Free Databricks Certified Data Analyst Associate Sample Questions

Correct answer: D — MCP Servers

• A. Datasets listings expose tabular or volume-based data for querying, not a standardized interface for an agent to call external tools.
• B. AI Models listings expose trained models for inference, not the tool-calling interface an agent uses to reach external systems.
• C. Notebooks listings expose reference code for a use case, not a running interface an agent can call at runtime.
• D. MCP Servers listings expose Model Context Protocol servers, which provide the tools and APIs an agent calls to interact with external systems, matching the team's need.

Correct answer: A — Run `CREATE TABLE catalog.schema.table (...)` without a `LOCATION` clause, letting Unity Catalog place the files under the schema's own managed storage path.

• A. Correct. Omitting `LOCATION` produces a managed table, and Unity Catalog automatically stores its data under the managed storage path configured for the containing schema, catalog, or metastore.
• B. Incorrect. Supplying an explicit `LOCATION` at creation makes the table external, with Unity Catalog governing only metadata over a team-controlled bucket rather than its own managed path.
• C. Incorrect. There is no `ALTER TABLE ... SET MANAGED` conversion path; a table's managed-versus-external type is fixed at creation and is not changed by altering it afterward.
• D. Incorrect. `MSCK REPAIR TABLE` is a Hive-style partition-discovery command for existing metastore tables, not a mechanism for adopting arbitrary uploaded files as a Unity Catalog managed table.

Correct answer: A — Lakeflow Jobs, which orchestrates dependent tasks across ingestion, pipelines, notebooks, and ML steps on a schedule, with retries and failure alerts built in.

• A. Coordinating multiple heterogeneous steps in a fixed order, with retries and alerting, is precisely the orchestration role that this component fills across ingestion, pipelines, notebooks, and model-training tasks.
• B. Governance metadata and lineage tracking describe a cataloging function, not a scheduler, so this component cannot sequence the ingestion, pipeline, and notebook steps the scenario describes.
• C. A SQL analytics surface for querying finished results is a downstream consumer of the workflow's output, not the mechanism that schedules or sequences the upstream tasks that produce it.
• D. A storage format's transaction log records what was written and when, but it has no scheduler or dependency graph, so it cannot coordinate the ordering of separate ingestion, pipeline, and notebook tasks.

Correct answer: A — Enter `certificationStatus:certified` in the Catalog Explorer search box, which filters results to assets tagged with the certified system tag.

• A. The `certificationStatus:certified` search keyword filters Catalog Explorer results to assets carrying the system certification tag with a certified value. This is the built-in mechanism for discovering only trustworthy, vetted datasets without opening each asset individually.
• B. There is no `tag:certified` search syntax in Catalog Explorer, and certification is not implemented as an arbitrary user-defined tag value. Using this keyword would not reliably match certified assets.
• C. Creation date has no relationship to certification status, since tables can be certified at any point after creation, including years later. Sorting by creation date would not surface certified assets reliably.
• D. Certification status is not stored as a permissions grant, so there is no `CERTIFIED` entry to find on a Permissions tab. Access control and certification are separate, independently managed concepts in Unity Catalog.

Correct answer: A — Column-level lineage for the new table's columns is missing, because Unity Catalog cannot capture column lineage when a source is referenced by path rather than by table name.

• A. This is correct because column-level lineage cannot be captured when a source is referenced by file path rather than by table name, which is a documented limitation distinct from table-level tracking.
• B. This is incorrect because Unity Catalog does not silently resolve arbitrary storage paths back to registered tables for lineage purposes; the path reference itself breaks column tracing.
• C. This is incorrect because the limitation applies specifically to column-level detail; table-level lineage can still be recorded even when a source is referenced by storage path.
• D. This is incorrect because the path-based limitation is not restricted by data type; no column, numeric or otherwise, gets column-level lineage when the source is a storage path.

Correct answer: A — Every row in the table is deleted, since an omitted predicate is treated as matching all rows

• A. Databricks documentation states that when no predicate is provided, DELETE FROM deletes all rows in the table, so omitting WHERE clears the entire table rather than doing nothing or requiring extra syntax.
• B. Delta Lake does not enforce a mandatory WHERE predicate on DELETE statements; the syntax explicitly allows an optional predicate, so no validation error is raised.
• C. DELETE FROM has no special default behavior that targets only NULL-containing rows; without a predicate it applies to the entire table regardless of which columns hold NULLs.
• D. The statement executes immediately rather than staying inert, and its effect is to remove all rows, not to silently do nothing until a condition is supplied.

Correct answer: A — The notebook's `File` menu offers `Upload files to volume`, which opens the same upload dialog used elsewhere and writes the selected file into the chosen volume path.

• A. The notebook's `File` menu includes an `Upload files to volume` action that opens the standard upload dialog, letting the analyst stage the local CSV into a Unity Catalog volume without switching to Catalog Explorer.
• B. There is no `Run` menu command that mounts a laptop's local file system to DBFS; DBFS mounts are configured against cloud object storage, not a local machine, and files uploaded through the UI land in Unity Catalog volumes rather than DBFS.
• C. Databricks notebooks do not support a `%upload` magic command, and uploaded files are staged into Unity Catalog volumes rather than pushed into a cluster's ephemeral driver storage.
• D. The notebook results pane has no file-path field that fetches files from a local machine over SSH; Databricks has no mechanism to reach into an analyst's laptop this way.

Correct answer: A — Increase the maximum cluster count so the warehouse can spin up additional clusters and distribute the concurrent queries across them

• A. Raising the maximum cluster count is correct because SQL warehouses handle high concurrency by adding clusters that share the incoming query load, rather than by making one cluster larger; more clusters means more queries can run in parallel instead of queuing.
• B. Cluster size (t-shirt size) mainly affects how much compute a single query gets to run faster, not how many separate queries can run at once, so growing it further does little to relieve queuing from many simultaneous users.
• C. Classic warehouses are not specifically optimized for concurrency; they offer entry-level performance without the intelligent workload management that helps distribute concurrent load, so switching to Classic would not resolve this queuing.
• D. Photon is a vectorized query engine that speeds up query execution rather than a resource competing with concurrency, so disabling it would not free capacity for more simultaneous queries and could slow individual queries down.

Correct answer: A — A materialized view persists and caches its query results for reuse, while a plain view is evaluated on demand and recomputed on every query.

• A. A materialized view caches its computed output so downstream reads reuse precomputed data, whereas a plain view has no persisted state and recomputes its query logic every time it is read.
• B. This reverses the actual behavior: a materialized view is the one that caches and persists results, while a plain view is the one recomputed on demand for each query.
• C. A plain view does not persist any results to storage at all, so claiming both object types persist their output misstates how plain views work.
• D. Standalone materialized views can be queried broadly through Unity Catalog like other tables, and pipeline scoping applies to plain views defined inside a Lakeflow pipeline, not to materialized views generally.

Correct answer: A — Genie Code is governed by the requesting user's Unity Catalog permissions, so it can only access data and operations that user already has grants for.

• A. Genie Code operates within the requesting user's own Unity Catalog permissions, so it can only access and surface data and operations that user is already entitled to, which directly addresses the admin's governance concern.
• B. Genie Code does not bypass the requesting analyst's grants through a privileged shared identity; its access is scoped to what that specific user can already see.
• C. Genie Code does not operate from a static cached snapshot of the metastore that ignores live permission checks; access is scoped per request to what the user can see.
• D. There is no manual per-table allowlisting step required from the admin; access is governed automatically by the existing Unity Catalog grants for the requesting user.

Correct answer: A — `ON shipments.order_id = pricing.order_id AND shipments.warehouse_region = pricing.warehouse_region`, matching on both columns together

• A. Combining both columns with AND in the ON clause requires each joined row to match on order_id and warehouse_region simultaneously, which is exactly what is needed to avoid pairing an order with the wrong region's cost.
• B. Filtering with a separate WHERE clause after an order_id-only join can work logically similarly to an AND condition for an inner join, but it still first materializes the incorrect cross-region matches and is not what the ON clause itself was asked to express; the join predicate should carry the multi-key logic directly.
• C. USING(order_id) only deduplicates and matches the single named column order_id; it does not automatically pull in warehouse_region as part of the match condition, so cross-region mismatches would still occur.
• D. An OR condition matches rows where either column is equal, which is looser than required and would still produce incorrect matches whenever order_id matches but the region differs.

Correct answer: A — Rows sort ascending by category first, then within each category tie sort descending by price, since ORDER BY accepts SELECT-list positions per key

• A. Correct: ORDER BY supports referencing SELECT-list columns by their integer position, so 2 refers to category and 3 refers to price; each key sorts independently unless a shared direction is stated, so category sorts ascending by default while price sorts descending as explicitly marked.
• B. Incorrect: Databricks SQL explicitly supports ordering by column position as an alternative to naming the column or alias directly, so this query parses and runs successfully.
• C. Incorrect: a DESC keyword only applies to the sort key it immediately follows; it does not retroactively apply to earlier keys in the same ORDER BY list, so category still sorts ascending here.
• D. Incorrect: positional references are fully honored by ORDER BY rather than ignored, and product_id is never used as a sort key in this query since it was not referenced by position or name.

Correct answer: A — `DESCRIBE HISTORY pricing;` to list each commit's version, timestamp, operation type, and the user or job that performed it

• A. Correct: `DESCRIBE HISTORY` returns the audit trail of a Delta table's operations, including version, timestamp, operation type, and the user or job identity, making it the direct way to identify which commits to inspect further.
• B. Incorrect: `SHOW TBLPROPERTIES` returns configuration settings such as retention durations, not a log of individual write events, so it cannot reveal who changed what or when.
• C. Incorrect: reading only the earliest version and manually diffing every row against the current table is a slow, error-prone substitute for the operation log and would not surface the acting user or job for each change.
• D. Incorrect: `EXPLAIN` shows the physical or logical execution plan for the current query, not a historical record of past write operations performed on the table.

Correct answer: D — `finance_catalog.sales_schema.orders_delta`, because Unity Catalog resolves objects through a three-level catalog, schema, and table hierarchy rather than a single database name.

• A. Omitting the catalog only works when a default catalog has already been set for the session, so it is not the reliable, fully qualified reference the scenario calls for when joining across business units.
• B. Every table in Unity Catalog belongs to a schema, so there is no mode where the schema and table collapse into a single flattened identifier under the catalog.
• C. Unity Catalog identifiers use dot separators between catalog, schema, and table, not an at-sign, so this syntax does not match how object names are resolved.
• D. Unity Catalog always resolves objects through the three-level `catalog.schema.table` hierarchy, so naming all three levels explicitly is the fully qualified reference the join requires.

Correct answer: A — `CREATE FOREIGN CATALOG pg_catalog USING CONNECTION pg_inventory OPTIONS (database 'retail_db')`, which mirrors the remote database's schemas and tables.

• A. `CREATE FOREIGN CATALOG` is the correct statement for federation; it takes an existing connection and a remote database name, then mirrors that database's schemas and tables as a foreign catalog in Unity Catalog.
• B. Plain `CREATE CATALOG` provisions a managed catalog backed by Databricks-managed storage, not a mirror of a remote federated database, so it is the wrong statement for this scenario.
• C. There is no `CREATE FOREIGN SCHEMA ... FROM CONNECTION` statement in Lakehouse Federation; foreign catalogs are the unit created on top of a connection, and schemas underneath sync automatically.
• D. `ALTER CONNECTION` modifies properties of an existing connection object; it cannot be used to create or attach a new foreign catalog, which requires its own `CREATE FOREIGN CATALOG` statement.

Correct answer: A — Photon accelerates the Parquet file writes underlying Delta Lake table writes, reducing the time spent encoding and persisting the ingested batch data to storage.

• A. This is correct: Photon accelerates the underlying Parquet write path used by Delta Lake (and Iceberg), so encoding and persisting the batch output to storage completes faster.
• B. This is incorrect because Photon still writes standard Parquet files under Delta Lake's open format; it does not introduce a proprietary binary layout, and tables remain readable by any Delta-compatible engine.
• C. This is incorrect because Photon does not alter Delta Lake's transactional guarantees; every write still commits to the transaction log as normal, and log updates are not deferred to a later OPTIMIZE run.
• D. This is incorrect because Photon does not automatically apply Z-ordering on writes; Z-ordering is a separate, explicitly configured optimization technique, not a side effect of Photon's write acceleration.

Correct answer: A — Append `LIMIT 1` to the statement so only the single most recent commit, which is listed first, is returned

• A. Correct: `DESCRIBE HISTORY` returns commits in reverse chronological order, so appending `LIMIT 1` isolates just the most recent commit, which is the fastest way to check what last night's run did.
• B. Incorrect: filtering on `operation = 'DELETE'` with no limit returns every delete across the table's entire lifetime, which is far more output than needed for a quick check of just the latest run.
• C. Incorrect: sorting ascending by version and limiting to one row returns the oldest commit in the table's history, the opposite of the most recent commit the analyst needs to inspect.
• D. Incorrect: grouping by operation type collapses rows across the entire history into aggregated buckets, losing the per-commit timestamp and version detail needed to confirm what happened in last night's specific run.

Correct answer: A — The existing partitioning must be removed as part of the migration, since liquid clustering cannot be combined with partitioning or Z-order on the same table

• A. Liquid clustering is documented as incompatible with both Hive-style partitioning and Z-ordering, so migrating means the team drops the partitioning scheme and stops the Z-order jobs in favor of a `CLUSTER BY` declaration on the columns that matter for their queries.
• B. Delta Lake does not support running partitioning and liquid clustering on the same table at once, so the two layout strategies cannot be layered together rather than one replacing the other.
• C. Once a table is declared with `CLUSTER BY`, Z-order compaction is no longer the applicable maintenance operation, so continuing to run `ZORDER BY` jobs would conflict with rather than complement the new clustering configuration.
• D. Migrating to liquid clustering does not require deleting historical files; existing data can remain in place and be reorganized later through an `OPTIMIZE FULL` operation rather than starting from an empty table.

Correct answer: A — The remote result cache is stored outside the cluster and persists across a warehouse restart, so the identical query reuses the earlier computed result.

• A. The remote result cache is a workspace-level store separate from any single cluster, has a lifecycle of up to 24 hours, and survives a warehouse being stopped and restarted, so an unchanged identical query returns the cached result immediately.
• B. The disk cache lives on local SSD attached to compute nodes and is cleared when a cluster or warehouse restarts, so it could not have survived the overnight stop to serve this result.
• C. Running a query in the SQL editor does not create a materialized view of the result; no persistent view object is created just by executing a SELECT statement.
• D. A stopped serverless warehouse is not still running underneath the billing layer; the near-instant response comes from the persistent remote result cache, not from hidden active compute.

Correct answer: A — `UNION ALL` keeps every row from both inputs without checking for duplicates, so the matching row count is expected; switching to plain `UNION` removes exact duplicate rows across the combined result.

• A. This is correct: `UNION ALL` concatenates all rows from both queries with no deduplication step at all, which is exactly why the combined row count equals the sum of the two extracts; switching to `UNION` performs a distinct-style removal of exact duplicate rows.
• B. This reverses the actual behavior: `UNION ALL` is specifically the variant that skips deduplication, while plain `UNION` is the one that removes duplicate rows, so no bug or workaround is involved here.
• C. `UNION ALL` performs no internal deduplication logic based on keys or any other columns, so describing the matching row count as coincidental misrepresents how the operator is defined to behave.
• D. This describes the behavior of plain `UNION`, not `UNION ALL`; the whole reason the combined count matches the sum of the two extracts is that no such column-wise duplicate comparison occurs during the union.

Correct answer: A — A pie chart, since it is designed to show proportionality between categories and is not intended for time-series data.

• A. A pie chart shows proportionality between metrics as slices of a whole and is documented as not intended for time-series data, matching this single-snapshot category breakdown.
• B. A line chart is built to present how metrics change across time periods, which does not match a single-point snapshot of category proportions for one month.
• C. An area chart also emphasizes change over time by combining line and bar elements, so it does not suit a one-time proportional breakdown of categories.
• D. A cohort chart tracks retention or behavior for grouped users across time periods, which is unrelated to a single snapshot of sales share by product category.

Correct answer: A — Heatmap and table visualizations

• A. Heatmap and table visualizations are the two types documented with the 64K row or 10MB underlying-data limit, since both render every individual data point or cell directly.
• B. Counter and pivot visualizations are not the types called out with this specific row and size limit in the documentation.
• C. Bar and line visualizations are not the types documented with this particular row and size limit.
• D. Scatter and bubble visualizations are not the types documented with this particular row and size limit.

Correct answer: A — The recipient is prompted to authenticate through the identity provider or an email one-time passcode, then sees a dashboard-only, view-focused experience.

• A. Recipients without workspace access still need to authenticate, either through the account's identity provider or a one-time passcode sent by email, before landing in a simplified, dashboard-only view.
• B. Shareable links are not anonymous; a login step through the identity provider or a one-time passcode is required before the recipient can see the dashboard's contents.
• C. Shareable links are specifically designed to extend access to registered account members outside the workspace, so lacking workspace membership does not by itself result in denied access.
• D. Opening a shareable link authenticates the recipient as a registered account member for a view-only copy of the dashboard; it does not create or grant any workspace membership.

Correct answer: C — Set the alert's Data Value option to evaluate the first value of the `error_count` column, so only the top row of the sorted result set is compared against the threshold.

• A. Summing across all rows combines every hour's error count into one total, which still mixes in older rows instead of isolating just the most recent hour's value.
• B. Averaging across all rows blends the newest count together with every older row's count, so it does not isolate the most recent hour on its own.
• C. The first-value setting evaluates only the top row of the query result, which is the most recent hour given the sort order, matching exactly what the team wants.
• D. The notification template only formats the text of an already-fired alert message; a `LIMIT 1` placed there has no effect on which row the condition itself compares.

Correct answer: A — Keyword

• A. The keyword is the literal identifier that appears after the colon in the SQL query, so it can only be renamed by editing that query text; the parameter editor's gear-icon panel does not offer a separate control to rename it.
• B. The display name is a label shown in filter editors and widgets and can be freely edited from the parameter configuration panel without touching the underlying query text.
• C. The parameter's type, such as String, Date, or Numeric, is a dropdown choice set in the configuration panel and does not require editing the SQL query.
• D. The multi-select checkbox toggling array-based filtering is set directly in the configuration panel and does not require any change to the query text.

Correct answer: A — Pie visualizations show proportionality between metrics but are not meant for conveying time series data, since slices do not communicate change over time.

• A. Pie visualizations are documented as showing proportionality between metrics but not being meant for conveying time series data, because static slices cannot represent how a value changes across a continuous time axis.
• B. Pie visualizations are not restricted to two slices; they can represent several categories at once, though usability degrades as the number of slices grows, so this stated hard limit is incorrect.
• C. Pie visualizations encode proportion through slice size based on category values, not through a numeric x-axis and y-axis pair, so there is no such two-continuous-variable requirement.
• D. Pie visualizations, like other dashboard widgets, can be built from datasets defined in the SQL editor or notebooks, not only from Genie space datasets, so this restriction does not exist.

Correct answer: C — A trusted asset defined as an example SQL query for the churn calculation, so Genie reuses vetted logic instead of generating a new formula each time.

• A. Dashboard parameters belong to AI/BI dashboards, not Genie spaces, and they control filter values in a fixed visualization rather than standardizing a conversational calculation.
• B. A governed tag classifies and can restrict access to a column, but it has no effect on which formula Genie chooses when generating SQL for a churn question.
• C. This is correct because a trusted asset, a vetted example SQL query, gives Genie one approved calculation to reference instead of re-deriving the formula for each question.
• D. A scheduled alert notifies users when a value crosses a threshold using whatever formula last computed it; it does not standardize or correct the SQL logic Genie uses to compute that value.

Correct answer: A — In Settings > Security > External access, set the embed dashboards policy to Allow approved domains and add the support console's domain to the approved list.

• A. Correct. Workspace admins control embedding policy under Settings > Security > External access, and an author cannot successfully embed a space in an external domain until that domain is allowed or explicitly approved there.
• B. CAN MANAGE on the space controls who can administer or share that specific Genie space, but it does not control which external domains are permitted to host an embedded iframe.
• C. Creating a new catalog and migrating tables changes data organization, not embedding permissions, and is unrelated to whether a domain is allowed to embed the space.
• D. Photon is a query execution engine setting that affects performance, not whether an external domain is authorized to embed a Genie space.

Correct answer: A — Use the request-review option on that response so an administrator can inspect the prompt, the SQL, and an optional comment.

• A. Requesting review on that exact response routes the specific prompt, SQL, and any comment straight to a space administrator for verification, which matches wanting an in-tool check without leaving the space.
• B. A general IT ticket routes to a help desk that has no direct visibility into that Genie space's SQL or curated instructions, adding an unnecessary detour.
• C. A thumbs down records general sentiment but does not guarantee an administrator inspects this particular response, and there is no defined digest workflow tied to it.
• D. Editing SQL directly requires the user to already know the correct query, which defeats the purpose of asking Genie and does not flag anything for an administrator to fix.

Correct answer: A — CAN USE permission on the SQL warehouse, which is the minimum access level required before it can be attached to a Genie space

• A. This is correct because CAN USE is documented as the minimum permission level a user needs on a SQL warehouse before that warehouse becomes available to attach to a Genie space.
• B. This is incorrect because Genie spaces do not require administrative control of the warehouse; the far lower CAN USE level is sufficient for attaching it.
• C. This is incorrect because attaching a warehouse to a Genie space only requires the ability to submit queries against it, not a separate restart-level permission.
• D. This is incorrect because read-only visibility does not let the warehouse actually execute queries submitted by the space; CAN USE, not CAN VIEW, is the required level.

Correct answer: A — A link table connecting the product and promotion hubs, capturing the many-to-many association between the two business concepts.

• A. Link tables exist precisely to represent relationships, including many-to-many associations, between two or more hub entities, which matches the product-to-promotion relationship described here.
• B. Satellites hold descriptive attribute history for a hub or link, not references to other hubs; representing a many-to-many relationship through a satellite would not match its intended role.
• C. Hub tables are meant to hold only a business key and minimal metadata, and a single foreign key column could not represent a many-to-many relationship where either side can have multiple matches.
• D. Fact tables belong to star schema Gold-layer modeling, and many-to-many relationships are commonly represented there through bridge tables; a data vault model uses link tables instead, not fact tables, for this purpose.

Correct answer: A — This is the Bronze layer, because its purpose is capturing source data as-is with minimal transformation, before any conforming or dimensional modeling happens.

• A. Keeping original field names with minimal transformation and only an audit timestamp is exactly the Bronze layer's role: capture data as-is for auditability and historical replay before any cleansing or dimensional modeling is applied downstream.
• B. Adding a single timestamp column is not the same as the deduplication and enterprise-wide conformance work that defines the Silver layer; the data described here is still raw and unconformed, which keeps it at Bronze.
• C. A raw, minimally transformed table with original field names is not business-ready and has no aggregate or dimensional structure, so it does not meet the definition of the Gold layer regardless of having a timestamp column.
• D. A hub table's business key identifies a specific business entity for satellites to attach history to; an ingestion timestamp added for auditing raw file loads is not a business key and does not make this a data vault hub.

Correct answer: A — `SELECT table_schema, table_name FROM finance.information_schema.tables;`, since every catalog exposes its own `information_schema` schema listing the objects it contains.

• A. Each catalog in Unity Catalog has its own `information_schema` schema that surfaces metadata for the objects inside that specific catalog, so qualifying it as `finance.information_schema.tables` correctly scopes the metadata query to the `finance` catalog.
• B. This is incorrect because `information_schema` is not a single global schema shared across catalogs; without qualifying it with a catalog name, the reference resolves to whichever catalog is currently set as the session default, not necessarily `finance`.
• C. This is incorrect because `SHOW TABLES` requires a schema to be specified (or a current schema to be set); it does not accept a bare catalog name and list every table across all of that catalog's schemas in one call.
• D. This is incorrect because `DESCRIBE CATALOG` returns catalog-level metadata such as its owner and comment, not a full listing of every schema, table, and volume it contains.

Correct answer: C — Yes, because a privilege granted on a container object like a catalog automatically applies to all current and future child schemas and tables created inside it.

• A. This describes the opposite of how Unity Catalog inheritance actually behaves; a grant on a catalog is not a one-time snapshot of existing objects and does apply to objects created afterward, so the new table would not need a separate grant.
• B. There is no scheduled or periodic refresh required for privilege inheritance to take effect; Unity Catalog evaluates inherited privileges at query time, so re-running the grant statement is unnecessary.
• C. Unity Catalog privilege inheritance flows down from container objects to both current and future child objects, so a `SELECT` grant on the catalog automatically covers the newly created schema and table without any further action from the owner.
• D. Inheritance in Unity Catalog does not stop at the schema level; a privilege granted on a catalog continues cascading down through schemas to the tables they contain, so this describes an inheritance limit that does not exist.

Correct answer: B — The user who runs the `CREATE TABLE` statement owns the new object by default, and workspace administrators automatically own the workspace catalog created when Unity Catalog is auto-enabled.

• A. Ownership is not assigned to a group by default when an object is created, and the workspace catalog's owner comes from workspace administrator status, not from inheriting a metastore root owner.
• B. The creator of an object is its default owner in Unity Catalog, and when Unity Catalog is automatically enabled for a workspace, workspace administrators are set as owners of the resulting workspace catalog.
• C. Metastore admins do not automatically own every new object; ownership defaults to the creating user, and requiring a separate grant before that user can modify their own object is not how Unity Catalog behaves.
• D. New objects are not owned by the parent schema's owner by default, and the workspace catalog is not left ownerless; every securable object, including the workspace catalog, must have an owner from creation.