How can I get DAX to return just the record with a date closest to the slicer?

Question

I'm hoping someone can help as I've completely run out of ideas.

I'm working on performance reporting data, producing a number of visuals to summarise the most recent data. To allow users to retrospectively produce reports from previous quarters, I have added a date slicer as a way to "View data as at xxxx date".

Here's a rough representation of my data table - the due dates are in English format (dd/mm/yyyy):

The ratings are calculated in another system (based on a set of targets), so there are no calculated columns here. In reality, there are a lot more measures that report on different time periods (some weekly, some annually, etc) and there are different lags before the data is "due".

I eventually managed to get a measure that returned the latest actual:

MostRecentActual =
VAR SlicerDate = MAX ( Dates[Day] )
RETURN
    CALCULATE (
        SUM ( Data[Actual] ),
        Data[Data due] <= SlicerDate,
        LASTDATE ( Data[Data due] )
    )

I'm not completely sure I've done it right but it seems to work. I'd be happier if I understood it properly, so explanations or alternatives would be welcomed.

What I'm trying to do now is a basic summary pie chart at the beginning which shows the proportion of the measures that were red, amber, green or unrated as at the date selected. So I would need it to count the number of each rating, but only one for each measure and only for the date that is closest to (but before) the slicer date, which would vary depending on the measure. So using the above three measures, if the slicer was set to 10/10/2019 (English format - dd/mm/yyyy), it would count the RAGs for Q3 2019/20 for measures A an C and for Q2 2019/20 for measure B as there is a time lag which means the data isn't ready until the end of the month. Results:- A: Amber, B: Green, C:Red.

If I were able to create the measure that counted these RAGs, I would then want to add it to a pie chart, with a legend that is "Rating", so it would split the chart up appropriately. I currently can't seem to be able to do that without it counting all dates before the slicer (not just the most recent) or somehow missing ratings from the total for reasons I don't understand.

Any help would be very gratefully received.

Many thanks

Ben

---

Further update. I've been working on this for a while!

I have created a COUNTAX measure to try to do what I was wanting to do. In some circumstances, it works, but not all and not in the crucial ones. My measure is:

TestCountaxpt2 =
VAR SlicerDate = MAX ( Dates[Date] )
VAR MinDiff =
    MINX (
        FILTER (
            ALL ( Data ),
            Data[Ref] IN VALUES ( Data[Ref] ) &&
            Data[Data due] <= SlicerDate
        ),
        ABS ( SlicerDate - Data[Data due] )
    )
VAR thisdate =
    MINX (
        FILTER (
            ALL ( Data ),
            Data[Ref] IN VALUES ( Data[Ref] ) &&
            ABS ( SlicerDate - Data[Data due] ) = MinDiff
        ),
        Data[Data due]
    )
RETURN
    COUNTAX (
        FILTER ( Data, Data[Data due] = thisdate && Data[Ref] IN VALUES ( Data[Ref] ) ),
        Data[RAG]
    )

It produces the following table for a subset of the performance measures, which looks almost ok:

Table showing the result of the TestCountaxpt2 measure:

The third column is the measure above and it seems to be counting one RAG per measure and the dates look correct as the slicer is set to 3rd January 2020. The total for column 3 confuses me. I don't know what that is counting and I don't understand why it doesn't add up to 7.

If I add in the RAG column from the data table, it goes a bit more wrong:

Same table but with RAG Rating added:

The pie chart that is produced is also wrong. It should show 2 Green, 2 Red, 2 Grey (no rating) and 1 Amber. This is what happens.......

Pie chart for the DAX measure, with RAG Rating in the legend:

I can see what it is doing, which is to work out the most recent due date to the slicer in the whole table and using that (which is 1st Jan 2020) whereas I want it to calculate this separately for each measure.

Link to PBIX:

https://drive.google.com/file/d/1RTokOjAUADGHNXvZcnCCSS3Dskgc_4Cc/view?usp=sharing

Reworking the formula to count the ratings:

RAGCount =
VAR SlicerDate =
    MAX ( Dates[Day] )
RETURN
    COUNTAX (
        ADDCOLUMNS (
            SUMMARIZE (
                FILTER ( Data, Data[Data due] <= SlicerDate ),
                Data[Ref],
                "LastDateDue", LASTDATE ( Data[Data due] )
            ),
            "CountRAG", CALCULATE (
                            COUNTA ( Data[RAG] ),
                            Data[Data due] = EARLIER ( [LastDateDue] )
                        )
        ),
        [CountRAG]
    )

Here's the table it produces:

Answer 1

The reason for Total = 4 for the third column is straightforward. The SelectDate is maximal over all of the Refs in the table and there are only four Refs that match that date.

To fix this and get the totals you're after, you'll need to iterate over each Ref and calculate the SlicerDate for each independently and only then do your lookups or sums.

I haven't tested this code but it should give you an idea of a direction to try:

MostRecentActual =
VAR SlicerDate = MAX ( Dates[Day] )
RETURN
    SUMX (
        ADDCOLUMNS (
            SUMMARIZE (
                FILTER ( Data, Data[Data due] <= SlicerDate ),
                Data[Ref],
                "LastDateDue", LASTDATE ( Data[Data due] )
            ),
            "SumActual", CALCULATE (
                             SUM ( Data[Actual] ),
                             Data[Data due] = EARLIER ( [LastDateDue] )
                         )
        ),
        [SumActual]
    )

Going inside to outside,

1. FILTER the table to ignore any dates beyond the SlicerDate.
2. Calculate the LastDateDue for each Ref using SUMMARIZE.
3. Sum the Actual column for each Ref value using its specific LastDateDue.
4. Iterate over this summary table to add up SumActual across all Refs in the current scope.

Note that for 4, only the Total row in your visual will contain multiple Refs since the innermost Data table inside FILTER is not the entire Data table but only the piece visible in the local filter context.