Can I just point out to anyone from the QuickSight team just how badly this cripples the running* functions… surely the whole point of these is NOT to have to mess around in SQL to get a simple consistent running total.
Using these with a date range is such an obvious use case and you can’t expect every date bucket in that range not to have some sparse coverage especially if you are further segmenting it. Think first couple of days of the month when something hasn’t happened yet.
I totally get the difficulty of working with sparse data sets behind the scenes but not having to resort to JOINing in a full calendar date range into your data each time just to get a simple running total analysis would be such a huge win for this tool. When I’m having to do that I might as well just go and do the totting up in SQL anyway.