How to pass Microsoft Power BI Data Analyst (PL-300)
18 min read4 domains coveredFree practice, no sign-up
The Microsoft Power BI Data Analyst (PL-300) tests whether you can take raw data and turn it into a trustworthy, performant, shareable Power BI solution end to end. The work splits into four stages: connecting to and cleaning data in Power Query, building a sound star-schema model with DAX, designing reports that answer the business question, and deploying those assets safely in the Power BI service with refresh, endorsement, and row-level security. The exam wants you to do the analyst's whole job, not just one slice of it.
It suits practising data analysts, BI developers, and report builders who already work in Power BI and want to prove it formally. The questions assume you have actually used Power Query Editor, written DAX measures, set relationships, and published to a workspace. You do not need to be a data engineer, but you do need to know why a measure beats a calculated column, why a star schema beats a flat table, and which workspace role can publish an app.
What makes PL-300 pass-or-fail is precision on the small documented behaviours, not broad familiarity. Many items hinge on one exact detail: that a multi-column merge aligns by selection order rather than column name, that only gauges, KPIs, and cards support data alerts, that Member is the least-privileged role that can publish a new app, or that you must take over a semantic model before you can update its credentials. Knowing roughly how Power BI works is not enough; the exam rewards knowing the rule the documentation actually states.
PL-300 is the full analyst workflow tested on documented specifics: prepare in Power Query, model with a star schema and DAX, visualise for the audience, and deploy with refresh and row-level security, and the right answer is almost always the one exact behaviour Microsoft's docs describe.
Difficulty
Intermediate
Best for
Working data analysts, BI developers, and report authors who already build in Power BI Desktop and publish to the service, and want an associate-level credential that proves they can prepare, model, visualise, and deploy data the way Microsoft documents it.
Prerequisites
None enforced. In practice you want real hands-on time in Power Query Editor, writing DAX measures and time intelligence, setting relationships, and publishing to a workspace with refresh and row-level security. Familiarity with the Power BI service, not just Desktop, is what carries the deploy domain.
Typically 40 to 60 questions
Questions
100 min
Time allowed
700 / 1000
Pass mark
$165
Exam cost (USD)
224
Practice questions
How this exam thinks
One habit decides PL-300: read the scenario for what the analyst is trying to achieve, then pick the option that matches the exact behaviour Microsoft documents, not the one that merely sounds plausible. The distractors are written to be reasonable. They invent settings that do not exist, swap one role or licence for another, or describe a feature working the way you might assume rather than the way it actually does. The right answer is the precise documented rule.
The exam frames most questions as a small task with a constraint: relate two dimensions in a many-to-many association, reduce an Import model that is too large, give a report author the least privilege needed to publish an app, or protect sensitive data in a subscription email. Each has a single best-practice answer that Microsoft recommends, and the test rewards knowing that recommendation. When a question stresses best practice or the recommended pattern, reach for the canonical answer: a bridging table for many-to-many, a measure over a calculated column, a star schema over a flat table, pre-summarising during ETL for the biggest size reduction, the least-privileged role that still does the job.
The rest is a set of discriminations the exam leans on, each resolved by one detail. Measure versus calculated column turns on where the calculation runs and context transition; Member versus Contributor turns on who can publish a new app; gauges, KPIs, and cards versus everything else turns on what supports data alerts; the standalone Copilot versus the report pane turns on scope. Name what the scenario asks for, then choose the option whose documented behaviour fits it exactly.
What each domain tests and how to study it
The PL-300 blueprint is split across 4 domains. Weights are the official share of the exam; see the official exam guide for the authoritative breakdown.
What you must be able to do. Connect to the right source, shape and clean the data correctly in Power Query, profile it to judge quality, and apply the documented service rules for endorsement, ownership, and credentials.
In one sentenceThe preparation layer: pull data from the right source, clean and shape it in Power Query, and respect the service's ownership and endorsement rules.
Recall check: answer these from memory first
In a multi-column merge, what aligns the join keys between the two tables, and what happens if you select the columns in a different order on each side?
Which Power Query feature stacks rows from a second query, and which one brings in columns from a second query on a matching key?
A semantic model has one owner and a coworker needs to update its cloud-source credentials. What must the coworker do first?
What it tests. Getting data in and making it clean before any modelling starts. Connecting to sources across files, databases, and cloud services; shaping data in Power Query Editor with merges, appends, type changes, and applied steps; and profiling data to judge quality, structure, and distribution. It also covers the service side of preparation: dataflows, the OneLake data hub, endorsement badges, and the ownership and credential rules that govern a shared semantic model.
How to study it. Spend real time in Power Query Editor rather than reading about it, because the questions test documented behaviours you only remember by doing. Practise multi-column merges and watch that the join aligns by the order you select columns, not by name. Learn the difference between merge (combine columns from another query on a key) and append (stack rows). Use column quality, column distribution, and column profile to assess data, and know what each pane reports. On the service side, learn the discovery experience: where the certifier name appears on an endorsed model, and the rule that a semantic model has one owner so a non-owner must take over the model before updating credentials.
Easy to confuse
Merge queries versus Append queries. Merge joins two queries on a key column to bring extra columns alongside existing rows, like a SQL join; append stacks the rows of one query under another with the same columns, like a UNION. Merge widens the table; append lengthens it.
Power Query Online (dataflows) versus Power Query in Desktop. They are two experiences of the same engine and Microsoft documents them as providing almost the same user experience in every scenario; the difference is where they run, Online in the service for dataflows and Desktop for local modelling, not what the shaping interface can do.
Worked example from the PL-300 bank
lock_openFree samplePrepare the Datahard
An analyst configures a multi-column merge between two tables by Ctrl-clicking to pick the join columns. Which behaviour does the Merge queries article describe for the column selection order?
AThe columns must have identical names in both tables for the merge to find matches correctly.
BThe columns can be selected in any order on each side; Power Query auto-aligns them by data type compatibility.
CThe order in which columns are selected in each table must match between the left and right tables.check_circle Correct
DThe columns must be sorted ascending on both tables before opening the Merge dialog or join keys mismatch.
Multi-column merges align by selection order, not by column name; clicks must be in the same order on each side. The Merge queries article warns that when multiple columns are selected for a join, the order of selection in each table must match, because the first selected column on the left is matched with the first on the right, and so on. Otherwise the merge can yield incorrect results.
Why A is wrong: Identical column names are not required; the article explicitly notes that matching column names is not a requirement for the merge operation.
Why B is wrong: Power Query does not auto-align multi-column joins. The first selection on the left matches the first selection on the right, and so on; selection order is the alignment rule.
Why C is correct: Correct. The Merge queries article warns that when multiple columns are selected for a join, the order of selection in each table must match, because the first selected column on the left is matched with the first on the right, and so on.
Why D is wrong: Pre-sorting is not a requirement of the Merge dialog. Match alignment uses selection order, not sort order.
What you must be able to do. Build a correct star-schema model, choose the right cardinality and relationship pattern, write DAX measures and time intelligence with correct context behaviour, and reduce model size with the most effective lever.
In one sentenceThe modelling core: a clean star schema, the right cardinality, correct DAX with context transition, and the biggest size lever for an oversized model.
Recall check: answer these from memory first
Salespeople map to many regions and regions to many salespeople. What is the recommended modelling pattern, and what does that table store?
Connecting a dimension with a unique key to a fact with duplicate keys, what cardinality does Power BI assign by default and which table is on the one side?
For an 800-million-row Import model where only monthly totals are needed, which single technique gives the largest model-size reduction and why?
What it tests. The heaviest domain by weight: turning prepared tables into a model that is correct, fast, and easy to calculate over. Designing relationships and a star schema, including how to handle a many-to-many association with a bridging table and what cardinality Power BI assigns by default; writing DAX measures, calculated columns, and time intelligence, including how context transition works; and optimising the model by reducing cardinality, choosing storage modes, and pre-summarising data to cut size.
How to study it. This is 35 percent of the exam, so it earns the most time, and most of it is two skills. First, model design: drill the star schema until it is automatic, relate dimensions to facts on a unique key (the default is one-to-many with the dimension on the one side), and use a bridging table to relate two dimensions in a many-to-many. Second, DAX: learn the measure-versus-calculated-column distinction cold and understand context transition, that a model measure used in a row context transitions automatically while an inline aggregation like SUM needs an explicit CALCULATE. For optimisation, rank the size levers and know that pre-summarising during ETL to raise the fact grain gives the largest single reduction; disabling Auto date/time and removing unused columns help less.
Easy to confuse
Measure versus calculated column. A measure is evaluated at query time in filter context and stored as a formula, ideal for aggregations that respond to slicers; a calculated column is computed row by row at refresh and stored in the model, widening it. Reaching for a calculated column when a measure would do bloats the model and is the common trap.
Bridging table versus a direct many-to-many relationship. The recommended pattern for two dimensions in a many-to-many association is a bridging (factless fact) table holding both keys, with each dimension joined one-to-many to the bridge; a direct many-to-many relationship works but is the less robust option Microsoft does not recommend as the default.
Worked example from the PL-300 bank
lock_openFree sampleModel the Datamedium
An Import model loads 800 million order-line rows. Reporting needs are limited to monthly revenue by product and region; nobody needs day-level or order-line detail. Which data reduction technique gives the largest model-size reduction?
ARemove unused columns from the fact table to reduce per-row width.
BFilter the date range to the most recent five years of order-line history.
CDisable Auto date/time to remove the hidden per-column date hierarchies.
DGroup rows during ETL to load pre-summarised month-level data into the fact table.check_circle Correct
Pre-summarising during ETL raises the fact grain and is the largest single lever on Import model size. Loading pre-summarised data is described as perhaps the most effective technique to reduce a model size. Grouping by date at the month level can deliver near-99% size reduction when day-level detail is not required, at the cost of losing the lower-grain detail.
Why A is wrong: Vertical filtering helps, but removing columns from an 800 million row table delivers less reduction than collapsing the row count to monthly grain.
Why B is wrong: Time-based horizontal filtering helps, but keeping order-line grain still leaves hundreds of millions of rows in scope.
Why C is wrong: Disabling Auto date/time saves some space, but it does not collapse the fact table; it removes hidden helper tables.
Why D is correct: Correct. Loading pre-summarised data is described as perhaps the most effective technique to reduce a model size.
What you must be able to do. Pick the visual that matches the analytical intent, enhance reports for usability and accessibility, and apply the correct Copilot scope and licensing rules.
In one sentenceThe visualisation layer: the right visual for the question, accessible and well-formatted reports, and the documented scope and licensing of Copilot.
Recall check: answer these from memory first
Which visual does Microsoft document for revealing correlation and clusters between two numerical measures across many items?
What capacity and licence does Copilot in Power BI require, and which common options (trial SKU, Pro-only) do not qualify?
A user wants to ask Copilot questions spanning multiple reports and models they can access. Which Copilot surface covers that, and which one is limited to the open report?
What it tests. Turning the model into reports that answer the business question and surface insight. Choosing the right visual for the data and audience, such as a scatter chart to reveal correlation and clusters between two measures; enhancing reports for usability, storytelling, and accessibility with formatting, themes, and layout; and identifying patterns with analytics features and AI visuals, including the Copilot experiences and their licensing and scope rules.
How to study it. Build a visual-selection reflex: match the analytical intent to the documented visual, scatter for correlation between two numerical measures, and know when a card, matrix, or chart fits instead. Learn the accessibility and storytelling tools, themes, alt text, tab order, and bookmarks, as the answer to usability requirements. For the AI and Copilot questions, learn the two facts the exam tests: the licensing floor, that Copilot needs a paid Fabric F2 or higher or Power BI Premium P1 or higher (trial and Pro alone do not qualify), and the scope difference, that the standalone Copilot reaches any accessible item while the report-pane Copilot is scoped to the open report. Note that Q&A is being retired in favour of Copilot for natural-language querying.
Easy to confuse
Standalone Copilot versus the report Copilot pane. The standalone Copilot, opened from the left navigation, finds and analyses any report, semantic model, or Fabric data agent the user can access; the report Copilot pane on the right of an open report is scoped only to that report. Cross-item questions need the standalone experience.
Scatter chart versus stacked column chart. A scatter chart plots two numerical measures against each other to reveal correlation and clustering across many data points; a stacked column chart compares category totals broken into segments. Correlation between price and volume is the scatter chart's documented job, not the column chart's.
Worked example from the PL-300 bank
lock_openFree sampleVisualize and Analyze the Datamedium
A workspace admin wants to enable Copilot in Power BI for report authors in a single workspace. Which capacity-licensing combination satisfies the documented requirement?
AGrant every report author a Power BI Pro licence and leave the workspace on shared capacity.
BAssign the workspace to a Fabric trial SKU and enable the Copilot tenant setting.
CGrant every report author a Premium Per User licence in the default My Workspace context.
DAssign the workspace to a paid Fabric F2 (or higher) or Power BI Premium P1 (or higher) capacity.check_circle Correct
Copilot for Power BI requires paid Fabric F2 or higher, or Power BI Premium P1 or higher; trial and Pro-only do not qualify. Copilot requires a paid Fabric capacity (F2 or higher) or Power BI Premium (P1 or higher) on the workspace. Trial SKUs and Pro or PPU licences alone are explicitly excluded.
Why A is wrong: A Pro licence alone is not sufficient; Copilot requires organisational paid capacity at the workspace level.
Why B is wrong: Copilot in Microsoft Fabric is not supported on trial SKUs or trial capacities, only on paid SKUs.
Why C is wrong: PPU is not sufficient on its own; the workspace must be linked to paid Fabric F2 or higher, or Power BI Premium P1 or higher.
Why D is correct: Correct. Copilot requires a paid Fabric capacity (F2 or higher) or Power BI Premium (P1 or higher) on the workspace.
What you must be able to do. Choose the least-privileged role and correct licence for a deployment task, configure refresh and endorsement, protect sensitive data, and implement row-level security correctly.
In one sentenceThe deployment layer: the right role and licence, working refresh and endorsement, sensitivity-label limits, and row-level security that restricts data by identity.
Recall check: answer these from memory first
Which is the least-privileged workspace role that can publish a brand-new app, and what can the Contributor role not do here?
Which licence does a user need before the New workspace command appears in the service?
Which dashboard tile types support data alerts, and where must those tiles have been pinned from?
What it tests. Getting assets into the service and keeping them running safely. Creating and managing workspaces, apps, and deployment pipelines, including which workspace role can do what and which licence lets a user create a workspace; managing semantic models with scheduled refresh, endorsement, and query caching; protecting data through sensitivity labels and subscriptions; and implementing row-level security to restrict data by user identity. Data alerts and dashboard tiles also live here.
How to study it. Memorise the workspace role matrix, because the exam tests it directly: Admin and Member can publish, unpublish, and reassign an app, so Member is the least-privileged role that can publish a new app; Contributor can update a delegated app but not publish the first version; Viewer only consumes. Learn the licensing floor for collaborative workspace creation (Power BI Pro or Premium Per User, free users see no New workspace). For row-level security, practise defining roles with DAX filters and testing them with View as. Learn the documented edge cases the exam loves: sensitivity labels apply to subscription attachments but not the email body or preview image, and data alerts work only on gauges, KPIs, and cards pinned from report visuals.
Easy to confuse
Member role versus Contributor role. Member can publish, unpublish, and change permissions for an app and is the least-privileged role that can publish a new app; Contributor can update an app only when explicitly delegated and cannot publish the first version. The gap is publishing rights, not editing content.
Sensitivity label on a subscription attachment versus the preview image or email body. Sensitivity labels apply to subscription email attachments but not to the email body text or the embedded preview image; to protect sensitive data in the email itself you must turn off the preview image, since the label will not cover it.
Worked example from the PL-300 bank
lock_openFree sampleDeploy and Maintain Assetsmedium
An analyst with the Contributor role tries to publish a brand-new app from a workspace and finds the Publish app command is unavailable. Which role grants the least privilege required to publish a new app?
AAssign the analyst the Contributor role on the workspace.
BAssign the analyst the Member role on the workspace.check_circle Correct
CAssign the analyst the Viewer role on the workspace.
DAssign the analyst a tenant-level Power BI admin role to publish the app.
Member is the least-privileged workspace role that can publish or unpublish an app for the first time. The workspace role matrix shows that publishing, unpublishing, and changing permissions for an app is a Yes for Admin and Member only. Contributors can update an app when explicitly delegated, but cannot publish the first version.
Why A is wrong: Contributors can update an app only if a workspace Admin has delegated update rights; they cannot create or publish an app for the first time.
Why B is correct: Correct. The workspace role matrix shows that publishing, unpublishing, and changing permissions for an app is a Yes for Admin and Member only.
Why C is wrong: Viewers can only read and interact with items; they have no app-publishing capability whatsoever.
Why D is wrong: Tenant admin roles oversee policy; they are not the least privilege for publishing an app from a single workspace.
A study plan that works
Map the blueprint and book a date
Day 1
Read the official PL-300 skills outline and the four domains with their weights. Book a provisional date now, because a fixed date turns open-ended study into a plan and is the strongest predictor of actually sitting. Note that Model the Data is 35 percent on its own and Prepare plus Model together are most of the exam, so weight your time accordingly.
Drill Power Query hands-on
Week 1
Spend the first week inside Power Query Editor. Practise connecting to files, databases, and cloud sources; do multi-column merges and confirm the selection-order rule; tell merge from append by building both; and use column quality, distribution, and profile to assess data. Learn the service-side rules on endorsement, single-owner semantic models, and taking over a model to change credentials.
Go deep on modelling and DAX
Weeks 2 to 3
Model the Data is the largest domain, so it gets the most time. Build star schemas, set relationships, and learn the default one-to-many cardinality and the bridging-table pattern for many-to-many. Write measures and time intelligence, and nail context transition: a model measure transitions automatically in row context, an inline SUM needs CALCULATE. Rank the model-size levers and know pre-summarising during ETL is the biggest.
Build visualisation and Copilot fluency
Week 4
Practise matching visuals to intent (scatter for correlation, card and matrix where they fit) and enhancing reports with themes, accessibility, and bookmarks. Learn the Copilot facts the exam tests: the F2 or P1 licensing floor, the standalone-versus-report-pane scope difference, and that Q&A is being replaced by Copilot for natural-language querying.
Master deploy, security, and the role matrix
Week 5
Memorise the workspace role matrix until Member-versus-Contributor publishing rights are automatic, and the Pro or PPU licence floor for creating a workspace. Configure scheduled refresh and endorsement, implement and test row-level security with View as, and learn the documented edge cases on sensitivity labels in subscriptions and which tiles support data alerts.
Drill weak domains, then space the review
Week 6
Use your per-domain accuracy to attack the domains dragging you down rather than re-reading what you already know. Then space it: revisit each domain's recall prompts after a few days and again a week later. Spacing roughly doubles what sticks compared with cramming the night before.
Sit a timed mock and calibrate
Weeks 6 to 7
Take at least one full timed mock under exam conditions to rehearse pacing and the flag-and-return habit. Treat the score as a per-domain readiness signal, not a single number, and review every missed question, naming the exact documented rule you misread, before you book or sit.
Know when you're ready
Readiness for PL-300 is a measured score on scenario questions you have not seen before, not a feeling that Power BI is familiar. Those are different things, and the gap is where people fail. Clicking through reports all day builds fluency, and fluency feels like knowledge, so confidence rises while precise recall does not. The fix is to test yourself: if you can read a fresh scenario, name what the analyst needs, and pick the option whose documented behaviour fits while explaining why each other option is wrong, you know it; if you can only nod along to an explanation, you do not yet.
Be especially wary of early confidence on modelling and the service rules. Knowing what a measure or a workspace role is feels like enough, but the exam tests the exact distinction, measure versus calculated column, Member versus Contributor, which tiles support alerts, and those are the items people drop. Trust your measured per-domain accuracy over your gut, and set the bar at clearing every domain comfortably on unseen questions across more than one session, with Model the Data, the heaviest domain, well above a bare pass before you book.
Ready to put this into practice?
Free PL-300 questions with worked explanations. No sign-up.
Read the scenario for what the analyst needs first, then match it to the exact documented behaviour. Distractors are written to sound reasonable; the right answer is the precise rule Microsoft's docs state, not the one that feels intuitive.
When a question says best practice or recommended, reach for the canonical pattern: a star schema over a flat table, a measure over a calculated column, a bridging table for many-to-many, and pre-summarising during ETL for the biggest model-size cut.
On any role or licence question, pick the least-privileged role and the minimum licence that still does the job. Member, not Admin, publishes a new app; Pro or PPU, not free, creates a workspace.
For DAX, decide measure versus calculated column by where it runs: query-time aggregation that responds to filters is a measure; a per-row value stored at refresh is a calculated column, and it widens the model.
Watch the documented edge cases the exam loves: data alerts only on gauges, KPIs, and cards pinned from report visuals; sensitivity labels cover subscription attachments but not the email body or preview image.
For Copilot questions, separate scope from licensing: the standalone Copilot spans any accessible item while the report pane is scoped to the open report, and Copilot needs paid Fabric F2 or Premium P1 and above.
Flag and move on. Cover every question once before you sink time into a hard one, so you collect the clear marks first and protect the items you actually know.
Frequently asked questions
Is the PL-300 hard?
It is an intermediate, associate-level exam, and the difficulty is precision rather than breadth. Knowing roughly how Power BI works is not enough; many items hinge on one documented detail, like the multi-column merge selection-order rule or which workspace role can publish a new app. Scenario practice with worked explanations matters far more than re-reading feature lists.
How long should I study for the PL-300?
Most candidates with real Power BI experience are ready in five to seven weeks of steady study. Less hands-on time means more weeks on the two heavy areas, modelling and DAX, and on the service-side deployment rules that people who live in Desktop tend to miss.
Do I need to know DAX for this exam?
Yes. Model the Data is the largest domain and DAX sits at its centre. You need to write measures and time intelligence, tell a measure from a calculated column, and understand context transition, that a model measure transitions automatically in row context while an inline aggregation needs CALCULATE. You do not need to be a DAX expert, but you cannot skip it.
Do I need to know the Power BI service, not just Desktop?
Yes, and it is where many candidates lose marks. The deploy domain tests workspace roles, app publishing, licensing for workspace creation, scheduled refresh, endorsement, sensitivity labels, and row-level security in the service. If you only build in Desktop and never publish, spend extra time in a workspace before you sit.
Which domains should I focus on?
Model the Data at 35 percent is the single biggest domain and deserves the most time, with Prepare the Data at 24 percent close behind. Visualise and Analyse at 23 percent and Deploy and Maintain at 18 percent round it out, but the deploy domain punches above its weight on precise rules, so do not leave it short.
How is the exam scored and what should I aim for?
Power BI exams are scored on a scale and reported as a pass or fail against the published bar, shown in the facts panel above. Because individual question weights are not visible, aim to clear every domain comfortably on unseen practice questions rather than chasing one raw figure, with the heaviest domain well above a bare pass.
How many practice questions should I do before booking?
Enough that every domain clears comfortably on questions you have not seen and a full timed mock feels comfortable on pacing. Quality of review beats raw volume: on every question, read the explanation and name the documented rule that picked the answer, including on the ones you got right.
Will the exam test Copilot and other newer features?
Yes. The current blueprint includes AI features, so know the Copilot facts the exam tests: the paid Fabric F2 or Premium P1 licensing floor, the standalone-versus-report-pane scope difference, and that Q&A is being replaced by Copilot for natural-language querying. Treat these as documented rules to memorise, not optional extras.
Is the Power BI Data Analyst certification worth it?
PL-300 is worth it for analysts, BI developers, and data professionals who work in organisations running Microsoft 365 or Azure, where Power BI is already the standard reporting layer. The certification validates skills across the full pipeline from data preparation and modelling through to deployment and access control, which are exactly the skills that translate into visible, day-to-day work. It is also a natural progression point for those who already hold a Microsoft fundamentals or associate badge and want to specialise on the data side.
Examworthy is not affiliated with or endorsed by Microsoft. This guide is original study material based on the public exam blueprint. We never reproduce live exam items. PL-300 and related marks belong to their respective owners.
