Quiz DP-600: Microsoft Implementing Analytics Solutions Using Microsoft Fabric

The correct answer is defining DAX filter expressions in RLS roles. Row-level security in Fabric semantic models
uses DAX expressions within roles to dynamically filter data based on the user's identity and context. This
allows different users to see different subsets of the same data without creating separate models or tables. You
would typically use functions like USERNAME() or USERPRINCIPALNAME() combined with a lookup table to
map users to their regions, then filter the fact table accordingly.
The other options are incorrect because: creating separate models is inefficient and difficult to maintain;
column-level security restricts column access, not row access; object-level security prevents access to entire
objects, not filtered subsets; and separating fact tables by region defeats the purpose of a unified semantic
model and doesn't properly implement RLS.

The correct answers are implementing a star schema, aggregating inventory data, and creating views. For
analytics workloads, a star schema provides the best balance of query performance and maintainability.
Creating aggregations for inventory snapshots is appropriate for supporting historical analysis and reducing
query load. Views can provide a logical star schema layer on top of normalized data, offering flexibility without
requiring full denormalization.
Denormalizing everything into a single flat table would create severe data redundancy, make updates difficult,
and actually hurt performance on complex queries due to table size. A fully normalized structure without
analytical optimization is not ideal for reporting, which is why the star schema approach is preferred.

The correct answers are enabling Direct Lake on OneLake, implementing incremental refresh, and creating a
composite model. Direct Lake is the modern optimization approach for large semantic models in Fabric,
allowing queries to execute against Parquet files directly, dramatically reducing memory footprint and
improving query speed. Incremental refresh reduces refresh time and load on the system by processing only
deltas. Composite models allow strategic use of different storage modes—Direct Lake for base tables and
Import for pre-aggregated tables—to optimize for both query performance and flexibility.
DirectQuery mode would push all queries to the source database, which can create bottlenecks and doesn't
solve the problem of concurrent user load. Simply increasing capacity without architectural improvements is
not cost-effective and doesn't address the fundamental performance bottlenecks. Direct Lake is preferred over
DirectQuery for large analytics models because it leverages the Parquet format and eliminates repeated
database hits.

The correct answer demonstrates understanding of Fabric's data connection and ingestion patterns. Dataflow
Gen2 with native connectors is the standard approach for SQL Server ingestion with scheduled refresh,
balancing latency and operational overhead. OneLake shortcuts for CSV files leverage the lakehouse's
OneLake integration without duplicating data. Dataflow Gen2 with Web connector is the recommended
pattern for REST API ingestion within Fabric's managed environment, supporting scheduled refresh and
monitoring.
The first option incorrectly suggests Warehouse with Direct Lake for a sync scenario (Direct Lake is for query
performance on existing medallion data, not for ingestion), and tries to use shortcuts for a real-time API which
won't work. The third option introduces unnecessary complexity with custom notebooks and Data Factory when
native Fabric connectors are available and simpler. The fourth option suggests direct shortcuts to a database
(not supported) and relies on Power BI Desktop rather than Fabric's data preparation tools.

The correct answer combines several advanced Fabric concepts appropriately. Bridge tables are the
recommended pattern for many-to-many RLS scenarios (Finance managers to multiple cost centers). Large
Semantic Model storage format with Direct Lake on OneLake provides the best performance for 500GB
datasets with 200 concurrent users by querying columnar storage directly without import overhead.
Incremental refresh on the fact table minimizes processing time and resource consumption for large models.
The second option uses object-level security (which controls who sees entire tables, not rows) rather than RLS,
and DirectQuery mode would cause performance issues at this scale. The third option relies on source-level
filtering which doesn't scale well for RLS and compresses data without Direct Lake, losing the main
performance benefit. The fourth option suggests SQL analytics endpoint over OneLake (though both are valid,
OneLake is simpler and preferred for Fabric-native scenarios) and mischaracterizes Direct Lake
configuration—detection of data changes is not a Direct Lake feature.

The correct answer implements the key Fabric features for managing this scenario. Impact analysis from the
dataflows feature reveals all downstream dependencies (reports, other dataflows, semantic models), enabling
proactive identification of affected items before changes. Deployment pipelines with test and production
stages allow validation before changes reach production users. Power BI Desktop projects (.pbip) in Git
provide version control and history for the semantic model with ability to revert breaking changes.
The first option distributes multiple copies of the semantic model (.pbit files) which defeats the purpose of a
shared model and makes changes even harder to track. The third option relies on manual documentation which
doesn't scale and doesn't prevent issues. The fourth option references OneLake Data Governance app (which
monitors data lineage but doesn't prevent dataflow issues), and suggests DirectQuery which adds complexity
without addressing the core dependency visibility problem. Workspace-level access controls don't provide the
granular visibility needed.
CASE STUDY 1
Overview
Litware. Inc. is a manufacturing company that has offices throughout North America. The analytics team at Litware
contains data engineers, analytics engineers, data analysts, and data scientists.
Existing Environment
litware has been using a Microsoft Power Bl tenant for three years. Litware has NOT enabled any Fabric capacities and
features.
Fabric Environment
Litware has data that must be analyzed as shown in the following table.

Tap the image to zoom
The Product data contains a single table and the following columns.

Tap the image to zoom
The customer satisfaction data contains the following tables:
* Survey
* Question
* Response
For each survey submitted, the following occurs:
* One row is added to the Survey table.
* One row is added to the Response table for each question in the survey.
The Question table contains the text of each survey question. The third question in each survey response is an overall
satisfaction score. Customers can submit a survey after each purchase.
User Problems
The analytics team has large volumes of data, some of which is semi-structured. The team wants to use Fabric to
create a new data store.
Product data is often classified into three pricing groups: high, medium, and low. This logic is implemented in several
databases and semantic models, but the logic does NOT always match across implementations.
Planned Changes
Litware plans to enable Fabric features in the existing tenant. The analytics team will create a new data store as a
proof of concept (PoC). The remaining Litware users will only get access to the Fabric features once the PoC is
complete. The PoC will be completed by using a Fabric trial capacity.
The following three workspaces will be created:
* AnalyticsPOC: Will contain the data store, semantic models, reports, pipelines, dataflows, and notebooks used to
populate the data store
* DataEngPOC: Will contain all the pipelines, dataflows, and notebooks used to populate Onelake
* DataSciPOC: Will contain all the notebooks and reports created by the data scientists
The following will be created in the AnalyticsPOC workspace:
* A data store (type to be decided)
* A custom semantic model
* A default semantic model
* Interactive reports
The data engineers will create data pipelines to load data to OneLake either hourly or daily depending on the data
source. The analytics engineers will create processes to ingest transform, and load the data to the data store in the
AnalyticsPOC workspace daily. Whenever possible, the data engineers will use low-code tools for data ingestion. The
choice of which data cleansing and transformation tools to use will be at the data engineers' discretion.
All the semantic models and reports in the Analytics POC workspace will use the data store as the sole data source.
Technical Requirements
The data store must support the following:
* Read access by using T-SQL or Python
* Semi-structured and unstructured data
* Row-level security (RLS) for users executing T-SQL queries
Files loaded by the data engineers to OneLake will be stored in the Parquet format and will meet Delta Lake
specifications.
Data will be loaded without transformation in one area of the AnalyticsPOC data store. The data will then be cleansed,
merged, and transformed into a dimensional model.
The data load process must ensure that the raw and cleansed data is updated completely before populating the
dimensional model.
The dimensional model must contain a date dimension. There is no existing data source for the date dimension. The
Litware fiscal year matches the calendar year. The date dimension must always contain dates from 2010 through the
end of the current year.
The product pricing group logic must be maintained by the analytics engineers in a single location. The pricing group
data must be made available in the data store for T-SQL queries and in the default semantic model. The follo

For ingesting customer data into the data store in the AnalyticsPOC workspace, a dataflow (D) should be
recommended. Dataflows are designed within the Power BI service to ingest, cleanse, transform, and load data
into the Power BI environment. They allow for the low-code ingestion and transformation of data as needed by
Litware's technical requirements. Reference = You can learn more about dataflows and their use in Power BI
environments in Microsoft's Power BI documentation.