Hey @eugenesmith
The event segmentation chart’s “Compare to Past” compares absolute numbers of event/user counts from a given day with another comparable granularity in the past. When trying to compare retention, we essentially compare the individual cohorts retention on how they are doing in their D1/W1/M1 relative to their acquisition dates.
So they should be interpreted slightly differently than a direct compare to past. The closest way to do this will be using the Change Over Time view in your retention chart. https://analytics.amplitude.com/demo/chart/new/l419dpf
Let me know if I have interpreted your use case correctly.