Data Modeling Practices & Calculations
using Power BI Desktop
Paul TurleyMicrosoft Data Platform [email protected]
Data Modeling
Desktop Model
multiple data sources Desktop user
Deployed Desktop Model
published model, reports & dashboard
multiple organization users
Enterprise Data Sources
scheduled refresh SSAS live queries DirectQuery
ScalingUp and Out
Solution ScopePower BI
Dashboards
Interactive ReportsPower Pivot
Data Model
Power Query Transformations
Power BI Dashboards
Interactive Reports
Server-Hosted Analysis Services
Model
Dimensional Data Mart
ETL Process
Master Data ManagementSmall-scale Solution Using Only Power BI
Large-scale Solution using Power BI with On-premises or Cloud Services
application data sources
HR/Payroll
Order entry
CRM/ERP
operational data store
managed semantic models
Analytic & operational reports
ETL process
Self-service BI
Copyright 2016, Intelligent Business LLC, SqlServerBiBlog.com
application data sources
HR/Payroll
Order entry
CRM/ERP
staging /operational data
store
enterprise data warehouse
subject area data marts
managed semantic models
Ad-hoc models & visuals
Executive dashboards
Analytic & operational reports
ETL processes
Full-scale BI Solution
Copyright 2016, Intelligent Business LLC, SqlServerBiBlog.com
staging EDW
application data sources
HR/Payroll
Order entry
CRM/ERP
staging /operational data
store
subject area data mart
managed semantic models
Ad-hoc models & visuals
Executive dashboards
Analytic & operational reports
ETL processes
Scaled-down BI Solution
Copyright 2016, Intelligent Business LLC, SqlServerBiBlog.com
data sources stagingmanaged
semantic models
Ad-hoc models & visuals
Executive dashboards
Analytic & operational reports
Data Factory
Scaled-out Cloud Solution
Copyright 2016, Intelligent Business LLC, SqlServerBiBlog.com
Machine Learning
Data Lake
HD Insight
SQL Database/Data
Warehouse
structured
So Many Choices
Fewer Choices
More Choices
Magic Quadrant for Business Intelligence and Analytics PlatformsFeb, 2016
Ability to execute
Completeness of vision
in the Microsoft tool BI set
Airline Performance Analysis
• Bureau of Transportation Statistics
• Federal Aviation Administration
• Airline On-time Performance
Don’t Get Hung Up On Relationships(and other universal life lessons)
Paul’s Model Design Rules
• There are no “rules”…only “guidelines”
• OK, there really are rules but understand their purpose
• Import only what you need
• All object names should be user-friendly
• Expose each column or member in only one table
• Hide base measure columns
• Create explicit measures (or at least be consistent)
• Format all numeric columns and measures
• Apply dimensional modelling rules where possible (i.e. consolidate measures into fact tables, separate lookup tables)
• OK to use flat, mixed-use tables to keep things simple – sometimes
• Understand the trade-off decisions (scalability, quality, maintainability)
Datarace Runner Analysis
• About 4,000 competitive racing events, most in southern Europe
• About 2 million competitors
• Data from Timing Data Service
Bi-Directional Relationships(keep an open mind)
1 to M
M to 1
M to M
(drumroll please)