06: Modelling Data In Power

BI Desktop
Chris Webb
[email protected]
Agenda
• What is the Data Model?
• Tables
• Relationships
• How much data can you import into Power BI?
• Formatting columns
• Sorting columns by other columns
• Other properties
What Is The Data Model?
• The Power BI Data Model is a representation of all of the data you
need for your reports
• The Data Model consists of tables connected by relationships
• It is only visible if you are importing data into Power BI or using
DirectQuery mode
• If you are using a live connection to SQL Server Analysis Services, then
you use the Analysis Services Tabular model/cube instead
Tables
• Tables in the Data Model have a fixed number of columns and a
variable number of rows
• Tables can be previewed in the Tables view in Power BI Desktop
• Each column has a data type
• Tables and columns are visible to end users in the Fields pane on the
Report view by default
• Columns are what you drag and drop to build visualisations
• You can hide columns and entire tables from the Report pane though
Relationships
• Tables can be joined to each other using relationships
• Think of a relationship as being like a lookup
• They may be built automatically when you load data, or you may have
to build them yourself manually in the Relationships view
• Relationships can only join using one column on each table
• If there is a value on the ‘many’ side that doesn’t match a value on
the ‘one’ side, it is matched to a blank value instead
Relationships
• Relationships can be either:
• One-to-one
• One-to-many
• Relationships can also be either:
• Active
• Inactive – that’s to say present, but they don’t do anything. Inactive
relationships can be activated inside DAX calculations.
Relationship Filter Direction
• Relationships have a cross filter direction
• If this is set to Both then making a selection on a column in a table on
one side of the relationship will filter the columns in the table on the
other
• If this is set to Single then
• Selecting something from the one side of the relationship will filter the many
side of the relationship
• Selecting something from the many side of the relationship will not filter the
one side of the relationship
• Filters can travel across multiple relationships
How Much Data Can You Import?
• As always: it depends. There is no easy answer!
• The amount of RAM on your desktop PC is one factor – though if you
deploy to PowerBI.com, model size is limited to 1GB
• The 32-bit version can’t address as much memory as 64-bit
• Compression is the other factor:
• Number of rows is not so important
• Neither is data type
• The number of distinct values in each column is important!
• Power BI can compress data extremely well: it is normal for the Data
Model to be 10-20% of the size of the original data
Reducing Memory Usage
• Do not import columns you have no use for
• Think carefully about importing columns with a large number of
distinct values
• Reduce the number of distinct values in columns by:
• Rounding numbers
• Avoiding junk dimensions
• Splitting single values into two or more separate columns, eg datetime values
into dates and times
• Avoiding calculated columns that ‘stage’ values for other calculations
Formatting Columns
• Values in fields can have formats applied in the Data view
• Different formatting options are available for different data types
• The formats only affect the values as they are displayed in the report,
not the values that are actually stored
Sorting Columns By Other Columns
• By default values will appear in alphabetical order in reports
• Though sort order can be changed in a particular report component
• Some values, like days of the week or months of the year, should
always appear in a set order in reports
• In these cases you can use the Sort By Column button in the Data view
to sort values in one column by those in another
• The values in the column you’re using for the sort order must:
• Sort in ascending order to get the order you want
• Have a one-to-one relationship with the values in the column you’re sorting
Hierarchies
• Hierarchies allow you to combine multiple columns together to let
users drill down through the data
• For example, a Calendar hierarchy would allow users to drill from Year to
Month to Date
• Hierarchies can be created by right-clicking on a column in the Fields
pane in the Report view
• Levels in a hierarchy can have different named to the underlying
columns
Other Properties
• Hide In Report View stops a column being visible when building
reports – useful for key columns
• Data Category tells Power BI what type of data is held in a column,
and therefore what kind of default visualisation is appropriate
• Default Summarization tells Power BI how a column should be
aggregated when it is dragged onto a report
• The Do Not Summarize option is particularly useful for numeric columns like
Years that should never be aggregated
• Synonyms give Power BI alternative words for Q&A querying