Category: DAX

Nikola Ilic takes a look:

A few months ago, while scrolling through posts on social media, one of them immediately grabbed my attention! It was about a new tool, called DAX Optimizer, which promised to identify and remove performance bottlenecks in your DAX formulas. For all of us dealing with optimizing Power BI reports on a day-to-day basis, that was a huge promise (and one I was impatiently waiting to see in action).

One important note is that this is not a free tool, as Nikola mentions. Read on for more thoughts about how it works, what it picks up, and whether it’s a good fit for your environment given the price.

Comments closed

Marco Russo and Alberto Ferrari make a comparison:

DAX offers a rich set of functions, some of which overlap in their functionalities. Among the many, two functions perform grouping: SUMMARIZE and GROUPBY. These are not the only two: SUMMARIZECOLUMNS and GROUPCROSSAPPLY perform similar operations. However, the article is about SUMMARIZE and GROUPBY, as the other functions have many more functionalities, so a comparison would be unfair.

To make a long story short: GROUPBY should be used to group by local columns, columns created on the fly by DAX functions. SUMMARIZE should be used to group by model and query columns. Be mindful that both functions support both scenarios: both functions can group by model and local columns. However, using the wrong function translates into a strong decrease in performance.

Read on for a detailed explanation.

Comments closed

Gilbert Quevauvilliers pulls back the curtain:

I was once again working with the grouping feature in Power BI Desktop.

This time I had a lot of values that I wanted to put into multiple groups. I also did not want to select the individual values (I have previously shown this in my blog post Creating a group with all values (Even if not in the data) in Power BI)

It is amazing that sometimes it takes me a while to fully understand how to use some features in Power BI and today was one of those days!

Click through to see what Gilbert learned and how you can use that information to group by ranges instead of individual values.

Comments closed

Marco Russo and Alberto Ferrari do some slicing and filtering:

Slicers with too many values might be inconvenient for users, as they must search for the desired selection among too many lines. In such cases, a common solution is to build a hierarchy and use slicers with multiple columns inside, or multiple slicers, each with one column. However, this solution works only in structures with a natural hierarchy, like continents and countries. Indeed, each country belongs to only one continent so the hierarchy can be easily created with a new column.

If the hierarchy is non-natural, the relationship between the parent and the children is many-to-many, requiring a specific type of relationship. 

Click through to see what that relationship looks like and how you can build it.

Comments closed

Gilbert Quevauvilliers builds a data validation tool in Excel:

I had a requirement where the customer wanted to make sure that the right user could only see the colour from the stock item assigned to them. Essentially using the Excel file for Row Level Security.

For the Row level security to work correctly it required the colour name to be an exact match.

For me to ensure that this would happen, I wanted to make sure I could get the data from my Power BI dataset, and then using the Excel data validation make sure that the user could either type in the value 100% correct or pick it from the drop-down list.

There are quite a few steps, but I can see business people being happy about the final outcome.

Comments closed

Marco Russo and Alberto Ferrari break the ties:

DAX already offered two ranking functions so far: RANK.EQ and RANKX. While RANK.EQ is very seldom used, RANKX has been the primary ranking function for a long time. We wrote several articles about RANKX: one of them deals with ranking on multiple columns, a scenario that requires some DAX acrobatics to be solved as described in RANKX on multiple columns with DAX and Power BI.

The new RANK function makes ranking on multiple columns much easier because it offers sorting by multiple columns as a native feature, because it belongs to the family of window functions. Thanks to the concept of current row in window functions, RANK also helps in a subtle issue with RANKX – which despite being rare, might affect ranking on decimal numbers as described in Use of RANKX with decimal numbers in DAX.

Read on to see how you can use RANK() and how it compares to RANKX().

Comments closed

Allison Kennedy has a new title:

Today’s post is a short one with a video to demonstrate what I have learned. I’m a big believer in providing your end user with as much information as possible. Someone new should be able to pick up the report and understand it without needing to get a degree in Power BI or go digging through your company’s intranet to learn what’s what. 

One of the neat features that Power BI provides, is the ability for any visual to act as a filter or ‘cross-filter’ to other visuals. This provides us with comparative values and highlighted bars, but Power BI tooltip just says ‘highlighted’, which isn’t very helpful.

Click through for a bit of DAX code which can help with that, as well as a video demonstrating how it all comes together.

Comments closed

Leila Etaati smooths the curve:

The calculation will be done for all rows, till we don’t have the anymore data ahead ( 3 rows ahead) as you can see it will finished on row 2018 as 2018 is the average of 2018, 2019 and 2020.

so in the moving average we will miss some data point but replacing with average of data.

Less data point but help us to see a clear trend.

Click through to see how to do this. Moving averages is really good for the reports in which you want to get a rough magnitude but don’t care about exact values and don’t want oscillations to throw you off. For example, if I need to know how many seats to reserve for an event, I might use a moving average of the last several events as my baseline, rather than the prior event’s number.

Comments closed

Marco Russo and Alberto Ferrari build a regression:

LINEST and LINESTX are two DAX functions that calculate a linear regression by using the Least Squares method. Both functions return multiple values, represented in a table that has a single row and one column for each of the values returned.

LINEST gets column references as arguments, whereas LINESTX explicitly iterates over the table provided in the first argument and executes the other arguments in a row context. Internally, LINEST invokes LINESTX and provides to it the table that contains the column references specified in the LINEST arguments. This article describes the more generic function LINESTX.

This is pretty neat for adding a visual element, though I’d probably include an R or Python visual and do the regression in there, myself.

Comments closed