roadmap
- 01
- 02
- 03
- 04
- 05
- 06
- 07
- 08
Prepare for Power BI interview questions on modeling data, writing DAX and building reports. Covers star schemas, Power Query, DAX measures, choosing visuals, the Service and row-level security.
82 of 82 questions in your plan: 0 known, 0 in progress, 0 due, 0 skipped, 82 new. 0 cut, outside your plan.
Power BI questions come up in data analyst, BI analyst and BI developer loops, usually in a technical round after a SQL screen. Some loops swap that round for a take-home: a report built from a supplied dataset, then walked through with the panel. Either way the conversation starts from the model, not the canvas. Why does a matrix total disagree with the sum of its rows? What happens when a second relationship to the date table is added for ship date, and how does USERELATIONSHIP bring it back into one measure?
What lifts an answer is prediction: saying which number a visual will show, and why, before Power BI shows it. A measure that is right on a card and wrong inside a matrix is a question about which filters reach the expression, and about what context transition does when CALCULATE runs inside a row context. Past junior level the follow-ups weigh cost and control: a calculated column stored at every refresh where a measure would have been computed at query time, a Power Query step that breaks query folding, and why a workspace member sees every row that row-level security hides from a viewer.
Dimensional modeling comes first, with no Power BI in it, because the product expects a star and most of its modeling questions are star-schema questions in its own vocabulary. Grain leads, since every later choice depends on what one fact row means; star or snowflake is then a matter of how many relationships a filter crosses before it reaches the facts. Role-playing dimensions set up the inactive-relationship questions, and additivity the semi-additive ones, such as a balance that cannot be summed across months.
The Power BI module follows the data from source to reader. Power Query shapes what gets loaded and decides how much of that work the source does. The model turns the star into relationships and filter directions, and DAX measures come after it because they compute along those relationships. Visuals put measures on a page that the Service publishes and refreshes. Row-level security comes last for a reason: its roles are DAX filters written in the model, but whom they bind is settled only after publishing.
The final module leaves the product. Chart design is written for any plotting tool and asks whether a chart answers its question: whether the mark fits the comparison, what statistic the tool computed unasked (Power BI sums a numeric column by default, a year column too unless told otherwise), and when small multiples beat another legend color. The visuals section taught how a report is built; this one is about how it is read.
From Excel, Power Query will feel familiar, since Excel's Get & Transform runs the same M language; the gap is DAX, so give the measures section more hours than its place in the plan suggests. From SQL, the first module is review and the risk runs the other way: writing a measure as if it were a `GROUP BY`, when it is evaluated separately for every cell of a visual, under whatever filters reach that cell. Two weeks out, write measures instead of reading about them: once importance stops at must know and level stops at the job's seniority, answer each remaining DAX question in Desktop against a small model before opening the answer. Over two or three months, switch study order to by topics so the questions arrive in the plan's own sequence, and build one report alongside it, from a query that folds to a dynamic role checked with View as. The plan has no SQL module, though many analyst loops screen on SQL first; the [SQL roadmap](/roadmaps/sql) prepares that round. If the posting is a BI analyst role rather than a Power BI one, the [BI Analyst roadmap](/roadmaps/bi-analyst) covers the loop as a whole.
From Excel, Power Query will feel familiar, since Excel's Get & Transform runs the same M language; the gap is DAX, so give the measures section more hours than its place in the plan suggests. From SQL, the first module is review and the risk runs the other way: writing a measure as if it were a GROUP BY, when it is evaluated separately for every cell of a visual, under whatever filters reach that cell.
Two weeks out, write measures instead of reading about them: once importance stops at must know and level stops at the job's seniority, answer each remaining DAX question in Desktop against a small model before opening the answer. Over two or three months, switch study order to by topics so the questions arrive in the plan's own sequence, and build one report alongside it, from a query that folds to a dynamic role checked with View as.
The plan has no SQL module, though many analyst loops screen on SQL first; the SQL roadmap prepares that round. If the posting is a BI analyst role rather than a Power BI one, the BI Analyst roadmap covers the loop as a whole.