The New Year is almost upon us, meaning its time to put in place some resolutions for the year ahead. I can think of no better commitment then to learn more about Power BI in 2019, which is hopefully the reason why you are reading this right now šŸ™‚ . Welcome to the eighth post in my series concerning Microsoft Exam 70-778, where I hope to provide a learning/revision tool for anyone who is taking the exam or looking to increase their Power BI expertise. Last week, we investigated how to manage Key Performance Indicator (KPI) reporting using Power BI. We now move into another topic area that is also tied closely to visualizations – Create hierarchies. The related skills for this exam area are:

Create date hierarchies; create hierarchies based on business needs; add columns to tables to support desired hierarchy

In the sections that follow, I will provide an overview of the two different types of Hierarchies configurable within Power BI, before then providing a step-by-step guide on how to create one. As with previous posts in this series, this subject does bring together some other skill areas, such as importing data with Power Query and using DAX calculated columns. Knowledge of these topics will hold you in good stead when starting with hierarchies.

What is a Hierarchy?

Hierarchies form the cornerstone of a broad aspect of human existence, which makes it natural they are a common theme when it comes to Business Intelligence (BI) solutions. A hierarchy is definable as a logical, often visual, representation of an order of precedence. Hierarchies, in strict Power BI terms, do not differ much from this general definition; they allow developers to order data by preference, priority or anything in between. When configured and used in isolation, they offer minimal benefit to Power BI report consumers. They start to become really valuable when utilised alongside visualizations, providing an additional interaction point for key data points and allowing report users to tailor a visualization to suit their needs. Some potential usage scenarios for hierarchical data may include:

  • Providing top-level and subcategory classifications for product records.
  • Modelling a business reporting line hierarchy, from CEO down to Developer.
  • Splitting sale date by quarter, year or even month.

Hierarchies come in two flavours within Power BI – Date Hierarchies and Custom (or User Defined) Hierarchies. The next two sections delve deeper into the inner workings of each.

Date Hierarchies

Date Hierarchies are best described as a logical breakdown of the constituent components of a date value, based on four different levels:

  • Year
  • Quarter
  • Month
  • Day

For each field that contains a Date Hierarchy, the appropriate value for each of the above is exposed out for utilisation across Power BI Desktop. So, for example, the underlying hierarchy values for the 3rd March 2017 would be:

  • Year: 2017
  • Quarter: Qtr 1
  • Month: March
  • Day: 3

All columns with a data type of Date or Date/Time will have a Date Hierarchy created for them automatically. The hierarchy will then become accessible for use in the following areas of Power BI Desktop:

  • When building out formulas using DAX. For example, the screenshot below shows how it is possible to access the hierarchy field values listed above from the Date/Time field LastEditedWhen:
  • When working with a Table visualization. Adding a Date or Date/Time column into the Values well will automatically create a table containing all of the hierarchy fields. An example of how this looks for the LastEditedWhen field is seen in the image below:

It is possible to disable the automatic creation of Date Hierarchies within the File -> Options area of Power BI Desktop, by toggling the Auto Date/Time option. Take care when toggling this though, as doing so will remove any existing Date Hierarchies within your model:

This fact is confirmed when returning to the Table created using the LastEditedWhen field, which automatically reverts to the base column value when theĀ Auto Date/Time option is disabled:

Custom Hierarchies

For more bespoke requirements, a Custom Hierarchy affords the same kind of functionality discussed in the previous section. They can be created from the right-hand Fields pane by selecting a field within a Power BI table and choosing the New hierarchy option:

Once established, you can then:

  • Rename the hierarchy. By default, it will be named using the convention <Field Name> Hierarchy.
  • Include additional fields by dragging and dropping them into the hierarchy.
  • Reorder the hierarchy, by dragging and dropping the fields into your preferred order, from top to bottom.
  • Delete the hierarchy through right-clicking it and selecting the Delete option. All fields included in the Hierarchy will remain as part of the underlying query.

A Custom Hierarchy typically relies on a self-relationship that a query might have on itself. An excellent example of this, already alluded to, is an Employee table, where an individuals Manager is trackable via a ManagerID column, that resolves back to an EmployeeID record. In these cases, there are several DAX functions available that will assist in getting custom columns set up to support your hierarchy, described collectively as Parent and Child Functions. Usage of these functions is essential when building out Hierarchies involving parent/child relationships.

TheĀ exercise at the end of this post will go through the detailed steps involved in creating a Custom Hierarchy.

Using Hierarchies with Visualizations

A good reason to include hierarchies as part of your Power Bi Reports is the options they unlock from a data drill-down perspective. We’ve seen already the behaviour Date Hierarchies adopt when using the Table visualization, which offers nothing in this respect. Other visualization types support, by comparison, far richer options for “homing in” on a particular piece of data, epitomised by the following buttons that appear at the top of the supported visualization:

In order, from left to right, these let you:

  • Drill up your data one level in the hierarchy.
  • Enable the Click to turn on Drill Down feature, allowing you to click on any part of a visual to drill-down to the next level of the hierarchy.
  • Go to the next level in the hierarchy.
  • Expand all down one level in the hierarchy. This behaves differently to the Click to turn on Drill Down option mentioned already, by grouping together the previous hierarchy level each time you drill-down.

An example of how all this works as part of a Pie chart visualization is viewable in the sequence below:

As should hopefully be clear, hierarchies are very much the icing on the cake in building an engaging and wholly interactive reporting solution within Power BI.

Example: Creating a Hierarchy and adding it to a Visualization

The steps that follow provide instructions on how to build out a product categorisation hierarchy, using sample data, and then how to apply this to a Donut chart visualization. All that is required to work through these steps is access to the Power BI Desktop application:

  1. In Power BI Desktop, click onĀ Edit Queries to open the Power Query Editor. Navigate to the New Source button and select the Blank Query option:
  2. Select the newly created Query1 in the Queries pane and click the Advanced Editor button. Copy and paste the following M code into the window and press Done:
    1. let
          Source = Table.FromRows(Json.Document(Binary.Decompress(Binary.FromText("nZNNbsMgEIWvgrxqJRb438mybZZRonrpZoGScYNiYwvsqu11epOerBjstHKwE3WBhJj3vjdoIMsc18HO4/MabXjBOCzRijcgasEkoDtJy7qA+5sUgVrfXwEOCXF2OHO8kWsrqhykZBWnxRR5SmPYHo6jRLP9kS9tKD9QcZji2uqGGeE4cTUz6D0piDe2v7i85ThSS1tDtVm974+Uv0Kf+Vc3Uzojus0mz7sEPwptafPVgROPlC1rJinj2sBIfnVPIE9NVVsIlsrgX1yq8kqgNTtI9gnopSXEiyZe0X99iZ5mgheumaZL7LC0pEWBHlqpBiGlenBQsra80sF1k4kPfM+Ed38qPVIB24rxxjL32WKkYS4h2Pf623SfKT191KD7OXdiAd8iMgEhDn2F3/0A", BinaryEncoding.Base64), Compression.Deflate)), let _t = ((type text) meta [Serialized.Text = true]) in type table [ProductID = _t, Name = _t, #"Product ID" = _t, Parent = _t, TotalSales = _t]),
          #"Changed Type" = Table.TransformColumnTypes(Source,{{"ProductID", Int64.Type}, {"Name", type text}, {"Product ID", type text}, {"Parent", Int64.Type}, {"TotalSales", Currency.Type}}),
          #"Replaced Value" = Table.ReplaceValue(#"Changed Type",null,0,Replacer.ReplaceValue,{"TotalSales"})
      in
          #"Replaced Value"
      
  3. This will then load up a table object that resembles the below, which derives from the sample Product data from Dynamics CRM/365 Customer Engagement:
  4. Right-click the Query1 query and use theĀ Rename option to change its name to Products. Select the Close & Apply button to exit the Power Query Editor.
  5. Open the Data pane, select the Products query and use the following DAX formulas to create New Columns for the table:
    • ProductHierarchy = PATH(Products[ProductID], Products[Parent])
    • Product Category = LOOKUPVALUE(Products[Name], Products[ProductID], PATHITEM(Products[ProductHierarchy], 1, INTEGER) )
    • Product Grouping = LOOKUPVALUE(Products[Name], Products[ProductID], PATHITEM(Products[ProductHierarchy], 2, INTEGER) )
    • Base Product = LOOKUPVALUE(Products[Name], Products[ProductID], PATHITEM(Products[ProductHierarchy], 3, INTEGER) )
  6. The first field creates the required hierarchy list in a delimited format; the remaining columns then list out the three levels of the hierarchy as separate column values. The data in your table should resemble the below if done correctly:
  7. Right-click the Product Category field in the Fields pane and select New Hierarchy, to create a new hierarchy as indicated below:
  8. Modify the newly created hierarchy so that:
    • It is named Product Hierarchy
    • It contains, in addition to the Product Category field, the Product Grouping and Base Product fields.
    • The order of the newly added fields reflect the following:
      • Product Category
      • Product Grouping
      • Base Product
  9. If done correctly, the Hierarchy should resemble the following:
  10. Click on the Report tab and add an empty Donut chart visualization to the report. Populate the Field well values as shown below:
  11. The Donut chart should update accordingly, and with the appropriate drill-down options available at the top of the visualization. Notice also that hovering over each section of the Donut chart displays an aggregated total for that level of the hierarchy:

Key Takeaways

  • Hierarchies within Power BI provide a means of logically categorising data into an order of preference or precedence, providing greater flexibility to Power BI report users when they interact with visualizations.
  • Date Hierarchies are created and managed automatically by Power BI for each Date or Date/Time field defined within your model. These automatically create fields that contain the Year, Quarter, Month & Day values from the respective date fields. These fields can then be utilised as part of a Table visualization or within a DAX formula. Date Hierarchies can also be completely disabled if required.
  • Custom (or User-Defined) Hierarchies need to be created manually and provide additional customisation options when compared to Date Hierarchies, such as the number of fields they contain, the order and its name. A Custom Hierarchy will typically make use of one of several Parent/Child DAX functions, such as PATH or PATHITEM.
  • Including a hierarchy as part of a chart visualization, such as a Pie chart or Donut chart, opens up other drill-down capabilities around your data. Indicated by the additional arrow icons included at the top of the visualization, they provide the means for users to interrogate data points that interest them the most straightforwardly.

The past couple of posts have already started to introduce some of the visualization options available within Power BI. In next week’s post, we will take a much closer look at all of the options available within the application to display data, as well as look at features such as Bookmarks and the various report page customisation tasks available to developers when building out a report.

As we move into the festive period, now is the time to put your feet up, relax, take stock for the year ahead…or, if you are reading this over Christmas, grab the opportunity to learn more about Power BI šŸ™‚ . In the seventh post in my series concerning Microsoft Exam 70-778, we move away from the suitably large subject area of DAX into a topic much more visually focused and accessible – Measure performance by using KPIs, gauges and cards. The related skills for this exam area are:

Calculate the actual; calculate the target; calculate actual to target; configure values for gauges; use the format settings to manually set values

As part of this week’s post, I hope to delve deeper into each of these skill areas, with the aim of providing a revision tool for the exam or a general learning aid for those getting to grips with Power BI for the first time. To follow on as part of any examples that follow, make sure you have downloaded the WideWorldImporters sample database and hooked up Power BI Desktop to the tables described in my first post in this series.

KPI Overview

A standard reporting requirement, and one which typically forms the basis of any contractual performance monitoring is the studying of Key Performance Indicator (KPI) figures. These are designed to provide an executive level view to answer questions such as “How is our sales pipeline progressing?” or “Have we met our objectives to reduce the number of monthly complaints down to X number?“. The whole process of collating together diffuse data sources to present this information may, in the past, have been exhausting. With Power BI, it is possible to deploy a dedicated KPI visualization that allows you to straightforwardly and consistently report headline figures, track variance over time and receive immediate visual cues to flag up any potential issues.

To get going with your first KPI visualization, you must supply two critical pieces of data:

  • Indicator: This is the “actual” value that needs reporting against a target, and will typically be some form of aggregation contained within a Measure.
  • Trend axis: Representing the performance of your indicator based on a range of data, the most typical axis type to use here would be some form of date value – such as month or year.

A KPI configured with just these two properties will work but looks…well…bland and not very useful at all:

This is why a Target Goal should additionally be specified. Adding this will allow for the appropriate colour to be applied to the visualization, to indicate how well the Indicator is performing against your target. The great thing about this is that up to 2 Target Goal values can be specified if needed. You potentially lose some functionality here, as the Distance percentage value will no longer display correctly; but this may be useful if, for example, you have an aspirational target that there would be a benefit in hitting, but is not a strict requirement. When the value of one Target Goal is determined to be fulfilled, but the other isn’t, the Visualization will display Amber by default, as indicated below:

The example at the end of this post will walk through the steps involved in setting up a KPI.

Gauge Visualizations

We got a sneak peek into Gauge’s when working with What-If Parameters in last week’s post, but a more in-depth discussion regarding them is necessary. Gauges provide a different means of viewing how a subset of data is performing against a target. They differ from KPI’s primarily due to the additional configuration options they afford and the potential they have to provide a faster means of interacting with key data points on the visualization. The configurable field well properties for a Gauge is viewable below, in the screenshot and descriptive bullet point provided:

  • Value: This is the value that requires tracking, similar to the Indicator within KPI’s.
  • Minimum value: The minimum amount that needs to be hit by the Value before the Gauge will be filled in.
  • Maximum value: The maximum amount that needs to be hit by the Value before the Gauge stops filling (i.e. a Gauge with a Value of 105 and a Maximum value of 100 would display as full).
  • Target value: The target that needs hitting, which will be presented as a black line on the Gauge, as shown below:
  • Tooltips: These are additional field values that will be displayed to the user when hovering over the Gauge. These fields can either be aggregated variants of columns or a Measure.

Gauges behave differently from KPI’s in that they will “work” with only one field well specified. For example, adding a Measure with a value of 8 million as the Value will display a half-full Gauge, as indicated below:

Notice that, in this scenario, the Gauge assumes the maximum amount to be double of what is specified in the Value field well – in this case, 16 million. A Gauge configured so frugally has little purpose and you should, as a minimum, ensure that the Maximum value and Target value field wells are also set, as demonstrated below:

A nice little feature of Gauges is the ability to specify any of the potential field well values manually, within the Formatting options (discussed in the next section).Ā  TheĀ Value field well must be populated already within your Gauge to support this feature. You will then have the ability to manually specify the other values within the Gauge axis options area:

Any manually defined figure will be automatically overwritten and deleted if you choose to use a field from a table instead, so take care here to avoid any data loss.

Gauges address a similar need to KPI’s, in that they allow you to view possible progress against a maximum or target value. A significant downside in using them is that they present a less visual means of showcasing potential issues with your data. You should weigh these advantages/disadvantages up when determining whether to use a Gauge or a KPI as part of your report.

Visualization Format Settings

All Visualization’s within Power BI can be customised, often to the extent where they can look indistinguishable from when you first add them onto your report. The Formatting tab on the right-hand pane is your go-to destination in this regard, and is accessible by clicking the paintbrush icon while selecting any Visualization:

The types of things that are achievable via these options include, but are not limited to:

  • Modifying all aspects of the visualizations Title, such as its visibility, the text displayed, font colour/size/type and alignment.
  • Adding a colour background to the visualization.
  • Including a description for the visualization.
  • Toggling a border for the visualization and its colour.
  • Fine-tuning the exact location of the visualization on the report, based on its X & Y Position values.

Each Visualization will also have a set of specific, unique options. We can see an example of how this looks by going back to KPI’s again and examining the top four options listed:

  • Indicator: This area provides options on how to display the Indicator field value on the KPI. It is possible to round up figures to the thousands, millions and even trillions, and also to adjust the number of decimal places shown. Auto is the default option used and will generally assign the best rounding option, depending on the underlying value.
  • Trend Axis: Here, it is possible to enable/disable the trend visual on the KPI. The example below demonstrates how this looks when disabled:
  • Goals: With the Goal toggle, it is possible to remove the Goal figure underneath the Indicator value completely; the Distance toggle will remove the percentage distance value as well. It is possible to mix and match options here, as indicated in the example below:
  • Color coding: The default RAG (Red, Amber, Green) traffic light system should be suitable for most situations but, if you have a specific branding requirement, it is possible to override the Good, Neutral and Bad colours with a custom Hex colour value. The Direction option also lets you reverse how the KPI works. So, for example, if raising more than 5 IT service requests would lead to a Bad outcome, any number underneath this would appear as Good instead.

When it comes to this subject area for the exam, there is no need to study the whole range of formatting options available; however, a general awareness will hold you in good stead.

Example: Creating a KPI

The process involved in building out a KPI can be a lot more involved then you may suspect, based on the descriptions provided so far. Therefore, the example that follows is designed to show you the type of preparatory modelling steps that will be needed to put in place a working solution utilising KPIs. To follow through these steps, make sure you have connected Power BI up to the WideWorldImporters sample database and that the Sales.OrderLines table is within your Power BI data model:

  1. In Power BI Desktop, on the Sales OrderLines table, use the New Column button to add on the following new calculated columns:
    • GrossPrice = ‘Sales OrderLines'[UnitPrice] * ‘Sales OrderLines'[Quantity]
    • NetPrice = ‘Sales OrderLines'[GrossPrice] + (‘Sales OrderLines'[GrossPrice] * (‘Sales OrderLines'[TaxRate] / 100))
    • OrderDate = RELATED(‘Sales Orders'[OrderDate])
  2. Next, create a Calculated Table via the New Table option, that uses the following DAX formula:
    • SalesOrderAgg2016 = FILTER(SUMMARIZE(‘Sales OrderLines’,’Sales OrderLines'[OrderDate], “Total Gross Price”, SUM(‘Sales OrderLines'[GrossPrice]), “Total Net Price”, SUM(‘Sales OrderLines'[NetPrice])), ‘Sales OrderLines'[OrderDate] >= DATE(2016, 1, 1) && ‘Sales OrderLines'[OrderDate] <= DATE(2016, 12, 31))
  3. Doing this will create a new table object that looks similar to the below, providing a total sum of the calculated columns created in step 1), grouped by OrderDate. I would also recommend, at this stage, to change the format of the new columns to your preferred Currency value:
  4. Add a new Calculated Column to the SalesOrderAgg2016 table to get the Month Name label from the OrderDate column:
    • OrderDateMonthName = FORMAT(DATEVALUE(SalesOrderAgg2016[OrderDate]), “MMMM”)
  5. With the base data ready, we can now look at creating an Actual and Target Measure on the SalesOrderAgg2016 table, using the following DAX formulas:
    • Actual Total Net Price = SUM(SalesOrderAgg[Total Net Price])
    • Target Total Net Price = 5750000
    • Actual Total Gross Price = SUM(SalesOrderAgg[Total Gross Price])
    • Target Total Gross Price = 5500000
  6. On the Report tab, under the Visualizations area, click on the KPI icon to add an empty KPI visualization to the report:
  7. Click on the newly created Visualization and, under the Fields area, drag and drop the appropriate SalesOrderAgg2016 fields into the wells indicated below:
  8. The visualization will then update accordingly, showing us the figure trend over each month and coloured red to indicate that this KPI is currently off target:
  9. To see how a KPI looks when ahead of target, we can repeat steps 5) and 6), but this time using the Net figures instead:

Key Takeaways

  • There are two principle visualization types available within Power BI to help track actual-to-target progress – KPIs and Gauges.
  • KPIs provide a more visually unique means of a binary success/fail determination when tracking towards a target. It is also possible to use KPI’s to track variance over time via the Trend axis. The Indicator will typically be the result of some form of aggregation or Measure.
  • Gauges provide a less visually distinctive, but non-binary, mechanism of viewing progress towards a target. Gauges support more potential field well values when compared with KPIs, nearly all of which are optional in some way. You can also manually define some of these values, for situations where your data model does not contain the required information.
  • All visualizations within Power BI are modifiable from a display or formatting perspective. The same basic options will generally be supported – such as changing a font type or background colour – with more specific configuration properties available per unique visualization type. For example, a KPI visualization can be customised to hide the background Trend Axis entirely. All of these options are designed to give developers greater control over the look and feel of their reports and to mirror them as closely as possible to any potential branding requirement.
  • When building out a solution designed to monitor progress within Power BI, the steps involved will typically be more in-depth than merely creating a new visualization. In most cases, there will be a requirement to bring together a lot of the other skills that have been discussed previously within this series – such as creating DAX formulas, modifying data within Power Query or bringing together different data sources into a single model. It is essential, therefore, not to underestimate the amount of time and effort involved in creating a practical solution that takes advantage of KPIs or Gauges.

I hope that this weeks post has been a little easier to bear when compared with DAX šŸ™‚ . In next weeks post, we will take a closer look at data hierarchies and how to apply these to visualizations within Power BI.

Welcome to post number 6 in my series concerning Microsoft Exam 70-778. The series aims to provide a revision tool for those who are looking at taking the exam and to also provide an introduction into some of the fundamental concepts around Power BI. As alluded to previously on the blog, Power BI is increasingly a topic that Dynamics 365 Customer Engagement professionals need to grapple with, particularly if they wish to implement a reporting solution that helps to ensure the long-term success of their CRM deployments. Going for Exam 70-778 gives you an ideal opportunity to familiarise yourself with this exciting technology area.

We moved into the Modeling and Visualizing Data theme for the first time last week, and we now jump into a relatively big subject this week – Create calculated columns, calculated tables, and measures. The skill areas covered here are as follows:

Create DAX formulas for calculated columns, calculated tables, and measures; Use What If parameters

To follow on as part of the examples that follow, make sure you have downloaded the WideWorldImporters sample database and hooked up Power BI Desktop to the tables described in my first post in this series.

Now, to begin this week’s post, we must first ask ourselves an essential question:

What is DAX?

Data Analysis Expressions, or DAX, has its roots in SQL Server Analysis Services (SSAS). First released as part of theĀ PowerPivot for Excel 2010 Add-in, DAX has quickly become the preferred language to use when working with Excel PowerPivot, SSAS and – most importantly – Power BI. DAX bears a lot of similarity to standard Excel functions, but a critical differentiator is that DAXĀ does not operate in the context of cells and is considered a strongly typed language (i.e. can lead to a higher risk of errors at runtime). A benefit of using DAX over PowerQuery/M is that the language handles a lot of data conversions implicitly. It also supports the usual list of data types associated with SQL type databases – Decimal/Whole Numbers, DateTime, Text, Boolean etc. A successful Power BI Reporting solution typically relies heavily on the building of complex DAX formulas, which are then complemented by the most appropriate visualisation at the report level.

Context in DAX

An important concept to grapple with DAX is that of context, and how this applies to the various formulas that you build out. There are two types of context:

  • Row Context: Essentially meaning the “current row”, functions that operate in this manner will be processed for every single row on your dataset, outputting the result required each time. For example, the IF function works on a row context.
  • Filter Context: Functions that support this will take into account any filters defined at the report level when performing the relevant calculation. Consequently, a lot of the work involved in contextually updating visuals is handled automatically by your DAX formulas, with no need to configure specific Measures/calculated columns as a workaround.

It is essential to understand these two types of context and on how they can impact on each other (the “context transition“). More critically, it’s also crucial to be aware of when a DAX function ignores a context completely. For example, some aggregation functions completely ignore row context. This article on the SQL BI website is a great resource that discusses this whole topic in depth.

Utilising DAX in Power BI

The Modeling tab is your go-to area when working with the three types of DAX calculations within Power BI Desktop – Measures,Ā Calculated Columns and Calculated Tables:

The sections that follow will discuss each of these in further detail, but there are some useful points to highlight that apply when using DAX generally:

  • Unlike the M Query Editor, there is full Intellisense support provided for DAX within Power BI. Straightaway, this makes this a much more streamlined tool that should be gravitated towards automatically as a consequence.
  • You should always use fully-qualified names that reference both the table object and field name that you are attempting to access. An example of how this should look can be seen below with the custom columnĀ EditedByDay on theĀ WideWorldImportersĀ Sales.OrderLines table:
  • It is also possible to add comments to your DAX code, with three flavours on offer; this is the part where we distinguish between the C# and SQL Server developers šŸ™‚ :
    • Single Line Comments: // or
    • Multi-Line Comments: /* and */
  • The screenshot below indicates how to utilise these as part of a DAX calculated column:

Let’s take a look now at the specific areas referenced above in greater detail:

Measures

Measures are best thought of as being fixed (or scalar) values, typically aggregations based on simple or more complex underlying formulas. A straightforward example could be just a Count of all rows on a table e.g

  • SalesOrderLinesCount = COUNTROWS(‘Sales OrderLines’)

Although Measures are compatible with any Visualization, the most common place to find them will be in a Card, as indicated below:

When first working with Measures, an advantageous feature at our disposal is theĀ New Quick MeasureĀ option, which provides a softer, GUI-focused approach to creating common types of Measure:

Beyond this, the sky is the limit when working with DAX Measures, with a whole range of different functions that can be used in isolation or tandem with each other.

Calculated Columns

We saw how it is possible to perform simplistic custom/calculated columns using the Power Query Editor as part of an earlier blog post. The functionality offered here can prove useful for straightforward requirements – for example, ifĀ Column1 equals “AAA”, thenĀ TRUE, elseĀ FALSE – but may soon become insufficient if you have more advanced needs for your new column. In this scenario, DAX calculated columns could come to the rescue and give you the ability to moreĀ familiarly define your business logic, particularly for those with their Excel head on.

A key consideration when working with DAX Calculated Columns is understanding the impact they have on performance; specifically, that they consume RAM and disk space within your report.Ā You should be aware also of the concept of circular dependencies. It is possible to build highly complex formulas within DAX, that may reference other DAX columns that you have built out. As a consequence, attempting to modify any dependent DAX formulas could result in errors. In addition to this, trying to define multiple calculated columns that perform the same operation may also lead to circular dependencies. There is another excellent article on the SQL BI website that jumps into this whole subject area in greater detail and is worth a read.

Similar to Measures, pretty much every type of DAX function listed here (minus ones that return table values) is useful in some way when working with Calculated Columns.

Calculated Tables

The types of DAX functions discussed so far focus on returning a single value – either in the sense of a fixed, often aggregated, value as part of a Measure or a single value on a row-set with a Calculated Column. The next step up from that, and where DAX enters a whole new realm of usefulness, is the ability to define Calculated Tables based on DAX formulas. There are a LOT of options at your disposal here. As well as being able to derive Calculated Tables directly from other objects within your model (e.g. the DAX formula SalesOrderLinesDupe = ‘Sales OrderLines’ would create a copy of theĀ Sales OrderLines table), there is also a wide array of different functions available that, when used individually or in combination, can satisfy any particular requirement:

  • The FILTER function returns a new table object filtered from another, based on a specific value. The CALCULATETABLE function is the “big brother” of this particular function, with the ability to define multiple, potentially complex filter conditions.
    • e.g. The following formula will return allĀ Sales OrderLine records with a PickedQuantity value greater than 7.
  • The ALL function returns all data from a table with underlying filters removed (i.e. filters that are defined using the options indicated in the screenshot below. It can be used to specify a whole table or single/multiple column(s). There are also variants of this, such as ALLEXCEPT (which will return all columns except the ones listed) and ALLNOBLANKROW (which removes any blank rows before returning the data).
  • VALUES and DISTINCT return a list of distinct rows only. VALUES can only work with physical tables, whereas DISTINCT also works with table expressions.
    • e.g. Both of the DAX functions below do the same thing – return 227 distinctĀ Description values from theĀ Sales OrderLines table:
  • SUMMARIZE groups a table by one or more columns, adding new ones where necessary. The grouping is optional. Only rows that contain data will return.
    • e.g. The follow DAX formula will return the total UnitPrice for each Sales OrderLines Description:
  • ADDCOLUMNS returns an existing table with new columns defined, typically as DAX formulas. SELECTCOLUMNS behaves similarly, but you have to specify the columns to return, including any new ones.
    • e.g. SalesOrderLinesWithCalcColumn = ADDCOLUMNS(‘Sales OrderLines’, “IsEndOfMonth”, IF(DAY(‘Sales OrderLines'[LastEditedBy]) >= 25, TRUE(), FALSE()))Ā returns theĀ Sales OrderLines table, with the exampleĀ IsEndOfMonth column included as a new column based on a DAX formula.
  • TOPN performs ranking based on conditions and returns the specified number of records based on the ranking criteria. Ties can also occur, meaning that the specified ranking number may not correlate to the actual number of rows returned.
    • e.g. The following formula will return the Top 25Ā Sales OrderLines records by UnitPrice. Notice the number of records returned exceeds 25; this is because of multiple ties within the underlying data.
  • There is a range of functions available to accommodate common table-join scenarios. CROSSJOIN allows for Cartesian product joins from multiple tables; NATURALINNERJOIN & NATURALLEFTOUTERJOIN perform self-explanatory joins, which seasoned SQL Server developers should have no trouble in guessing; and GENERATE & GENERATEALL allow for more bespoke table joins to take place.
    • e.g. the following formula will perform an inner join of theĀ Sales Orders andĀ Sales OrderLines table and return the ID values from both as a new table object
  • GENERATESERIES creates a single table with a list of numerical values. It requires the start & end number and the (optional) increment value.
    • e.g. SeriesExample = GENERATESERIES(1, 150, 1) will generate a single column table object with 150 rows, numbered 1 to 150.
  • CALENDAR generates a list of date/time values in a single table column object, based on the range specified. CALENDARAUTO behaves the same but creates a list of relevant date values after evaluating the whole model (e.g. if there is another table with dates between January 1st and November 31st 2018, then this would be the list of dates generated).
    • e.g. Using CalendarAuto with theĀ WideWorldImportersĀ Sales OrdersĀ andĀ Sales OrderLines tables will create date values from the 1st January 2013 through to the 31st November 2016.
  • ROW generates a single row table with the column values specified, based on a key/value pairing.
    • e.g.Ā RowExample = ROW(“My DAX”, “Brings all the nerds to the yard”, “And they’re like”, “Do you wanna write Measures”, “Damn right, I want to write Measures”, “I can teach you, but you must learn M first”)Ā produces the following table:
  • UNION combines two or more tables vertically, retaining duplicates. Column names do not need to match, but the number of columns does. INTERSECT is similar to UNION but maintains only values that exist the same in both tables. Nested UNIONS are usable alongside this, but the specified order can impact the result. EXCEPT is similar to UNION and INTERSECT, but outputs rows that exist only in the first table, not the second one.
  • DATATABLE allows the user to specify tables that contain manually entered data, similar to the Enter Data feature:
    • e.g.:

What If Parameters

There may be times within Power BI where you need to provide some predictive modelling capability. For example, show how potential sales will look based on an increased margin of 25%. Or it could be, as developers, we wish to test the functionality of some of our DAX formulas by having the ability to see how potential ranges of values will affect our calculations. For either scenario, What If Parameters can provide a solution. By default, on creation, a slicer control is provided that can be placed anywhere on your Report. To create a What If Parameter:

  1. Within Power BI Desktop, on the ModelingĀ tab, select the New Parameter option:
  2. On the What-if parameter dialog box, enter the details as indicated below and press OK:
  1. This will then add the following components to your report:
    • A Slicer control called TestWhatIfParameter
    • A new Calculated Table called TestWhatIfParameter:
  2. Under the Visualizations tab, select the Gauge visualization to add it your report. Define the properties for the Gauge visual as follows:
  1. Now, when you adjust the value in the TestWhatIfParameter, the Gauge will update accordingly. For example, setting it to 75 will update the report as follows:

This simple example does not do justice to the potential that this feature has, so I would recommend exploring it further yourself.

Key Takeaways

  • DAX is the primary formula language when working with datasets outside of Power Query. It includes, to date, more than 200 different types of functions that can assist in all sorts of data modelling.
  • An important concept to grasp within DAX is context and, specifically, row context (formulas that calculate a result for each row in a dataset) and filter context (formulas that automatically apply any filtering carried out at report level).
  • The sheer amount of DAX functions available makes it impossible to master and remember all of them, particularly when it comes to the exam. Your learning should, therefore, focus on learning the general syntax of DAX and the general types of functions available (aggregation, date/time etc.)
  • There are three principal means of utilising DAX with Power BI:
    • As Measures: These typically present a scalar value of some description, often an aggregation or a result of a complex formula. Using them in association with a Card visualization type is recommended, but this is not a strict requirement.
    • As Calculated Columns: Similar to the options available within Power Query, Calculated Columns provide a dynamic and versatile means of adding new columns onto your datasets. Compared with the options available within Power Query and the complexity of the M language, DAX Calculated Columns might represent a more straightforward means of adding custom columns onto your datasets.
    • As Calculated Tables: A powerful feature, mainly when used in conjunction with Calculated Columns, you have the ability here to create entirely new datasets within the model. These will typically derive from any existing datasets you have brought in from Power Query, but you also have functionality here to create Date tables, sequence numbers and manually defined datasets as well.
  • What-if Parameters provide of means of testing DAX formulas, as well as allowing report users to perform predictive adjustments that can affect multiple visualizations on a report.

DAX is a subject so vast that there are entire books devoted to it. Therefore, from an exam perspective, don’t worry too much about becoming a DAX master. In next weeks post, we’ll hopefully go down a few gears as we take a look at how to work with KPIs and how to apply them to visualizations.

Another week, another Power BI post šŸ™‚ This is the fifth post in my series focusing on Microsoft Exam 70-778, where I aim to provide a detailed overview and description of the skills needed to tackle the exam successfully. Last week’s post rounded off the opening theme, Consuming and Transforming Data By Using Power BI Desktop, as we focused on the options available to us to help tidy up data within the Power Query Editor. We now start to move outside of this area of the application with the Modelling and Visualising Data theme and, specifically, the following skill area:

Manage relationships; optimize models for reporting; manually type in data; use Power Query

As before, to follow on as part of any examples, make sure you have downloaded the WideWorldImporters sample database and hooked up Power BI Desktop to the tables described in my first post in this series.

Relationships

It will be a rare occurrence if you find yourself regularly working with a single table object within Power BI. The required data for most reporting situations will generally reside in different locations – whether within separate tables in a single SQL database or across different instances or applications. In most cases, there will be some field – or key – that will link records together and act as the bedrock when defining Power BI Relationships.

For those who are coming from a SQL or Dynamics 365 Customer Engagement background, I would expect that the concept of Relationships will present little difficulty in grasping. But, for those who are coming in cold, it is worth considering the following question – Why might you want to look at defining Relationships in the first place? The reasons will generally be specific to your business need, but can be generalised as follows:

  • Database normalization creates a degree of complexity that can be difficult to unweave from a reporting standpoint. Customer records may exist in one table, Address details in another, their orders in a third…you get the picture. Consequently, there is a need to bring this all together into a single/simplified view when developing reporting solutions; Relationships help to meet this objective.
  • The great benefit of using Power BI is its ability to provide a consistent, contextual filtering experience across Reports. So, for example, when adjusting the range on a Date slicer, all other visualisations are automatically refreshed to only show data from the updated period. This wizardry is all achieved by Relationships and the cross-filtering rules you define for them.
  • Power BI data models can grow in complexity over time. The defining of Relationships simplifies this to a degree and allows you to view all of your queries in a clean, visual manner.

We can see an example of how a Power BI Relationship looks in the picture below:

We can see above two one-to-many (1:N) Relationship between the Sales.Customers and Sales.Invoices tables from the WideWorldImporters database. The matching key, in this case, is the CustomerID field.

Relationships are created and managed in one of two ways:

  • Automatically when bringing data into Power BI Desktop the first time. Power BI will automatically detect and build any Relationships within a SQL Server-derived data source, based on any PRIMARY KEY/FOREIGN KEY CONSTRAINTs that exist already. The Select Related Tables option can be used to intelligently determine which tables can be brought in on this basis.
  • By going to the Manage Relationships button on the Home tab within the main Power BI window (NOT the Power Query Editor). This option will allow you to manage and create new Relationships based on your requirements. An Autodetect… option is also available, which behaves similarly to the data source auto-detection discussed previously. The example below shows the WideWorldImporters Relationships referenced earlier:

The walkthrough exercise at the end of this post will go through the process of creating a Relationship from start to finish, but it is useful at this stage to dive into the more detailed concepts involving Relationships:

  • The following types of Relationships (or Cardinality) can be specified:
    • One to Many (1:N) / Many to One (N:1)
    • One to One (1:1)
    • Many to Many (N:N) – This feature is currently in Preview and caution is advised when using it.
  • Each Relationship requires that a Cross filter direction is specified, which will determine how filters cascade out to related tables. These are configurable as either Single (in the direction of your choosing – left to right or right to left) or Both. 1:N Relationships must always use Both.
  • Only one Relationship is classifiable as Active between two tables at any given time. Despite this limitation, it is possible to define multiple Relationships, if required. It is essential to understand with this that only Active Relationships are utilisable when performing DAX column/measure calculation, which may lead to unintended results. You can get around this by either changing the Inactive Relationship to Active or by taking advantage the USERELATIONSHIP function to force DAX to pick your preferred Relationship.
  • The Assume Referential integrity option is only available when you are using DirectQuery and, if enabled, underlying queries will attempt to use INNER JOINs as opposed to OUTER JOINs in any query folding. While this can speed things up, it may also lead to missing or incomplete results.

Formatting Columns

We saw in last week’s post some of the more functional data formatting options that the Power Query Editor provides. I say the word functional in the sense of meaning formatting that serves a specific, data quality need, as opposed to a styling requirement. Moving outside of the Power Query Editor and we can discover a whole host of additional styling formatting options on the Modelling tab:

Let’s break all this down into a bit more detail:

Formatting

  • Data Type: Here, it is possible to override the data type for a particular column value. The options available are:
    • Decimal Number
    • Fixed decimal number
    • Whole Number
    • Date/Time
    • Date
    • Time
    • Text
    • True/False
    • Binary
  • Format: Depending on the Data Type value selected, the list of choices here will change accordingly. In most cases, such as forĀ Text values, you will only have a single selection available, but your options for Date/Time and Currency values are considerably richer. For example:
    • Date/Time values can be formatted in numerous different ways, both from a styling and locale perspective. This includes formatting dates in the good ol’ British way, meaning that the ISO 8601 date value 2018-01-01 can appear as 01-01-2018, 1 January 2018 or even Monday, 1 January 2018.
    • Currency values are modifiable to any global currency sign or shortcode. So, the value 50 is displayable as both Ā£50 and 50 GBP for British Pound currency figures.
  • $ Button: This is, in effect, an alternate way of changing the Currency value on a field, as described previously.
  • % Button: Converts a value to a percentage. For example, the Whole Number value of 75 would appear as 75%.
  • ‘ Button: Adds a thousand separator to a number value. Therefore, 100000 becomes 100,000.
  • .0/.00 Button: Adjusts the number of decimal places at the end of a numeric, up to a maximum limit of 15.

Properties

  • Data Category: This option, designed primarily for String values, allows you to indicate the type of data the field represents, with a specific focus towards mapping capabilities. For example, it is possible to mark a field as State or Province, to ensure that Power BI does not make an incorrect assumption when attempting to plot an address to a map. Some additional options here include:
    • Latitude/Longitude to, again, ensure accurate plotting to a map visual.
    • Classifying a field as containing either a Web URL or Image URL.
    • Categorising a field as Barcode value.
  • Default Summarization: When building out a report and adding fields to visualizations, Power BI assumes the preferred aggregation value to use. This will be pretty spot on in most cases – for example, Currency fields will default to a Sum aggregation. However, it is possible to override this option to use any of the following, easily guessable, alternatives:
    • Don’t summarize.
    • Sum
    • Average
    • Minimum
    • Maximum
    • Count
    • Count (Distinct)

Sort By Column

There may be a requirement to perform sorting of column values, based on some predefined logic as opposed to merely alphabetical or numeric order. The best example to provide here would be for Month Names. Instead of sorting in alphabetical order, like so:

April

August

December

February

etc.

We would want to sort by logical Month order i.e.:

January

February

March

April

etc.

Another scenario could be column sorting based on High, Medium and Low categorisation; all records with a value of High would need to appear first and the Medium second, as opposed to Low if sorted alphabetically.

Sort By Column gives us the flexibility to achieve these requirements, but it is crucial first to ensure that a proper sorting column resides within your query. Going into the Power Query Editor and modifying theĀ Sales CustomerTransactions table enables us to add on a Month Number column for the TransactionDate field by going to the Add Column tab and selecting Date -> Month -> Month option:

Coming out of the Power Query Editor using the Close & Apply option, go into the Sales CustomerTransactions table within theĀ Data tab and select theĀ Sort By Column option, selecting theĀ MonthNumberĀ field as the Sort by Column:

We should now see when adding a simple Table visualization to a report, that the Month values derived from TransactionDate appear in correct sort order:

Through this example, you can hopefully see how straightforward it is to accommodate more unique sorting requirements with minimal effort.

Manual Data Entry

Although Power BI assumes, for the most part, no requirement to define any manual data as part of your reports, there may be situations where this need arises. For example, on the WideWorldImporters Sales.CustomerTransactions table that has been built out through this series, you may want to define a manual list of all FullName values that can then be linked up to a slicer for filtering purposes. There are two ways this requirement can be met, with both options having their benefits and potential preference, based on how much you like to code šŸ™‚ :

  • The Enter Data button on the Power Bi Desktop Home tab allows you to create a new Table object based on manual data entry. The table can have as many columns as you like and, what’s nifty is that it supports full copy + paste functionality from other applications. It is a VERY convenient feature to bear in mind:
  • Using PowerQuery, it is also possible to define custom Tables populated with data. We saw an example of how to achieve this in last weeks post when creating some example data for cleansing, but there are some further options available to us:
    • Simple lists of data can appear within brackets. For example, the following M code:
      • let
            Source = { "This", "Is", "A", "Test"}
        in
            Source
    • Would generate the following data within Power Query:
    • Records with multiple rows/fields can also be generated using the example code below:
      • Table.FromRecords({[Did = "This", You = "is", Know = "Testing"]})
    • Which would appear as follows:
    • Finally, as shown last week, it is possible to create a Table object, using the example code below:
      • #table(
            {"Forename", "Surname", "Description"},
            {
                {"JANE","smith","   this describes the record"},
                {"alan", "JOHNSON", "    record description detected    "},
                {"   MARK", "CORRIGAN    ","another description"},
                {"JANE","smith","   this describes the record"}
            }
        )

The tools available here help to satisfy many potential ad-hoc requirements when using Power BI Desktop and when used in conjunction with other features described in this post and earlier in the series, start to become pretty powerful.

Example: Creating Relationships

In this very straightforward example, we will see how it is possible to create a Relationship manually from within Power BI Desktop, using the WideWorldImporters sample database as our connection source:

  1. It will be necessary to disable to automatic creation of Relationships when importing data for this exercise. This is done by going to the Options -> Data Load screen and ensuring that the appropriate option has been unticked:

  1. Follow the instructions discussed in this post to connect to the WideWorldImporters database but, on this occasion, ensure that the following tables only are selected for import:

  1. Once the data loads into Power Bi, navigate to the Relationships tab and you should see the two table objects indicated below (you may need to resize them accordingly for all fields to display):
  2. Because of the setting change carried out in step 1, the underlying Relationship between both of these tables in SQL has not been detected and added. To fix this, click on the Manage Relationships button on the Home tab and then the New button to open the Create relationship window, as indicated below:

  1. Define the following settings for the new Relationship, as described and illustrated below:
    • In the first drop-down box, select Sales Orders and verify that the OrderID column highlights itself.
    • In the second drop-down box, select Sales OrderLines and verify that the OrderID column highlights itself.
    • For Cardinality, ensure One to many (1:*) is selected.
    • For Cross filter direction, ensure Both is selected.
    • Ensure Make this relationship active is ticked.

  1. Press OK and then Done to create the Relationship. The Relationship window should refresh accordingly to indicate that a new Relationship exists between the two tables:

Key Takeaways

  • Relationships form the cornerstone of ensuring the long-term viability and scalability of a large data model. Assuming you are working with well-built out, existing data sources, Power BI will automatically detect and create Relationships for you. In situations where more granular control is required, these Relationships can be specified manually if needed. It is worth keeping in mind the following important features of Relationships:
    • They support one-to-one (1:N), one-to-many (1:N) and many-to-one (N:1) cardinality, with many-to-many (N:N) currently in preview.
    • Filter directions can be specified either one way or bi-directionally.
    • Only one relationship can be active on a table at any given time.
  • It is possible to sort columns using more highly tailored custom logic via the Sort By Column feature. The most common requirement for this generally involves the sorting of Month Names in date order but can be extended to cover other scenarios if required. To implement, you should ensure that your data has a numbered column to indicate the preferred sort order.
  • Moving outside of the Power Query Editor presents us with more flexibility when it comes to formatting data to suit particular styling or locale requirements. While the majority of this functionality provides date/time and currency formatting options, for the most part, it is also possible to categorise data based on Location, the type of URL it is or on whether or not it represents a Barcode value; these options can assist Power BI when rendering certain types of visualizations.
  • There may be ad-hoc requirements to add manually defined data into Power BI – for example, a list of values that need linking to a Slicer control. The Enter Data button is the “no-code” route to achieving this and supports the ability to copy & paste data from external sources. For more advanced scenarios, you also have at your disposal a range of M code functionality to create Lists, Records and Tables, which can be extended further as required.

Hopefully, by now, you are starting to get a good feel for how Power BI works and also the expected focus areas for the exam. Next weeks post is going to be a biggie, as we jump head first into DAX formulas and how they can be used for calculated columns and Measures. We’ll also introduce the concept of What-if Parameters and how they work in practice.

Welcome to the third post in my series on Microsoft Exam 70-778, where I aim to provide a detailed overview and description of the skills needed to tackle the exam successfully. We saw in the previous post how we could use the Power Query Editor to perform a multitude of different transformations against data sources; this will now be taken further, as we start to look at how to ensure optimal quality within our Power BI data models. The relevant skills for this area as follows:

Manage incomplete data; meet data quality requirements

To follow on as part of the examples below, make sure you have downloaded the WideWorldImporters sample database and hooked up Power BI Desktop to the tables described in my first post in this series. With all this done, it is crucial first to grapple a vital concept relating to the management of incomplete data, namely, the ability to…

Filter Data

If you are consuming an entire SQL tables worth of data within the Power Query Editor, the size of your model can grow over time. In this scenario, it may be necessary to apply column-level filters to your data directly within Power Query, as opposed to merely dumping a cartload of irrelevant data in front of your users. Fortunately, the whole experience of filtering data should be a cakewalk, given its similarity with Microsoft Excel filters, as we can see below when attempting to filter the FullName field on theĀ WideWorldImporters Sales.CustomerTransactions table:

This particular feature does have some differences compared with Excel, though:

  • For large datasets, only a subset of all total, distinct record values are returned. This fact is indicated above via the List may be incomplete warning sign. Clicking the Load more option will do exactly that, but may take some time to process.
  • The range of filters available will differ depending on the data type of the column. For example, String data types will have filters such as Begins With...,Ā  Does Not Begin With…, whereas Dates are filterable based on Year, Quarter, Month etc.
  • The Remove Empty option will do just that – remove any columns that have a blank or NULL value.
  • As discussed on the Performing Data Transformations post, when you start combining filters with Parameters, it is possible to transform particularly unwieldy filtering solutions into more simplistic variants.

Column Errors

When it comes to adding custom columns as part of your Power Query transformations, there is always the potential for these to error, due to a misconfigured calculation or some other kind of unforeseen issue. When this occurs, the corresponding column value is flagged accordingly within the Power Query Editor and can be inspected further to determine the root cause of the issue. To demonstrate this, add on the following custom column using the following formula onto the Sales.CustomerTransactions table

[AmountExcludingTax] * [Error]

The value for each row in the new column should resemble the following:

When this issue occurs, it may be most prudent to first try and address the issue with your underlying calculation, ideally fixing it so that no error occurs at all. Where this is not possible, you can then look towards removing any rows that contain such a value. We will see how this can be done in the Adding/Removing Rows section later in this post.

Blank Data

On the flip side of field value errors is blank data. In most cases, when working with SQL data sources, this will rear its head when there are NULL values in your database. For example, take a look at below at the CreditLimit field on the WideWorldImporters Sales.Customers table:

When these fields are fed through from Power Query and relied upon as part of DAX custom column/Measure creation, you may start to get some unintended consequences. For example, after filtering the same table above only to retain rows where the CreditLimit equals null, attempting to create a Measure that totals up all CreditLimit values results in the following when displayed as a Card visualisation:

If you, therefore, have a desire to perform additional aggregations or custom calculations on fields that contain blank/null values, then you should take the appropriate steps to either a) remove all rows that contain one of these two values or b) perform a Replace action on the column to ensure a proper, default value appears instead. For the CreditLimit field, this can be as simple as replacing all null values with 0:

Adding/Removing Rows

Often our data sources are not pristine clean from a data perspective – duplicate rows may be common, it could be that rows exist with completely blank or null values or your incoming data file could be a complete mess from a column header perspective. With this problem in mind, the Power Query Editor provides us with the functionality to keep or remove rows based on several different conditions:

The options granted here should be reasonably self-explanatory, but the list below contains some additional guidance if you need it:

  • Keep/Remove Top Rows: Keeps or removes the top number of rows, in ascending order, based on the amount you specify.
  • Keep/Remove Bottom Rows: Keeps or removes the bottom number of rows, in descending order, based on the number you specify.
  • Keep Range of Rows: Keeps the number of rows specified based on the starting row number. For example, for a 50-row table, if a First row value of 1 and a Number of rows value of 10 is selected, then the first ten rows will be retained.
  • Keep/Remove Duplicates: Based on the currently selected column(s), keeps or removes all rows with duplicate values.
  • Keep/Remove Errors: Based on the currently selected column(s), keeps or removes all rows that have an Error value.
  • Remove Blank Rows: Removes any row that has a blank or NULL value.

Formatting Column Data

Data from a live, production system, such as Dynamics 365 Customer Engagement, can sometimes be a complete mess from a readability perspective; incorrect casing and invalid characters are typically commonplace in this situation. Fortunately, there are a range of options at our disposal with the Power Query Editor, on the Transform tab:

Most of these are self-explanatory, with the exception of the Trim and Clean options:

  • Trim removes any leading/trailing whitespace characters from a string value.
  • Clean detects and removes any non-printable characters from a string value.

Although not technically a data cleansing options, there are some clear usage scenarios for the Add Prefix & Add Suffix options, such as creating unique reference code for each column value, based on the unique record number value (e.g. ABCD-1, ABCD-2 etc.).

Formatting options for other column types are not available from within Power Query. So if, for example, you wished to format all date values in the format YYYY-MM-DD, you would have to move outside of the Power Query Editor to achieve this. The steps involved to accomplish this will be a topic for a future post.

Example: Cleansing Data

Having reviewed each of the possible cleansing options at our disposal, let’s now take a look at an example of how to cleanse a troublesome dataset:

  1. Within the Power Query Editor, on the Home tab, select the New Source -> Blank Query option. This will create a new Query in the left-hand pane called Query1.
  2. Select Query1 and press F2 to allow you to rename it to CleanseExample.
  3. Right-click the CleanseExample query and select the Advanced Editor option:
  4. Within the Advanced Editor window, copy & paste the following code into the window:
    • #table(
          {"Forename", "Surname", "Description"},
          {
              {"JANE","smith","   this describes the record"},
              {"alan", "JOHNSON", "    record description detected    "},
              {"   MARK", "CORRIGAN    ","another description"},
              {"JANE","smith","   this describes the record"}
          }
      )
  5. It should resemble the below if done correctly:
  6. When you are ready, press the Done button. PowerQuery will then create a table object using the code specified, populating it with records, as indicated below:
  7. There are three key issues with this data that need resolving:
    • The inconsistent word casing on the Forename/Surname.
    • Whitespacing on the Description and ForeName fields.
    • Duplicate records.
  8. These issues are fixable by taking the following action:
    • For the casing issue, CTRL + left click to select the Forename & Surname fields, go to the Transform tab and select Format -> Capitalize Each Word. Your data will then be modified to resemble the below:
    • For the whitespace issue, select the Forename & Description fields and, on the Transform tab, select Format -> Trim:
    • Finally, to remove the duplicate record for Jane Smith, highlight the Forename & Surname fields, navigate to the Home tab and select Remove Rows -> Remove Duplicates. This will then leave us with three records, as illustrated below:
  9. As a final (optional) step, we can also look to clean up the Description field values by applying the Capitalize Each Word formatting option:

Et voilĆ ! We now have a tidy and clean table, ready for consumption within Power BI šŸ™‚

Key Takeaways

  • Data can be filtered directly within Power Query, using Excel-like functionality to assist you in only returning the most relevant data in your queries. The data type of each field plays a particularly important part of this, as only specific filter options will be at your disposal if, for example, you are working with numeric data.
  • From a data quality perspective, you typically will need to handle column values that contain one of two possible value types:
    • Errors: This will usually occur as a result of a calculated column field not working correctly. The best solution will always be to address any issues with your calculated column, such as by using a conditional statement to return a default value.
    • Blanks/NULLs: A common symptom when working with SQL derived data sources, your real problems with blank values start to appear when you attempt to implement DAX custom columns/Measures outside of the Power Query Editor. It is, therefore, recommended that these are dealt with via a Replace action, depending on your fields data types. For example, a number field with blank/NULL values should be replaced with 0.
  • The Remove Rows option(s) can act as a quick way of getting rid of any Error or Blank/NULL rows and can also be utilised further to remove duplicates or a range of rows. In most cases, you will have similar options available to you with Keep Rows instead.
  • There are a variety of formatting options available to us when working with text/string data types. These range from fixing capitalisation issues in data, through to removing whitespace/non-printable character sets and even the ability to prepend/append a new value.

Data cleansing is a reasonably short subject area in the grander scheme of things, but the steps covered represent key stages towards building out a competent and trustworthy reporting solution. The next post in the series will discuss the options available to us in building out more complex and bespoke data models.