6.1. Exam Strategy

The PL-300's center of gravity is the requirement-driven scenario: a short description of a data source, model, or report, a "You need to…" sentence, and a qualifier that does most of the discriminating. Common qualifiers are minimize model size, minimize development effort, minimize refresh time, ensure users can see only…, and follow the principle of least privilege. Read the qualifier first: several options may work, but only one minimizes what's asked. Model-size qualifiers favor removing columns, reducing cardinality, and summarizing in Power Query; development-effort qualifiers favor built-in features (Column From Examples, quick measures, Unpivot Other Columns, JSON themes, apps); least-privilege qualifiers favor Contributor over Member, Build over Contributor, and Viewer or app access over any edit role.

Time management: with about 40–60 questions in 100 minutes, budget roughly 90 seconds each and bank time on quick recall items. Case studies describe a company's current environment, data sources, report requirements, and security needs across several tabs; skim the questions first, then read only the tabs they need. Within a case study you can review answers before leaving it, but once you leave the case study, you can't return. For standalone questions, use mark for review and return later rather than stalling. Some sections (such as yes/no series that ask whether a solution meets a goal) can't be revisited; the interface warns you before you enter them.

Format-specific technique:

• DAX completion (hotspot): identify the evaluation context first (measure or calculated column? which filters must change?), then the function family (CALCULATE modifier, time intelligence, iterator). Eliminate options that operate on the fact table's date column instead of the date table, and options that change filters when only aggregation is needed.
• Power Query sequencing (drag-drop): anchor the first and last steps. Promote headers and set types early; remove columns and filter rows before expensive steps; unpivot before merging with a date dimension; disable load on staging queries last.
• Multi-select: each selection is graded; pairs often combine a configuration step (define the role in Desktop) with an assignment step (add the group in the service).
• Yes/No solution series: evaluate each proposed solution independently against the stated goal; the same goal appears with different solutions.

When two answers look right, re-read the stem for the storage mode (Import, DirectQuery, Direct Lake), the license or capacity (Pro, PPU, Fabric), who performs the action (author in Desktop or admin in the service), and whether the requirement is about seeing data (RLS, roles) or changing data (Power Query, model).

The night-before list: merge vs. append and the join kinds; reference vs. duplicate; Unpivot Other Columns; storage modes and Direct Lake fallback; CALCULATE overwrite vs. KEEPFILTERS; SAMEPERIODLASTYEAR and the marked date table; semi-additive LASTNONBLANK; bookmark options; drillthrough vs. drill down; sync slicers (sync vs. visible); workspace role matrix; RLS only for Viewers; Build permission; gateway modes; data alert tile types.