10 free sample questions from a bank of 1251, with the correct answers and explanations. No signup required — start practising right now.
1Snowflake provides a mechanism for its customers to override its natural clustering algorithms. This method is:
Micro-partitions
Clustering keys
Key partitions
Clustered partitions
Answer: B
The short version
B — Clustering keys let customers override natural clustering. Snowflake auto-groups micro-partitions, but keys steer co-location for query pruning. The settled key is B because keys are the documented override.
Key concepts in this question
Natural clustering: automatic load-order grouping that Snowflake maintains without admin keys.
Clustering keys: chosen columns that direct related rows into the same micro-partitions.
Micro-partitions: small immutable storage units with min-max metadata used for pruning.
Key partitions and clustered partitions: nonstandard labels, not the Snowflake override feature.
Override intent: explicit customer control over layout for large filtered tables.
Why B is correct
B is correct because Snowflake docs define clustering keys as lists of columns over which micro-partitions are organized. Adding a key re-clusters the table around those expressions and improves pruning for common filters. Micro-partitions themselves are the storage result, not the control, so only B names the override mechanism.
Why the others are wrong
A. Micro-partitions are the storage outcome, not the customer control that reorders them.
C. Key partitions is not a Snowflake feature name for clustering control.
D. Clustered partitions is likewise not the documented override construct.
SnowPro-Core exam tip
When the stem says override natural clustering choose clustering keys on filter columns.
2Which statement best describes clustering?
Clustering represents the way data is grouped together and stored within Snowflake's micro-partitions
The database administrator must define the clustering methodology for each Snowflake table
The clustering key must be included on the COPY command when loading data into Snowflake
Clustering can be disabled within a Snowflake account
Answer: A
The short version
A — Clustering is how data groups inside micro-partitions. It describes physical co-location that drives pruning. The settled key is A because that grouping definition is canonical.
Key concepts in this question
Clustering: logical ordering of rows that places similar values together in storage.
Micro-partitions: Snowflake units that record min-max per column for skip decisions.
Automatic behavior: Snowflake clusters by load order unless keys refine it, so DBA setup is not mandatory.
COPY linkage: loading does not require listing keys; clustering applies after ingest.
Disable idea: clustering cannot simply be switched off as an account feature.
Why A is correct
A is correct because Snowflake defines clustering as the way data is grouped together and stored within micro-partitions. That grouping lets the optimizer skip partitions whose ranges miss the predicate. Options about mandatory DBA design, COPY key lists and account disable misstate the managed nature, leaving A as the best description.
Why the others are wrong
B. DBA need not define methodology per table because Snowflake auto-clusters by default.
C. COPY need not include the key; clustering is a table property, not a load parameter.
D. Clustering cannot be disabled account-wide; it is intrinsic to micro-partition storage.
SnowPro-Core exam tip
Define clustering as grouping inside micro-partitions that enables pruning; keys only refine it.
3What is the most granular object that the Time Travel retention period can be defined on?
Account
Database
Schema
Table
Answer: D
The short version
D — Table. Time Travel retention can be overridden at account, database, schema, and table level, and the table is the finest-grained object that carries it.
Key concepts in this question
DATA_RETENTION_TIME_IN_DAYS: the single parameter that sets Time Travel retention; it is accepted on CREATE/ALTER DATABASE, CREATE/ALTER SCHEMA, CREATE/ALTER TABLE and at account level.
Account default: the account-level value seeds every object; a lower-level statement wins only for the object it names.
Granularity ladder: account → database → schema → table. There is no column-level or row-level retention.
Setting 0: retention of 0 disables Time Travel for that object; -1 means "use the parent setting".
Why D is correct
Retention is a property of a securable object, and Snowflake exposes it at exactly four levels, with the table being the most granular. Because the parameter is defined on tables, a table can hold history longer or shorter than its schema without affecting siblings. Nothing below the table exists — a column cannot retain data.
Why the others are wrong
A. Account is the broadest level: it only sets the inherited default for every object.
B. Database is one step coarser than schema and table, so it is not the most granular option.
C. Schema overrides the database default but is still coarser than an individual table.
SnowPro-Core exam tip
Retention climbs down the hierarchy — account, database, schema, table — so the deepest object on the list is the answer to any "most granular" question.
4Which task is supported by the use of Access History in Snowflake?
Data backups
Cost monitoring
Compliance auditing
Performance optimization
Answer: C
The short version
C — Access History supports compliance auditing. It records who read which columns and when. The settled key is C because lineage plus access logs serve audit needs.
Key concepts in this question
Access History: Snowflake views tracking query access to tables, columns and policies.
Compliance auditing: proof of who touched sensitive data for regulation and review.
Data backups: point-in-time recovery handled by Time Travel and Fail-safe, not Access History.
Cost monitoring: warehouse usage views, separate from column-access lineage.
Performance tuning: Query History plus pruning analysis, not access lineage.
Why C is correct
C is correct because Access History joins queries to accessed objects and policies, letting auditors trace PII or financial reads. That lineage answers who saw what and when for SOX, HIPAA style reviews. Backups, cost and speed use different features, so only compliance auditing matches the supported task.
Why the others are wrong
A. Backups rely on clone, Time Travel and Fail-safe, not on access lineage.
B. Cost monitoring uses metering and warehouse history, not column-access records.
D. Performance work uses Query History and pruning stats, not Access History lineage.
SnowPro-Core exam tip
Link Access History to who-read-what auditing; link Query History to speed and cost.
5Which feature of Snowflake’s Continuous Data Protection (CDP) has associated costs?
Fail-safe
Network policies
End-to-end encryption
Multi-Factor Authentication (MFA)
Answer: A
The short version
A — Fail-safe carries storage costs inside Continuous Data Protection. It extends recovery beyond Time Travel with Snowflake-managed vaulting. The settled key is A because only Fail-safe adds billable bytes.
Key concepts in this question
Continuous Data Protection: built-in encryption, Time Travel and Fail-safe for durability.
Fail-safe: post-Time-Travel recovery window maintained by Snowflake support.
Storage cost: charge for bytes retained during Fail-safe, distinct from free encryption.
Network policies: access controls with no storage billing implication.
MFA and encryption: security controls included without extra storage fees.
Why A is correct
A is correct because Snowflake bills Fail-safe storage while end-to-end encryption, network policies and MFA carry no comparable data-retention charge. Fail-safe preserves historical partitions after Time Travel expires, consuming capacity. That cost link is explicit in packaging docs, so A is the only CDP feature with associated costs here.
Why the others are wrong
B. Network policies gate logons and add no storage-cost footprint.
C. End-to-end encryption protects data without separate retention billing.
D. MFA strengthens auth and has no storage-cost relation.
SnowPro-Core exam tip
For CDP costs remember Fail-safe storage is billable; encryption, MFA and policies are not.
6When enabling access to unstructured data, which URL permits temporary access to a staged file without the need to grant privileges to the stage or to issue access tokens?
File URL
Scoped URL
Relative URL
Pre-Signed URL
Answer: B
The short version
B — Scoped URL gives temporary staged-file access without stage grants or tokens. It encodes short-lived permission for one file. The settled key is B because scoped links avoid privilege and token steps.
Key concepts in this question
Unstructured access: reading staged files through URLs for apps and sharing.
Scoped URL: short-lived link valid about a day that inherits requester rights without extra grants.
File URL: persistent path that still requires stage privileges to resolve.
Relative URL: shorthand resolved against a stage context, not a standalone grant.
Pre-signed URL: longer-lived link that needs access tokens and explicit issuance.
Why B is correct
B is correct because Snowflake documents scoped URLs as temporary access that needs neither stage privilege grants nor access tokens. The URL itself carries limited entitlement for the file and expires quickly. Pre-signed links need tokens, while file and relative forms still depend on stage rights, so only B meets the no-grant plus no-token condition.
Why the others are wrong
A. File URL still checks stage privileges, so it fails the no-grant need.
C. Relative URL resolves within stage context and does not bypass privilege checks.
D. Pre-signed URL requires tokens by design, violating the token-free requirement.
SnowPro-Core exam tip
Need token-free temp file access without stage grants, choose scoped URL expiring in a day.
7Which Snowflake function is maintained separately from the data and helps to support features such as Time Travel, Secure Data Sharing, and pruning?
Column compression
Data clustering
Micro-partitioning
Metadata management
Answer: D
The short version
D — Metadata management sits apart from data and powers Time Travel plus sharing plus pruning. Column stores hold values while metadata tracks versions and ranges. The settled key is D because only metadata is separate yet enabling.
Key concepts in this question
Metadata layer: Snowflake service tracking files, versions, min-max and sharing pointers.
Time Travel: historical reads enabled by retained versions plus metadata pointers.
Secure Data Sharing: pointer-based sharing without copying, driven by metadata.
Pruning: skipping micro-partitions using stored min-max, a metadata use.
Compression and clustering: data-layout tricks that live with the bytes, not apart from them.
Why D is correct
D is correct because Snowflake maintains metadata separately from column data to support cloning, travel, sharing and pruning at cloud scale. Compression, clustering and micro-partitioning all describe byte organization, not the separate control plane. Only metadata management explains a distinct service that still enables those three headline features.
Why the others are wrong
A. Column compression shrinks bytes but remains part of data storage, not separate management.
B. Data clustering orders rows inside storage rather than providing a separate service.
C. Micro-partitioning is the data layout itself, not the distinct metadata function.
SnowPro-Core exam tip
For separate-from-data plus Time Travel plus sharing plus pruning choose metadata management.
8A tag object has been assigned to a table (TABLE_A) in a schema within a Snowflake database.
Which CREATE object statement will automatically assign the TABLE_A tag to a target object?
CREATE TABLE [table_name] LIKE TABLE_A;
CREATE VIEW [view_name] AS SELECT * FROM TABLE_A;
CREATE TABLE [table_name] AS SELECT * FROM TABLE_A;
CREATE MATERIALIZED VIEW [view_name] AS SELECT * FROM TABLE_A;
Answer: A
The short version
A — CREATE TABLE LIKE copies the TABLE_A tag to the new table. LIKE clones structure plus tag lineage automatically. The settled key is A because only that form propagates tags.
Key concepts in this question
Tags: governance labels for classification and policy attached to tables and views.
CREATE LIKE: new table from source definition that inherits tags by design.
CTAS: new table from query results that does not auto-carry source tags here.
Views: separate objects whose tag inheritance differs from table LIKE behavior.
Materialized views: precomputed results with their own tagging rules, not LIKE propagation.
Why A is correct
A is correct because Snowflake specifies that CREATE TABLE new LIKE TABLE_A automatically assigns the source tags to the target. The other forms create fresh objects from SELECT results or view definitions without that automatic tag copy. That distinction makes LIKE the only statement that moves the TABLE_A tag by itself.
Why the others are wrong
B. Plain view over TABLE_A does not auto-assign the base-table tag to the view here.
C. CTAS builds from rows and starts without automatic source-tag assignment.
D. Materialized view over TABLE_A follows its own rules and does not inherit via LIKE logic.
SnowPro-Core exam tip
For automatic tag copy from a table remember CREATE TABLE LIKE; CTAS and views do not.
9In addition to performing all the standard steps to share data, which privilege must be granted on each database referenced by a secure view in order to be shared?
READ
REFERENCES
REFERENCE_USAGE
USAGE
Answer: C
The short version
C — REFERENCE_USAGE is required on each database behind a shared secure view. Secure views hide logic, so share consumers need that cross-database right. The settled key is C because it unlocks referenced data without exposing internals.
Key concepts in this question
Secure views: access-controlled definitions that hide SQL while sharing results.
Data sharing: provider-to-consumer flow requiring rights on every underlying database.
REFERENCE_USAGE: privilege that lets a share use objects from a database without full access.
USAGE and READ: coarser or incorrect grants that do not satisfy the secure-share rule.
REFERENCES: object-level right distinct from the share-specific usage grant.
Why C is correct
C is correct because Snowflake requires REFERENCE_USAGE on each database referenced by the secure view in addition to standard share steps. That grant lets the consumer resolve cross-database objects at query time while the view still masks logic. Other privileges either over-grant or miss the sharing semantic, so only C completes the share.
Why the others are wrong
A. READ is not the documented cross-database share grant for secure views.
B. REFERENCES governs object references generally, not the share-specific database right.
D. USAGE alone allows basic access but does not satisfy the secure-view sharing rule.
SnowPro-Core exam tip
For secure-view shares add REFERENCE_USAGE on every underlying database.
10Which function can be used with the COPY INTO [LOCATION] statement to convert rows from a relational table to a single VARIANT column, and to unload rows into a JSON file?
FLATTEN
OBJECT_AS
OBJECT_CONSTRUCT
TO_VARIANT
Answer: C
The short version
C — OBJECT_CONSTRUCT builds a VARIANT row for COPY INTO JSON unload. Relational columns become one semi-structured object per row. The settled key is C because that function targets JSON unload.
Key concepts in this question
COPY INTO LOCATION: unload path that writes table rows to staged files such as JSON.
OBJECT_CONSTRUCT: function that maps keys to column values into a single VARIANT.
VARIANT column: flexible type that holds the constructed object for JSON output.
FLATTEN: explodes semi-structured data, opposite of building an object.
TO_VARIANT and helpers: type casts that do not assemble multi-column objects alone.
Why C is correct
C is correct because OBJECT_CONSTRUCT takes column expressions and returns one VARIANT object ready for JSON unload. COPY INTO LOCATION can then write those objects as JSON lines. FLATTEN unnests rather than builds, while TO_VARIANT merely casts single values, so only C performs the row-to-object conversion the stem demands.
Why the others are wrong
A. FLATTEN expands VARIANT into rows, reverse of the required construct step.
B. OBJECT_AS is not the documented row-to-VARIANT builder for this unload.
D. TO_VARIANT casts one value and cannot combine many columns into an object.
SnowPro-Core exam tip
To unload rows as JSON VARIANT think OBJECT_CONSTRUCT; to explode think FLATTEN.