PL-300 Cheat Sheet.
The night-before summary, built like the exam.

The exam at a glance

Skills measured (as of 20 April 2026)

Storage modes

Power Query essentials

• Query folding: steps translated into a native source query. Remove columns and filter rows EARLY to preserve it. A greyed-out View Native Query means folding stopped there and the rest runs locally.
• Merge joins queries side by side on keys; append stacks rows from queries of the same shape.
• Unpivot turns column-per-month into one row per month; pivot does the reverse; transpose flips the whole table.
• Reference starts from another query's output so upstream edits flow through; duplicate copies the steps into an independent chain.
• Parameters centralize values such as file paths. RangeStart and RangeEnd are the reserved names incremental refresh requires.
• Column quality shows valid, error, and empty shares; column profile and column distribution give statistics.
• Change Type With Locale fixes regionally formatted dates imported as text.
• Privacy levels are set per data source and govern whether data may be folded across sources.

Modelling essentials

• Star schema: narrow fact table, descriptive dimensions, one-to-many, single filter direction. Beats a flattened table on both size and speed.
• Date table: contiguous dates, marked as the date table. Fact-table dates have gaps and break time intelligence.
• Role-playing dimension: second relationship inactive, activated per measure with USERELATIONSHIP inside CALCULATE.
• Measure vs calculated column: measure evaluates at query time in filter context and stores nothing; calculated column materializes per row at refresh and costs memory.
• Why totals don't add up: a measure is recomputed in the total row's filter context, so a ratio total is total-over-total, not the sum of row ratios.
• Semi-additive balances need LASTDATE or a closing-balance function — never a plain SUM across dates.
• DIVIDE for safe division; DISTINCTCOUNT for unique counts; REMOVEFILTERS to clear, KEEPFILTERS to preserve.
• Shrink a model by removing unused high-cardinality columns first; they compress worst.

Report features and what they are for

• Drillthrough: navigate to a detail page carrying the selected row's context.
• Bookmark: save and restore report state — filters, selection, visibility, sort — usually behind a button.
• Edit interactions: per visual pair, choose filter, highlight, or none.
• Sync slicers: keep a slicer selection aligned across pages.
• Field parameters: let a slicer switch which measure or field a visual shows.
• Visual calculations: written against the visual's result set, ideal for running totals and share of total.
• Analytics pane: trend lines, forecasts, reference and error bars. Analyze: explain an increase or decrease.
• Key influencers ranks factors behind an outcome; the decomposition tree drills, and high-value path auto-expands the biggest contributor.
• Accessibility: set tab order, add alt text, never encode meaning in color alone.
• Paginated report for fixed-layout printing; phone view for mobile layout; automatic page refresh for DirectQuery wall displays.

Manage and secure

• Workspace roles: Viewer consumes; Contributor publishes and edits; Member also shares and adds lower roles; Admin manages everything including users.
• Row-level security: DAX filter in a role, members assigned in the service. Defining a role in Desktop filters nobody until members are added.
• Report filters hide, RLS restricts. If the requirement is security, it is RLS.
• Gateway: required for the service to reach on-premises sources for refresh or DirectQuery.
• App: the way to distribute curated content read-only to a large audience without granting edit rights.
• Build permission: lets other authors create reports on a shared semantic model.
• Endorsement (Promoted, Certified) signals quality; certification is restricted to designated reviewers. Sensitivity labels signal confidentiality and persist into exports.
• Data alert fires on a dashboard card, gauge, or KPI tile at a threshold; subscription emails a snapshot on a schedule.
• Lineage view maps upstream and downstream dependencies before a breaking change.
• Shared capacity caps daily scheduled refreshes; dedicated capacity raises the ceiling.