RANKX By Group In Measure for Push Dataset

MFelix - I think you've keyed in on what I'm looking for; the ultimate goal is to only show the latest set of rows for each site based on that MAX of that site's status date.

However, there are two issues with your proposed solution given the context of a push dataset:

1) You can't reference push dataset RealTimeData table columns directly outside of a row context; so only interative functions and filters have access to them.  Only measures can be referenced at the "top" level, which is indeed very frustrating.  There are many ways I can get a row context within a nested calculation, but the issue becomes surfacing the result to the top level.  E.g., calculate the max status date for a given site, compare each row's status date for that site against the site max date, then return a 1 or 0 based on whether there is a match.

2) The problem with the "based on selection" approach is that in when a map visual is used and it is first loaded, there is no selection context.  So nothing is "selected" and the map only cares about the fields you've used for latitude, longitude, color, size, etc..  This makes it very difficult to filter the map visual to only show sites based on their most recent data. Using something like a page-level filter basd on status date, say within the last 5 minutes doesn't work either because if the visual loads in the middle of a data refresh, some site;s status rows may not have been committed to the push dataset on the Power BI quite yet, which then causes the map to only show a subset of the sites.  This is why I need some sort of "sliding window with site context" solution.

I hope the above makes sense.  Thanks!

parry2k MFelix - Here's what's happening:

Refresh while dataset update in progress:

Refresh after dataset update complete:

The push dataset is updated every 5 minutes.  It takes approximately 1 minute for all new rows to be committed to the dataset, and each site's rows are pushed to Power BI in a single batch per site.  So if the user views the report at some point during that 1 minute of pushing data, some sites may still have the previous refresh/status date and therefore aren't shown in the visual as filtered using the provided sliding date window.

parry2k - The data *is* in the dataset.  Basically what I want to do is show the latest set of rows for a given site.  That's why I'm trying to find a way to have the visuals show the latest data *per site*.  This way if some site has been updated to a later date, the *latest* date of another site will still be shown.  This would ensure that the user didn't see the missing sites simply because only the latest date was being used across the entire dataset.

This also protects against a failure of the push dataset refresh process. If the process that pushes the data to the push dataset fails to run properly, the most recent successful data pushed will still be shown.  If you just go with a sliding date filter across the entire dataset (which is the way this is behaving), you'll show either an incomplete set of data (as shown above) or no data at all (in the case of a data push failure).

parry2k - I hear you, but we're not going in a circle.  It's not behaving the way you describe.  Do you have a push dataset you can test this against?  Perhaps you are testing this with a local data model vs a push dataset which has very different behavior and limitations with some calculations.

Here's what I'm using based on what was provided by MFelix :

Latest Site Refresh Date = MAXX(SUMMARIZE(ALLSELECTED(RealTimeData), RealTimeData[Site], RealTimeData[Status Date]), RealTimeData[Status Date])

Is Latest Site Data = IF(MAX(RealTimeData[Status Date]) = [Latest Site Refresh Date], 1, 0)