Microsoft Power BI Data Analyst

Data almost never arrives ready. You profile columns for quality and distribution, fix types, errors and gaps, and write the clean-up steps once so every refresh repeats them without effort.

• Column quality, distribution and profile tools
• Fixing data types, errors and empty values
• Splitting, replacing and standardizing text
• Removing duplicates and filtering bad rows
• Reading and editing steps in the Applied Steps pane

You build A cleaned query set built from a deliberately messy extract, with every step named and explained.

A good model starts with tables shaped for analysis. You combine sources with merge and append, reshape with pivot and unpivot, and split a flat file into fact and dimension tables ready to load.

• Merging queries and choosing a join kind
• Appending tables from several sources
• Pivot, unpivot and group by
• Fact and dimension tables in a star schema
• Reference or duplicate: reusing queries safely
• Query folding and why it speeds a refresh

You build A star schema of fact and dimension queries, built from three source files.

Relationships decide what every number means. You set cardinality and filter direction, build hierarchies, and tidy the model so report authors find the right field and never the wrong one.

• Cardinality and filter direction
• Role-playing dimensions and inactive relationships
• Building and marking a date table
• Hierarchies, display folders and hidden fields
• Many-to-many relationships and when to avoid them

You build A clean semantic model with a date table, tested relationships and fields organized for report authors.

Measures are where a model earns trust. You learn row and filter context, write measures with CALCULATE and variables, and make sure a total still adds up when a slicer or a filter changes the picture.

• Measures against calculated columns
• Row context and filter context
• CALCULATE and modifying filters
• Iterators such as SUMX
• Variables for readable, faster formulas
• Safe division and handling blanks

You build A measure set for sales, margin and share of total that stays correct under any filter.

Business questions are almost always about time: this period against last. You write year-to-date, prior-period and rolling-window measures, and add ranking and top-N logic that responds to what the user selects.

• Year-to-date, quarter-to-date and month-to-date
• Comparing with the same period last year
• Rolling averages and running totals
• Ranking and top-N with RANKX and TOPN
• Fiscal calendars and custom periods

You build A time-intelligence page of measures: growth against last year, rolling averages and a ranked top ten.

A report that takes thirty seconds to load will not be used. You measure where the time goes, trim columns and cardinality, and rewrite heavy measures, so that a faster model is proven by numbers.

• Performance Analyzer: reading visual load times
• Model size, cardinality and unused columns
• Rewriting slow DAX
• Aggregation tables for large data
• Reducing query load from visuals

You build A performance review of a slow report, listing each fix and the measured change.

A report is an argument made with visuals. You choose the chart that fits each question, lay out a page so the eye lands where it should, and use formatting and tooltips to show what matters and hide what does not.

• Matching a visual to the question it answers
• Page layout, grids and visual hierarchy
• Themes and consistent formatting
• Conditional formatting and reference lines
• Slicers, filters and report page tooltips

You build A three-page report that tells one clear story from a single business question.

People explore a report in ways you did not plan. You add drillthrough, bookmarks and field parameters, design a mobile layout, and make the report usable with a keyboard and a screen reader.

• Drillthrough and drill down
• Bookmarks, buttons and navigation pages
• Field parameters and synced slicers
• Mobile layouts
• Alt text, tab order and accessible color

You build A navigable report with drillthrough, bookmarks and a mobile view, checked for accessibility.

Some answers are hard to see in a chart. You use AI visuals to explain drivers, break a measure down and flag anomalies, and use Copilot to draft pages and measures that you then check and correct.

• Key influencers: what drives a result
• Decomposition tree for root-cause exploration
• Anomaly detection and forecasting on line charts
• Smart narrative and Q&A
• Using Copilot to draft report pages and measures
• Checking AI output before sharing it

You build An analysis page that explains a change in a key metric using AI visuals, with each finding verified.

A report nobody can open helps nobody. You organize workspaces and roles, publish apps, schedule refresh, set up incremental refresh, and decide when a gateway is needed to reach data on a private network.

• Workspaces, roles and apps
• Scheduled refresh and its limits
• Incremental refresh for large tables
• On-premises data gateways and when you need one
• Endorsement, lineage and impact analysis

You build A published workspace and app with scheduled refresh, and a note explaining the gateway decision.

A shared model can leak data as easily as it can inform. You write static and dynamic row-level security, apply sensitivity labels, and control who may view, build on or reshare your content.

• Static row-level security roles
• Dynamic row-level security with a user table
• Testing a role as another user
• Sensitivity labels on reports and models
• Sharing, Build permission and reshare rights

You build A dynamic row-level security design tested against three user personas.

Everything meets in one project. You take a brief and raw data through preparation, modeling, report design and publishing, then present the result and defend the choices you made.

• Reading the brief and agreeing success measures
• Preparing and modeling the data
• Building and refining the report
• Publishing, refreshing and securing it
• Presenting the result to stakeholders

You build A complete Power BI solution: cleaned data, a tested model, a designed report and a secured workspace.