Microsoft Power BI Data Analyst PL-300 Exam Prep by Melarks PL-300 EXAM CHEAT SHEET Current curriculum baseline: Microsoft skills measured as of April 20, 2026. PREPARE THE DATA ---------------- • Import = in-memory copy; DirectQuery = source queried at interaction time; DirectLake = Fabric/OneLake storage mode. • Profile beyond the top 1,000 rows when uniqueness/quality conclusions matter. • Correct data type first; use locale-aware conversion for ambiguous dates/numbers. • Append adds rows. Merge joins columns by key. Left Anti returns unmatched left rows. • Reference inherits upstream query changes; Duplicate is an independent copy of steps. • Unpivot turns columns into attribute/value rows; Unpivot Other Columns is resilient to new period columns. • Remove unused rows/columns early and preserve query folding where possible. • Incremental refresh: Date/Time RangeStart + RangeEnd, commonly >= RangeStart and < RangeEnd. • Enable Load controls whether a query becomes a model table; refresh participation is a separate setting. • Privacy levels govern source isolation during combination; they are not RLS. DATA MODELING ------------- • Prefer a star schema: dimensions on the 1 side, fact table on the many side. • Define grain first: what exactly does one fact-table row represent? • Single-direction dimension → fact is the default best practice; bidirectional is a deliberate exception. • Many-to-many business relationships often need a bridge table. • Inactive relationships can be activated inside CALCULATE with USERELATIONSHIP. • Use role-playing date dimensions when independent Order Date and Ship Date slicers are required. • Date table: unique, non-null, contiguous date column; mark it as Date Table where appropriate. • Month Name sorts by Month Number; Month-Year needs a sequential YearMonth key. • Measures are dynamic filter-context calculations; calculated columns are stored row-level values. • High-cardinality unused text/IDs are common model-size targets. Hiding a column does not reduce model size. DAX & FILTER CONTEXT -------------------- • Measures already respond to filter context. CALCULATE is for changing context, not merely 'making filters work'. • Row context = current row; filter context = visible subset from visuals, slicers, filters and relationships. • CALCULATE modifies filter context and performs context transition when row context exists. • REMOVEFILTERS(Column) removes only that column's filter; unrelated filters remain. • KEEPFILTERS intersects a new same-column filter with the existing filter. • SUMX/AVERAGEX iterate a table; SUM/AVERAGE aggregate existing columns. • VALUES(Column) returns distinct visible values; useful for entity-grain averages. • DIVIDE is safer than / when denominator can be zero or blank. • SAMEPERIODLASTYEAR shifts the current date set one year; DATEADD shifts by a chosen interval. • Variables store results in the context where they are evaluated; they do not magically recalculate after a later context change. • Balances are often semi-additive: aggregate across entities, but use a closing/last-data-date logic across time. • Calculation groups reuse transformations across measures; SELECTEDMEASURE identifies the current measure. • DAX Query View uses queries such as EVALUATE to inspect model results. VISUALIZE & ANALYZE ------------------- • Line = trend; bar/column = category comparison; scatter = numeric relationship; matrix = hierarchy/crosstab. • Gauge = value vs target/range. KPI = status/trend with target over time. • Filter restricts a target visual; Highlight keeps broader context and emphasizes a subset. • Drill down stays within a hierarchy; Drillthrough navigates to a detail page with context. • Report page tooltip shows rich hover content without navigation. • Bookmarks can capture Data, Display and Current page; avoid Data when a bookmark should only show/hide visuals. • Top N requires a value measure; sorting descending alone does not limit rows. • Field parameter switches fields/measures; What-if parameter provides numeric scenario input. • Key influencers ranks factors associated with an outcome; decomposition tree breaks a measure across dimensions. • Q&A = user asks natural language. Smart narrative = Power BI generates text. • Analytics features can include reference lines, forecasting, anomaly/outlier detection, clustering and error bars. • Paginated reports are RDL, authored in Power BI Report Builder, and optimized for pixel-perfect/print/multi-page output. • Visual calculations operate on a visual's result structure and are not reusable semantic-model measures. • Accessibility: alt text, tab order, contrast, labels/icons—never rely on color alone. POWER BI SERVICE, SECURITY & GOVERNANCE --------------------------------------- • Workspace = collaboration/authoring. App = curated distribution to consumers. • Viewer is the typical workspace role when RLS must constrain a user; elevated editors are not constrained the same way. • Build permission allows creating reports/Analyze in Excel from a semantic model; Reshare allows passing access onward. • Dynamic RLS commonly uses a user-access mapping table + USERPRINCIPALNAME(). • Multiple RLS roles are additive (union/OR), not restrictive AND. • OLS protects model objects such as tables/columns; hiding fields is not security. • Dashboard = single-page Service canvas. Tiles are pinned/supported dashboard items and can link back to source reports. • App audiences control which content is visible; RLS controls which data rows are visible. • Gateway provides Service-to-on-prem connectivity; source credentials are separate. • Scheduled refresh updates imported data; automatic page refresh periodically re-queries supported DirectQuery pages. • Subscriptions deliver content on a schedule/condition. Data alerts are threshold-based notifications in supported dashboard scenarios. • Usage metrics = adoption. Performance Analyzer = slow visual/DAX diagnosis. • Lineage/impact analysis = downstream dependencies. Query Dependencies = Power Query query/source dependencies. • Promoted = recommended. Certified = formally reviewed/authoritative. • Sensitivity labels classify/protect content; they do not replace RLS. • Publishing deploys report/model design changes; refresh only updates data in an existing model. • Deployment pipelines support Dev → Test → Prod lifecycle. HIGH-VALUE EXAM GOTCHAS ----------------------- • Hide ≠ remove; hide ≠ security. • Append = rows; Merge = columns. • Refresh ≠ publish. • Gateway online ≠ source credentials valid. • Usage metrics ≠ Performance Analyzer. • Query Dependencies ≠ Lineage view. • Dashboard ≠ report. • App audience ≠ RLS. • Workspace role ≠ semantic-model Build permission. • Q&A ≠ Smart narrative. • Sort descending ≠ Top N. • Drill down ≠ Drillthrough. • Calculated column ≠ measure. • Power Query parameter ≠ What-if parameter ≠ Field parameter. • Paginated report ≠ ordinary interactive PBIX report. • Raw mock percentage ≠ Microsoft's scaled certification score. Reminder: This is an independent study resource. Always compare final preparation with the current official Microsoft PL-300 study guide. Progress on melarks.com is stored locally in the browser and is not synced across devices.