ago.
Start with the denominator
Every bar should state clearly whether the denominator is "patients screened" or "patients who signed consent." Mixing those mid-chart is the most common reason ops teams argue over percentages in standup.
Look for category drift after amendments
When inclusion lab thresholds change, reason codes often stay the same while the underlying population shifts. Ask your analyst to mark amendment effective dates as vertical reference lines. Without them, a declining lab-failure bar might look like site improvement when it is actually criterion relaxation.
Site-level overlays belong in the appendix
The main waterfall should stay at study level for steering readability. Site comparisons belong in a separate appendix table so CRAs can drill down without cluttering the committee slide.
Three footnotes we refuse to omit
Source file name and extraction timestamp, the mapping table from EDC codes to display categories, and a count of records excluded for missing reason codes. If any of those three is absent, treat the chart as draft.