PBIRS report does not match PBI Desktop report

Viewed 125

I have a dashboard published to PBIRS. The dashboard contains a few measures. When viewing a visual on the dashboard, the tooltip displays the correct value for the measure in the visual. When viewing the visual "as a table", the measure is incorrect (extremely, obviously incorrect). When viewing the table in PBI Desktop, the correct value is displayed. An example follows. Headers and axes omitted so as to protect the innocent (me). table viewed from PBIRS via browser (incorrect measure values) table viewed from PBI Desktop (correct measure values) As you can see it appears the PBIRS is calculating the table value incorrectly. That said, the visual representation of the measure (a line chart) displays the correct value. Can anyone give me a clue as to what is going on here? Thanks Jim

1 Answers

below is part of an exchange with Microsoft on this issue:

I was thinking about this a bit last night and came up with some useful information to debug what’s going on here a little. I ran the performance analyzer in PBI Desktop and captured the DAX that was generating the data for the visual in question. Here is the DAX:

// DAX Query
DEFINE
  VAR __DS0FilterTable = 
    TREATAS({"2021-07-20"}, 'dim_GPS_OSR_Date'[Slicer Date])

  VAR __DS0FilterTable2 = 
    FILTER(
      KEEPFILTERS(VALUES('dim_Division'[Division])),
      NOT(ISBLANK('dim_Division'[Division]))
    )

  VAR __DS0FilterTable3 = 
    FILTER(
      KEEPFILTERS(VALUES('dim_GPS_OSR_Date'[GPS OSR Date])),
      AND(
        'dim_GPS_OSR_Date'[GPS OSR Date] >= DATE(1938, 6, 20),
        'dim_GPS_OSR_Date'[GPS OSR Date] < DATE(2021, 10, 20)
      )
    )

  VAR __DS0Core = 
    SUMMARIZECOLUMNS(
      'dim_Report_Date'[Year-Month],
      __DS0FilterTable,
      __DS0FilterTable2,
      __DS0FilterTable3,
      "LBM_Conversions", 'fact_Conversions'[LBM Conversions],
      "Sales_Conversions", 'fact_Conversions'[Sales Conversions],
      "LBM_Percent_Converted", 'fact_Conversions'[LBM Percent Converted],
      "Sales_Percent_Converted", 'fact_Conversions'[Sales Percent Converted],
      "Conversions", IGNORE('fact_Conversions'[Conversions]),
      "Online_Conversions", 'fact_Conversions'[Online Conversions],
      "Online_Percent_Converted", 'fact_Conversions'[Online Percent Converted]
    )

  VAR __DS0PrimaryWindowed = 
    TOPN(1001, __DS0Core, 'dim_Report_Date'[Year-Month], 1)

EVALUATE
  __DS0PrimaryWindowed

ORDER BY
  'dim_Report_Date'[Year-Month]


// DAX Query
//DEFINE
//  VAR __DS0FilterTable = 
//    TREATAS({"2021-07-20"}, 'dim_GPS_OSR_Date'[Slicer Date])
//
//  VAR __DS0FilterTable2 = 
//    FILTER(
//      KEEPFILTERS(VALUES('dim_Division'[Division])),
//      NOT(ISBLANK('dim_Division'[Division]))
//    )
//
//  VAR __DS0FilterTable3 = 
//    FILTER(
//      KEEPFILTERS(VALUES('dim_GPS_OSR_Date'[GPS OSR Date])),
//      AND(
//        'dim_GPS_OSR_Date'[GPS OSR Date] >= DATE(1938, 6, 20),
//        'dim_GPS_OSR_Date'[GPS OSR Date] < DATE(2021, 10, 20)
//      )
//    )
//
//  VAR __DS0Core = 
//    SUMMARIZECOLUMNS(
//      'dim_Report_Date'[Year-Month],
//      __DS0FilterTable,
//      __DS0FilterTable2,
//      __DS0FilterTable3,
//      "LBM_Conversions", 'fact_Conversions'[LBM Conversions],
//      "Sales_Conversions", 'fact_Conversions'[Sales Conversions],
//      "LBM_Percent_Converted", 'fact_Conversions'[LBM Percent Converted],
//      "Sales_Percent_Converted", 'fact_Conversions'[Sales Percent Converted],
//      "Conversions", IGNORE('fact_Conversions'[Conversions]),
//      "Online_Conversions", 'fact_Conversions'[Online Conversions],
//      "Online_Percent_Converted", 'fact_Conversions'[Online Percent Converted]
//    )
//
//  VAR __DS0PrimaryWindowed = 
//    TOPN(1001, __DS0Core, 'dim_Report_Date'[Year-Month], 1)
//
//EVALUATE
//  __DS0PrimaryWindowed
//
//ORDER BY
//  'dim_Report_Date'[Year-Month]
//

Note that the same query was generated twice, which seems odd, but it’s not necessarily indicative of the problem with the table output per se.. Commenting out the second query and running the first produced the results attached. Note that the % conversion columns are in fact correct in the query output. This leads me to conclude there is a bug in the component that renders the table in the visual, at least in the May 2021 release of Power BI. Thought this might be helpful to share. We are still planning on upgrading to the Sep 2021 version of PBI Desktop and PBI RS in light of this issue and will reach out once that is complete and we have something to compare.

following up: Installing the Sep 2021 version of PBI Desktop and PBI RS solved the issue. This case can be closed.

Related