It is odd that they don’t normalize the width of the dozens of 3 cohort graphs. Apparently in order to show fully filled rows.

But it dramatically blunts the visual clarity of comparison between the differing percentages in each cohort associated with better and worse outcomes.