Forum Discussion

lauriedata's avatar
lauriedata
Resolver I
8 months ago
Solved

rank() function not working as expected (NOT USING RANKX)

Hello.  Whether or not I include "DENSE" in the formula, the result is identical.  The rank() function is giving me ties with skips in the numbering.  The rank is based on the total value of reward WITHIN THE DATE RANGE, NOT FOR A SINGLE DATE.    The ranking ORDER is correct, but I would expect the numbers to be 1,1,3,3,5 or if I use dense, to be 1,2,3,4,5 etc.  Neither is happening.

 

 

Here are the formulas:

Rank_TopN Reward Value =
var _TopN = selectedvalue('TopN'[Top N]) /* user chooses rank level to view */
Return
if(HASONEFILTER(v_flt_ci_cs_adjustment_rwd[Loyalty ID]),
switch(true(),rank(all(v_flt_ci_cs_adjustment_rwd[Loyalty ID]),orderby([CS Adjust RW],desc))<=_TopN,
rank(all(v_flt_ci_cs_adjustment_rwd[Loyalty ID]),orderby(v_flt_ci_cs_adjustment_rwd[CS Adjust RW],desc))))

NOTE: CS Adjust RW = calculate (sum(v_flt_ci_cs_adjustment_rwd[sum_reward_value]), allexcept(v_flt_ci_cs_adjustment_rwd, v_flt_ci_cs_adjustment_rwd[Loyalty ID])).

==========================

If I do the same formula for another field/table for points, it works out fine, so I cannot figure out what I am doing wrong. 

Here is the points result and formula that all seem to work as expected.

Rank_TopN Points = 
var _TopN = selectedvalue('TopN'[Top N]) /* user chooses rank level to view */
Return
if(HASONEFILTER(v_flt_ci_cs_adjustment_points[Loyalty ID]),
switch(true(),rank(all(v_flt_ci_cs_adjustment_points[Loyalty ID]),orderby([CS Adjust Points],desc))<=_TopN,
rank(all(v_flt_ci_cs_adjustment_points[Loyalty ID]),orderby([CS Adjust Points],desc))))
 
NOTE:  CS Adjust Points = 
        CALCULATE(
SUM(v_flt_ci_cs_adjustment_points[points_earned]),
ALLEXCEPT(
v_flt_ci_cs_adjustment_points,
v_flt_ci_cs_adjustment_points[Loyalty ID]
)
)
Need Help
  • MarkLaf's avatar
    MarkLaf
    8 months ago

    Thanks for sharing the pbix as that makes it much easier to troubleshoot.

     

    The reason that the measure is not working in your pbix is because of the [Rank_TopN Reward Value] is not blank filter on the visual. This adds a value filter to the visual's SUMMARIZECOLUMNS, which does not interact well with the ADDCOLUMNS I had originally used. We can circumvent this issue if we use SUMMARIZECOLUMNS instead. This worked for me with the is not blank filter in your shared pbix.

     

     

    Rank_TopN Reward Value fixed = 
VAR _topN = SELECTEDVALUE('TopN'[Top N])
VAR _core = 
    SUMMARIZECOLUMNS( 
        v_flt_ci_cs_adjustment_rwd[Loyalty ID],
        ALLSELECTED( 
            v_flt_ci_cs_adjustment_rwd[Loyalty ID], 
            v_flt_ci_cs_adjustment_rwd[reward_issue_date] 
        ), 
        "@selectedSum", CALCULATE( SUM( v_flt_ci_cs_adjustment_rwd[sum_reward_value] ) )
    )
VAR _rank = RANK( DENSE, _core, ORDERBY( [@selectedSum], DESC ) )
RETURN
IF( _rank > 0 && _rank <= _topN, _rank )

     

     

     

     

12 Replies

  • Ashish_Mathur's avatar
    Ashish_Mathur
    Super User
    8 months ago

    Hi,

    Share some data to work with and show the expected result.  Share data in a format that can be pasted in an MS Excel file.  Show the expected result there.

  • v-hashadapu's avatar
    v-hashadapu
    Community Support
    8 months ago

    Hi lauriedata , Thank you for reaching out to the Microsoft Community Forum.

     

    The ranking breaks because the rewards table is still producing multiple visible rows per Loyalty ID (one per date) and the RANK function always ranks each row of the table it is given. Even though your measure [CS Adjust RW] correctly returns the total reward per Loyalty ID, the visual’s row context forces RANK to evaluate that same total repeated across several rows, which leads to tied values but skipped rank numbers. That is why DENSE has no effect. The function is not ranking a clean, one-row-per-ID table. Your points table works only because its structure naturally gives one row per ID during ranking.

     

    You must force RANK to operate on a distinct list of Loyalty IDs, not the visual rows. Example:
    Rank_TopN Reward Value =
    VAR _TopN = SELECTEDVALUE('TopN'[Top N])
    VAR _Rank =
    RANKX(
    ALL(v_flt_ci_cs_adjustment_rwd[Loyalty ID]),
    CALCULATE([CS Adjust RW]),
    ,
    DESC,
    DENSE
    )
    RETURN
    IF(_Rank <= _TopN, _Rank)

    • lauriedata's avatar
      lauriedata
      Resolver I
      8 months ago

      Hi.  Thank you, I tried that, but the formula with rankx  with dense gave the exact same result as my rank.  1,1,4,4, etc.  When I refreshed the data, it also threw a cyclic error.  I'll keep working on this.  Thanks again.

      • v-hashadapu's avatar
        v-hashadapu
        Community Support
        8 months ago

        Hi lauriedata , Thank you for reaching out to the Microsoft Community Forum. 

        Please try this:
        Rank_TopN Reward Value =
        VAR _TopN = SELECTEDVALUE('TopN'[Top N])
        VAR RankTable =
        ADDCOLUMNS(
        VALUES(v_flt_ci_cs_adjustment_rwd[Loyalty ID]),
        "TotalRW", CALCULATE([CS Adjust RW])
        ) VAR _Rank = RANKX(RankTable, [TotalRW], , DESC, DENSE)
        RETURN IF(_Rank <= _TopN, _Rank)

         

        If you still see 1,1,4,4 or get a cyclic error, that strongly indicates a dependency loop, make sure [CS Adjust RW] does not reference any ranking measures or any measure that in turn references Rank_TopN. Also check for hidden decimal differences (use ROUND([CS Adjust RW],2) in the TotalRW column while testing).

         

        If that still doesn’t work, please provide me with a short sample data (preferably not as images) including your measures and all the relevant details and I will check it to see how I can help you.

        Need help uploading data? https://community.fabric.microsoft.com/t5/Community-Blog/How-to-provide-sample-data-in-the-Power-BI-...