Home / Data Analyst practice test / Understanding of Databricks Data Intelligence Platform

Free · 5 questions with explanations

Understanding of Databricks Data Intelligence Platform: Databricks Data Analyst Associate Practice Questions

Exam-style questions on Understanding of Databricks Data Intelligence Platform. Pick your answer, then open the explanation to see why it's right — and why the other options are wrong.

Updated November 10, 2026

1 Understanding of Databricks Data Intelligence Platform

A solutions architect is designing a new data platform on Databricks. She has three distinct requirements: (1) ingest streaming events from Apache Kafka and declaratively apply multi-hop transformations before writing clean data to Delta tables, (2) enforce fine-grained column-level access controls and capture data lineage across all data assets, and (3) schedule and orchestrate a nightly batch aggregation job that depends on the successful completion of the streaming pipeline. Which combination of Databricks platform components BEST maps to these three requirements, in order?

  1. ARequirement 1: Databricks SQL, which runs streaming SQL queries on SQL warehouses to ingest Kafka events and apply multi-hop transformations; Requirement 2: Delta Lake, which has built-in row-level and column-level access policies that integrate with LDAP directories and record lineage in the transaction log; Requirement 3: Mosaic AI, which provides MLOps-style orchestration for scheduling batch aggregation pipelines as recurring training jobs.
  2. BRequirement 1: Lakeflow Spark Declarative Pipelines, a declarative SQL/Python framework for multi-hop batch and streaming pipelines with a Kafka source connector; Requirement 2: Unity Catalog, which enforces fine-grained access with column masks and row filters and captures column-level lineage; Requirement 3: Lakeflow Jobs, which orchestrates workflows with task dependencies and starts runs on cron schedules or on upstream task success.
  3. CRequirement 1: Lakeflow Jobs, which consumes Kafka events in real time and applies streaming transformations through long-running background tasks; Requirement 2: Mosaic AI, which manages access permissions and lineage for features and tables through its Feature Store governance layer; Requirement 3: Delta Lake, which has a built-in cron scheduler that starts OPTIMIZE and aggregation jobs when new data partitions are detected.
  4. DRequirement 1: Unity Catalog, which has built-in Kafka ingestion connectors that route incoming messages to the right schema according to access control policies; Requirement 2: Databricks SQL, which offers a visual column masking editor in the SQL Editor and generates lineage diagrams from query patterns; Requirement 3: Lakeflow Spark Declarative Pipelines, the recommended scheduler for nightly compaction and aggregation batches.
  5. ERequirement 1: Delta Lake, which provides native Apache Kafka consumer clients that ingest streaming events and apply schema enforcement before writing transactions; Requirement 2: Lakeflow Jobs, which manages column-level security by running identity-aware job clusters that restrict columns to the job owner's permissions; Requirement 3: Unity Catalog, which has a built-in cron trigger that schedules nightly aggregation workflows on registered tables.
Show answer & explanation

Correct answer: B

WHY B: Each requirement maps to the described component. (1) Lakeflow Spark Declarative Pipelines (formerly Delta Live Tables) is a declarative framework for batch and streaming pipelines in Python/SQL. It includes connectors for sources such as Apache Kafka, Amazon Kinesis, and Google Pub/Sub, and supports multi-hop transformations through Bronze, Silver, and Gold layer patterns. (2) Unity Catalog is the unified governance layer that enforces column-level masking and row-level security through dynamic views and table policies, and it tracks column-level lineage, capturing which upstream tables, views, and notebooks produced each output column. (3) Lakeflow Jobs is the workflow orchestration service that schedules and runs tasks (notebooks, pipelines, SQL queries, Python scripts) with dependency chains, cron triggers, and retry logic. WHY NOT A: Databricks SQL is designed for BI analytics queries and dashboards, not for stateful streaming ingestion from Kafka; SQL warehouses are not long-running Kafka consumers. Delta Lake is a storage format with transaction support and has no access policy enforcement engine; that role belongs to Unity Catalog. Mosaic AI is the AI/ML platform, not a workflow scheduling service. WHY NOT C: Lakeflow Jobs is an orchestration service for scheduling tasks; it does not act as a Kafka streaming consumer or apply streaming transformations. Mosaic AI's Feature Store manages ML features, not general-purpose table access controls. Delta Lake has no built-in cron scheduler; scheduling is handled by Lakeflow Jobs. WHY NOT D: Unity Catalog is a governance layer and has no Kafka connectors or streaming ingestion capabilities. Databricks SQL does not provide a visual column masking editor that generates lineage; column masking is configured through Unity Catalog policies. Lakeflow Spark Declarative Pipelines is a pipeline framework, not a job scheduler for triggering batch compaction runs. WHY NOT E: Delta Lake is a storage layer format (ACID transactions, time travel, schema enforcement) and has no Kafka consumer clients; the ingestion framework with Kafka connectivity is Lakeflow Spark Declarative Pipelines. Lakeflow Jobs applies cluster-level access modes but does not enforce column-level security, which is Unity Catalog's role. Unity Catalog is a governance system, not a pipeline scheduler.

2 Understanding of Databricks Data Intelligence Platform

A data engineer creates a managed Delta table named analytics.sales.transactions in Unity Catalog by running CREATE TABLE analytics.sales.transactions AS SELECT * FROM raw.events. A second engineer independently creates an external table named finance.reporting.invoices using CREATE EXTERNAL TABLE finance.reporting.invoices LOCATION 's3://company-data/invoices/'. Both engineers later run DROP TABLE on their respective tables. Which statement CORRECTLY describes the outcome of both DROP operations?

  1. ABoth the managed and external table DROP operations delete the table's metadata entries from the Unity Catalog metastore AND permanently delete the underlying data files from cloud storage, because registering any table in Unity Catalog — whether managed or external — transfers full data lifecycle ownership to Unity Catalog, which then controls deletion behavior identically for both table types.
  2. BDropping the managed table analytics.sales.transactions removes its metadata AND, after a recovery period (7 days by default), deletes its data files. Dropping the external table finance.reporting.invoices removes only the metadata pointer; the files at s3://company-data/invoices/ stay intact.
  3. CBoth DROP operations remove only the metadata entry from the Unity Catalog metastore; neither operation ever deletes the underlying data files, because Unity Catalog operates exclusively at the governance layer and delegates all data file lifecycle management to the cloud provider's storage service, which requires a separate storage API call to permanently remove files.
  4. DDropping the managed table analytics.sales.transactions removes only the metadata entry and leaves the data files in place until an engineer manually runs VACUUM on the storage path. Dropping the external table finance.reporting.invoices immediately and permanently deletes both the metadata and all underlying data files in S3, because external tables transfer complete storage ownership to Unity Catalog upon initial table registration.
  5. EDropping either type of table in Unity Catalog automatically places the table in a 30-day recycle bin from which it can be fully restored, including all historical data versions. The underlying data files for both managed and external tables are preserved during this recycle period in a quarantine storage tier, and administrators must explicitly call PURGE TABLE with a FORCE flag to permanently remove them from storage.
Show answer & explanation

Correct answer: B

WHY B is correct: Unity Catalog distinguishes clearly between managed and external tables in terms of lifecycle management. For managed tables, Unity Catalog controls both the governance (metadata, access policies) AND the physical data storage. Therefore, when a managed table is dropped, Unity Catalog removes it from the metastore and deletes the underlying data files from the Unity Catalog-managed storage location after a recovery period (7 days by default), during which UNDROP TABLE can restore it. For external tables, Unity Catalog manages only the governance — access control and metadata — but does NOT own the data files, which reside in a user-specified cloud storage path. Dropping an external table removes only the metadata registration (the 'pointer') from the metastore; the data files at the specified external location remain unchanged. This is a critical distinction for data lifecycle management and disaster recovery planning. WHY NOT A: It is incorrect that Unity Catalog treats both table types identically for deletion. The managed/external distinction exists precisely to handle different lifecycle ownership models. External tables exist because organizations need to retain data file control outside of Databricks (for other tools, compliance, or cross-platform access), so Unity Catalog intentionally never deletes external data files on DROP. WHY NOT C: It is incorrect that dropping a managed table leaves the data files untouched. Managed tables are 'fully managed by Unity Catalog, which means Unity Catalog manages both the governance and the underlying data files.' Dropping a managed table is a complete deletion operation — both metadata and data files are removed. WHY NOT D: The description reverses the behavior. It is the managed table (not the external table) whose data files are deleted on DROP. External tables do NOT transfer ownership to Unity Catalog — that is the definition of an external table. Additionally, while VACUUM handles old Delta file cleanup, a DROP TABLE on a managed table removes the data files as part of the drop operation itself, not through a deferred VACUUM cycle. WHY NOT E: Unity Catalog does not provide a built-in 30-day recycle bin or quarantine storage tier for dropped tables as of the current platform version. While Delta Lake's time travel allows querying historical table versions before they are vacuumed, a DROP TABLE removes the table registration and (for managed tables) the data. There is no automatic recycle bin restore mechanism, and no PURGE TABLE command exists in this context.

3 Understanding of Databricks Data Intelligence Platform

A data analyst at a logistics company is exploring Databricks Marketplace to acquire a public geospatial reference dataset and a Databricks notebook demonstrating geospatial analysis patterns. The analyst has a Unity Catalog-enabled Databricks workspace. Which set of statements about Databricks Marketplace behavior and capabilities is ACCURATE?

  1. ATo browse Marketplace listings, the analyst first applies to become a registered Marketplace provider, as provider and consumer identities share one registration workflow. Read-only consumer browsing is unlocked once Databricks approves the application and assigns a provider profile to the workspace; after that, the geospatial dataset and the notebook can be requested from the workspace like any other listing.
  2. BMarketplace delivers data in a Databricks-proprietary format: a dataset acquired through a listing is converted from its original Parquet, CSV or JSON files into an encrypted binary archive before delivery. The analyst can query the converted dataset inside the workspace, but it is not readable from Apache Spark running outside Databricks or from other downstream analytics tools, and the conversion is not reversible.
  3. CThe Open Marketplace at marketplace.databricks.com can be browsed without logging in to a Databricks workspace, so the analyst can evaluate listings first. To request access and install the dataset or notebook, the analyst uses Marketplace inside the Unity Catalog-enabled workspace; once the provider approves, the dataset appears as a read-only Unity Catalog catalog that can be queried with SQL, and the notebook is installed to the workspace.
  4. DMarketplace listings, datasets and notebooks alike, are delivered as soon as a Databricks user clicks 'Get', with no provider approval step, no terms-of-use acceptance and no workspace permission requirement. Once acquired, the dataset appears as a writable catalog in the analyst's Unity Catalog, and local transformations can be written back through it to the provider's original data source.
  5. ENotebooks in Marketplace are static, read-only PDF exports of notebook code that are not importable into a workspace; to use one, the analyst copies the code from the PDF into a new blank notebook. Datasets, by contrast, are delivered as live queryable catalogs. The difference reflects that Databricks treats structured tabular data as the asset type with enough governance controls for live delivery.
Show answer & explanation

Correct answer: C

WHY C: This matches the documented behavior across two separate actions. The Open Marketplace (marketplace.databricks.com) is browsable without a Databricks account or workspace, which lets the analyst pre-evaluate listings. To request and install a data product, the consumer works from a Databricks workspace. For datasets shared through Delta Sharing, the approved product appears as a catalog in the consumer's Unity Catalog metastore, queryable with standard SQL like internal data but read-only (the consumer does not write back to the provider's source). A notebook listing is installed to the workspace and can be run or cloned straight away. Some listings are available instantly on terms acceptance; others need explicit provider approval first. WHY NOT A: Provider and consumer identities are separate in Marketplace. A consumer does not register as a provider to browse or request data products; the consumer workflow (browsing, requesting, querying) is independent of the provider workflow (listing, sharing, approving). Requiring provider registration would make the platform unusable as a broad data exchange. WHY NOT B: Marketplace uses Delta Sharing, an open standard that interoperates with Apache Spark outside Databricks, pandas, Power BI and other tools. Shared data is not converted to a proprietary format; open, format-agnostic interoperability is a design goal of Delta Sharing, and data stays in its original Delta, Parquet or other format. WHY NOT D: Not every listing is instant or unconditional: some require provider approval, and listings require the consumer to accept terms of use. Shared catalogs are read-only for consumers, so nothing is written back to the provider's data source; allowing that would violate the provider's data sovereignty, which the Delta Sharing protocol prevents. WHY NOT E: Marketplace notebooks are real installable, executable Databricks notebooks, not static PDF exports; an acquired notebook is installed to the workspace in runnable, editable form. Datasets and notebooks differ in how they are delivered (a Unity Catalog catalog versus a workspace notebook object), not in executability or governance.

4 Understanding of Databricks Data Intelligence Platform

A data engineering team is evaluating the Databricks Data Intelligence Platform for the first time. A team member lists several platform components and assigns each one a primary role. Which of the following descriptions CORRECTLY matches a platform component to its primary purpose?

  1. ALakeflow Spark Declarative Pipelines provides unified governance, access control and lineage tracking across the data and AI assets on the platform; it enforces fine-grained table permissions through ANSI SQL GRANT and REVOKE statements and supports cross-workspace data discovery through a centralized metadata catalog.
  2. BLakeflow Jobs is a declarative pipeline framework for defining and executing batch or streaming ETL transformations written in SQL or Python; its main contribution is the ability to define streaming tables and materialized views that refresh as new source data arrives, without hand-written trigger logic.
  3. CMosaic AI is the integrated AI and machine learning platform on Databricks for building, deploying and managing AI applications; it covers generative AI agents, Foundation Model APIs, Vector Search for RAG, AutoML for automated model training, Model Serving endpoints, and MLflow-based experiment tracking and model lifecycle management.
  4. DUnity Catalog is the performance-optimized query execution engine in the Databricks Runtime; it accelerates SQL and DataFrame workloads by replacing the standard Apache Spark execution engine with a vectorized native engine called Photon, which processes data at native CPU speed without code changes.
  5. EDatabricks SQL is a machine learning model registry and real-time deployment service; it stores trained model artifacts, manages model versions with approval workflows, and exposes LLMs and classic ML models as auto-scaling REST endpoints that serve inference traffic at low response latencies.
Show answer & explanation

Correct answer: C

WHY C: Mosaic AI is the integrated AI and machine learning pillar of the Databricks Data Intelligence Platform. It spans the AI lifecycle: building generative AI applications (Agent Bricks, Agent Framework, AI Playground), accessing foundation models through the Foundation Model APIs, storing and retrieving embeddings with Vector Search for RAG, training classic ML models with AutoML and MLflow tracking, deploying models through Model Serving endpoints, and governing models with Unity Catalog. WHY NOT A: Unified governance, access control, lineage, GRANT/REVOKE statements and a cross-workspace metadata catalog describe Unity Catalog. Lakeflow Spark Declarative Pipelines (formerly Delta Live Tables) is a framework for building batch and streaming pipelines with declarative SQL/Python, not a governance system. WHY NOT B: A declarative pipeline framework with streaming tables and materialized views describes Lakeflow Spark Declarative Pipelines, not Lakeflow Jobs. Lakeflow Jobs is the workflow orchestration service: it schedules and runs notebooks, scripts and pipeline triggers on a cron or event-driven basis. They are distinct products. WHY NOT D: A vectorized query engine called Photon describes the Photon component of the Data Intelligence Engine, not Unity Catalog. Unity Catalog's role is data governance (access control, auditing, lineage, metadata management), not query acceleration. WHY NOT E: Databricks SQL is the SQL analytics and BI layer: SQL warehouses, the SQL Editor, AI/BI Dashboards and Genie spaces for analysts to run queries and build dashboards. It is not a model registry or serving service; model serving and the MLflow model registry are functions of Mosaic AI.

5 Understanding of Databricks Data Intelligence Platform

A data governance team is conducting a quarterly audit of their Unity Catalog setup using the Catalog Explorer in Databricks. The team lead asks which of the following tasks can be accomplished DIRECTLY within the Catalog Explorer UI, without needing to write SQL, open a notebook, or use the CLI.

  1. ARunning long SQL transformation queries that repartition large Delta tables, previewing the DAG execution plan in a Spark UI embedded in Catalog Explorer, and stopping a stalled Lakeflow Jobs run from a pipeline operations panel in the Catalog Explorer sidebar, each of them performed directly within the Catalog Explorer interface.
  2. BProvisioning new interactive compute clusters and defining their auto-scaling policies, attaching init scripts and libraries to those clusters from the table detail panel, and browsing MLflow experiment run metrics and artifacts for registered models, each of them without leaving the Catalog Explorer tab or opening another section of the Databricks UI.
  3. CConfiguring Kafka ingestion source settings for a schema, assigning serverless compute sizes to individual SQL warehouses from the table detail panel, building row filter expressions in a drag-and-drop security policy editor, and enabling schema evolution for streaming tables, each of them from one configuration panel within the Catalog Explorer interface.
  4. DViewing column-level lineage that traces which upstream tables, views and notebooks fed a downstream Delta table; tagging a data asset with business metadata to improve discoverability; inspecting table ownership and granted permissions; and marking a table as certified to signal its quality to consumers, each of them directly within the Catalog Explorer interface.
  5. ERewriting the queries registered under a schema to use the latest Liquid Clustering keys, generating synthetic test data for empty tables to validate dashboard visualizations, and scheduling data quality alerts that send Slack notifications when freshness SLAs are missed, each of them offered as a native workflow inside the Catalog Explorer interface.
Show answer & explanation

Correct answer: D

WHY D: Catalog Explorer is Unity Catalog's primary visual management interface and supports these tasks. (1) Lineage: Unity Catalog captures column-level lineage, and Catalog Explorer shows it as a visual graph for tracing data upstream and downstream across tables, notebooks, and jobs. (2) Tagging: Catalog Explorer has a UI for adding and editing tags and comments on catalogs, schemas, tables, and columns to support discovery. (3) Permissions and ownership: administrators view and manage GRANT/REVOKE permissions from the table detail view. (4) Certified datasets: data stewards mark tables as certified (endorsements) in Catalog Explorer, which shows a quality indicator to consumers browsing the catalog. WHY NOT A: SQL transformation queries and Spark UI DAG views belong to the SQL Editor or notebooks; Catalog Explorer is a metadata browsing and governance tool, not a query execution environment. Starting and stopping Lakeflow Jobs runs is done in the Jobs (Workflows) UI. WHY NOT C: Kafka ingestion is configured in the Lakeflow Spark Declarative Pipelines pipeline configuration or cluster libraries, and SQL warehouse sizing in the SQL Warehouses section. Row filters and column masks are Unity Catalog features, but they are applied through SQL (ALTER TABLE) or the permissions panel, not through a drag-and-drop editor in Catalog Explorer. WHY NOT B: Cluster provisioning, auto-scaling policies, init scripts and libraries are managed in the Compute section, and MLflow experiments are browsed in the Experiments section under Machine Learning, not in Catalog Explorer. WHY NOT E: Query rewriting for Liquid Clustering, synthetic data generation and alert scheduling are not Catalog Explorer features. Liquid Clustering is configured with CLUSTER BY in DDL statements, and alerts are configured in the SQL Alerts section of Databricks SQL.

Take the full Data Analyst practice test →