Power BI interview questions

A calculated column is computed row by row at refresh and stored, so I use it when I need a value per row for slicing or grouping. A measure is calculated when a visual needs it, in the current filter context, so I use it for totals and ratios. For example, an order size band is a column, but margin percentage is a measure.

CALCULATE evaluates an expression after modifying the filter context. I use it for things like sales for one region or year-to-date sales, and with ALL to remove filters when I need a share of the grand total.

I would use a star schema with a Sales fact table and dimensions for Product, Customer, Region and Date. Each dimension has a one-to-many relationship to the fact table with single-direction filtering, and I mark a continuous date table for time intelligence.

Mostly in Power Query: setting types, removing columns, trimming text and merging tables, because those steps repeat on every refresh and keep the model small. I use DAX for calculations that depend on filter context.

I put the measure in a table visual broken down by the dimension involved to see which rows are off. Then I check relationships and key uniqueness, whether a column is being summed when it should be a measure, and whether a filter or inactive relationship is affecting it.

I publish from Power BI Desktop to the service and set scheduled refresh. If the data comes from a server on the local network, I make sure an on-premises data gateway is configured, and I check refresh history after the first few runs.