Snowflake SnowPro Core COF-C03 Practice Questions with Explanations

Free Snowflake SnowPro Core COF-C03 practice questions. 29 of them, each with the correct answer, a full explanation, and the reason every other option is wrong.

COF-C03 practice questions

Question 1

What is a key difference between the Snowflake CLI and SnowSQL?

  • A. SnowSQL is used for managing workloads and applications, while Snowflake CLI executes queries.
  • B. Snowflake CLI is primarily used for managing workloads and applications, while SnowSQL is used for executing operations.
  • C. SnowSQL supports only Data Definition Language (DDL) operations, while Snowflake CLI supports only Data Manipulation Language (DML) operations.
  • D. Snowflake CLI only supports query execution, while SnowSQL manages Python packages.
Show answer and explanation ▾

Correct answer: B

Snowflake CLI is the newer open-source command-line tool built for developer workloads: it creates, deploys and manages applications and objects such as Streamlit apps, Snowpark functions and procedures, Native App packages, notebooks and Snowpark Container Services, and it can also run SQL. SnowSQL is the older command-line client focused on connecting to Snowflake and executing SQL statements and operations, including loading and unloading data with PUT and GET. The deciding difference is therefore purpose: workload and application management for Snowflake CLI versus statement execution for SnowSQL.

Why the other options are wrong:

  • A. This reverses the roles of the two tools; application and workload management is the focus of Snowflake CLI, not SnowSQL.
  • C. Neither tool is limited to one class of SQL; both can run DDL and DML statements.
  • D. Snowflake CLI is not limited to queries, and SnowSQL is a SQL client rather than a Python package manager.

Question 2

Which services are handled within the Cloud Services layer? (Choose two.)

  • A. Authentication
  • B. Query processing
  • C. Secure socket
  • D. Virtual warehouse hosting
  • E. Metadata management
Show answer and explanation ▾

Correct answer: A, E

The cloud services layer is the coordinating tier of Snowflake's architecture. It handles authentication of users and clients, access control, infrastructure management, query parsing and optimization, and metadata management, including the statistics and object definitions that make features such as cloning, Time Travel and pruning possible. Authentication and metadata management therefore both belong to this layer. Actual query execution happens in the compute layer, where virtual warehouses process the data, while the storage layer holds the table data itself.

Why the other options are wrong:

  • B. Query processing (execution) is performed by virtual warehouses in the compute layer, not by cloud services.
  • C. Secure socket is not one of the services that Snowflake lists for the cloud services layer.
  • D. Virtual warehouses make up the separate query processing (compute) layer rather than being hosted as a cloud service.

Question 3

A user is automating a data pipeline using the Snowflake CLI and needs to pass variables into SQL scripts.

Which variable templating syntaxes should be used? (Choose two.)

  • A. STANDARD
  • B. LEGACY
  • C. JINJA
  • D. NONE
  • E. ALL
Show answer and explanation ▾

Correct answer: A, B

When Snowflake CLI runs SQL from a script, it can substitute variables using several templating syntaxes chosen with the enable-templating option. Two are switched on by default and need no extra configuration in an automated pipeline. STANDARD uses the <% variable_name %> form, which Snowflake recommends for new work. LEGACY uses the older &{ variable_name } form inherited from SnowSQL and remains enabled so existing scripts keep working. Together these are the syntaxes a pipeline should rely on when passing values into SQL scripts, with STANDARD as the preferred choice and LEGACY supported for compatibility.

Why the other options are wrong:

  • C. JINJA templating is disabled by default and must be turned on explicitly, so it is not one of the default syntaxes a pipeline should rely on.
  • D. NONE is a setting that turns variable templating off entirely, so it provides no syntax for passing variables.
  • E. ALL is a setting that enables every syntax at once, including JINJA; it is not a templating syntax itself.

Question 4

What is the MINIMUM Snowflake edition needed to apply 90 days of Time Travel to an object?

  • A. Standard
  • B. Enterprise
  • C. Business Critical
  • D. Virtual Private Snowflake (VPS)
Show answer and explanation ▾

Correct answer: B

Standard Edition allows Time Travel retention of only 0 or 1 day. Extended Time Travel, which lets DATA_RETENTION_TIME_IN_DAYS be set as high as 90 days on permanent objects, requires Enterprise Edition or higher. Because the stem asks for the minimum edition that supports a 90-day retention period, Enterprise is the lowest tier that qualifies. Business Critical and Virtual Private Snowflake also allow it, but they sit above Enterprise and are not the minimum.

Why the other options are wrong:

  • A. Standard Edition caps Time Travel retention at 1 day, so 90 days is not possible.
  • C. Business Critical supports 90 days but is a higher tier than the minimum required.
  • D. Virtual Private Snowflake supports 90 days but is the highest tier, not the minimum.

Question 5

Which objects can be recovered after being dropped? (Choose two.)

  • A. View
  • B. Transient table
  • C. Temporary table
  • D. Permanent table
  • E. Materialized view
Show answer and explanation ▾

Correct answer: B, D

Dropped tables are kept for the Time Travel retention period, and UNDROP TABLE restores them as long as that period has not expired. Permanent tables support Time Travel (1 day by default, configurable up to 90 days on Enterprise Edition or higher) followed by Fail-safe. Transient tables also have Time Travel, with a default and maximum of 1 day, so a dropped transient table can be restored within that window even though it has no Fail-safe. Both table types can therefore be recovered after a DROP.

Why the other options are wrong:

  • A. A dropped view cannot be recovered with UNDROP; it must be recreated from its definition.
  • C. A temporary table exists only for its session, has no Fail-safe and is purged when the session ends, so it cannot be recovered once its session is over.
  • E. Snowflake documents that dropped materialized views cannot be recovered and must be recreated.

Question 6

How does Snowflake support storage for external tables?

  • A. Snowflake automatically replicates and stores all external table data.
  • B. Snowflake stores the metadata from external tables.
  • C. Snowflake references external tables using internal directory tables.
  • D. Snowflake references external tables using internal transient tables.
Show answer and explanation ▾

Correct answer: B

An external table lets Snowflake query files that remain in an external stage, such as Amazon S3, Google Cloud Storage or Azure storage. The data files are not loaded into Snowflake storage. Instead, Snowflake keeps file-level metadata about them, such as file names, paths and version identifiers, which it uses to locate and scan the files and which can be refreshed automatically or manually as files change. Storage support for external tables is therefore limited to that metadata.

Why the other options are wrong:

  • A. External table data stays in the customer's cloud storage; Snowflake does not copy or replicate the files.
  • C. Directory tables catalog files on a stage, but they are not the mechanism that defines or stores external tables.
  • D. Transient tables are a native table type for Snowflake-managed data and play no role in referencing external files.

Question 7

Which Snowflake features support the ability to write to a Git repository?

  • A. Snowsight dashboards and stages
  • B. Snowpipe and dynamic tables
  • C. Stored procedures, tasks, and worksheets
  • D. Streamlit applications, workspaces, and notebooks
Show answer and explanation ▾

Correct answer: D

Snowflake's Git integration lets several Snowsight development surfaces both pull from and push to a remote repository. Workspaces can be connected to a Git repository so files can be committed and pushed, notebooks can be synced with a repository so changes are committed back, and Streamlit applications can likewise be connected to Git so app source is versioned there. These are the features that write to Git, whereas a Git repository clone used from SQL is read-only.

Why the other options are wrong:

  • A. Dashboards have no Git push support, and stages store files rather than committing to Git repositories.
  • B. Snowpipe loads files and dynamic tables refresh query results; neither writes to a Git repository.
  • C. Stored procedures and tasks can read code from a Git repository clone, but they do not commit or push to it.

Question 8

A Snowflake Practitioner is tasked with designing a data lakehouse architecture that requires:

1. Support for ACID transactions to ensure data consistency

2. The ability to evolve schemas over time without rewriting existing data

3. Efficient querying capabilities on large datasets

Which table type meets these requirements?

  • A. Permanent
  • B. Iceberg
  • C. External
  • D. Transient
Show answer and explanation ▾

Correct answer: B

Apache Iceberg tables in Snowflake are built on the open Iceberg table format, which is designed for data lakehouse architectures. The format provides ACID transactions so concurrent changes remain consistent, supports schema evolution through metadata changes rather than rewriting existing data files, and keeps file-level statistics that allow efficient pruning on large datasets. Snowflake adds its query engine on top, so the table type satisfies all three stated requirements while keeping the data in open storage.

Why the other options are wrong:

  • A. Permanent tables store data in Snowflake's internal format rather than an open lakehouse format.
  • C. External tables are read-only references to files and do not offer ACID writes or managed schema evolution.
  • D. Transient tables are internal tables without Fail-safe and are not an open lakehouse format.

Question 9

What is SnowSQL used for?

  • A. Developing dashboards
  • B. Loading and unloading data
  • C. Developing Snowflake Native Apps
  • D. Monitoring data transformation activity
Show answer and explanation ▾

Correct answer: B

SnowSQL is Snowflake's command-line client for connecting to an account and executing SQL queries and DDL and DML operations. A common use is bulk data movement: it runs PUT to upload local files to an internal stage before COPY INTO loads them, and GET to download unloaded files from a stage to a local machine. Loading and unloading data is therefore its defining use among the listed options.

Why the other options are wrong:

  • A. Dashboards are built in Snowsight, not in a command-line client.
  • C. Native App development is handled with Snowflake CLI and application packages, not SnowSQL.
  • D. Monitoring of transformation activity is done in Snowsight and with views such as task and query history.

Question 10

What is the impact on storage when a user clones a 1 TB sized table and updates 10% of the rows in the cloned table?

  • A. Only metadata is stored
  • B. Full storage is duplicated
  • C. No additional storage is used
  • D. Only modified data consumes additional storage
Show answer and explanation ▾

Correct answer: D

Cloning is zero-copy: the new table initially points to the same micro-partitions as the source, so creating it adds no storage. When rows in the clone are updated, Snowflake writes new micro-partitions for the changed data, and only those new partitions are owned by and billed to the clone. Updating 10% of the rows therefore adds storage roughly proportional to the micro-partitions rewritten for those changes, while the unchanged data remains shared with the original table.

Why the other options are wrong:

  • A. Metadata-only storage applies at the moment of cloning, but updates write new micro-partitions that consume storage.
  • B. Cloning never copies the full table; unchanged micro-partitions stay shared with the source.
  • C. Modifying rows in the clone creates new micro-partitions, so some additional storage is consumed.

Getting these right? Take the first 58 with you.

The free PDF is the first half of the full Snowflake SnowPro Core COF-C03 pack: 58 questions with every answer explained, as one PDF in your inbox.

Get the free PDF

What people say about the full Snowflake SnowPro Core COF-C03 pack 4.5 from 2

I passed with the Snowflake COF-C03 learning materials, Thank you so much.
Erica

Question 11

What is the recommended use for the COPY INTO <location> SINGLE = TRUE; command?

  • A. To load data without duplications
  • B. To load an individual file
  • C. To copy data and parse the results into single stages
  • D. To unload a single file from a table into a stage
Show answer and explanation ▾

Correct answer: D

COPY INTO <location> unloads table or query results to a stage, and by default it writes data in parallel to several files. Setting SINGLE = TRUE tells Snowflake to write all the output to one file instead. This is useful when a downstream consumer expects exactly one file, and MAX_FILE_SIZE can be raised so the single file can hold larger result sets.

Why the other options are wrong:

  • A. Avoiding duplicate loads depends on load metadata for COPY INTO <table>, not on the SINGLE option.
  • B. SINGLE is an unloading option of COPY INTO <location>; it is not used to load files.
  • C. SINGLE controls the number of output files, not the creation or parsing of stages.

Question 12

Which option will define the column type when unloading data into Parquet-formatted files?

  • A. BINARY_AS_TEXT = TRUE
  • B. EMPTY_FIELD_AS_NULL = FALSE
  • C. CAST or :: in the COPY INTO statement
  • D. SINGLE = TRUE
Show answer and explanation ▾

Correct answer: C

When unloading to Parquet, the column types in the output files follow the data types produced by the SELECT inside COPY INTO <location>. To control the Parquet type of a column, the query casts it explicitly with the CAST function or the :: operator, for example converting a value to a specific numeric or date type. This shapes the file schema without changing the source table.

Why the other options are wrong:

  • A. BINARY_AS_TEXT governs how binary columns are interpreted when reading Parquet, not how output types are defined.
  • B. EMPTY_FIELD_AS_NULL deals with empty fields in delimited files and does not set Parquet column types.
  • D. SINGLE = TRUE unloads output to one file and has no effect on column types.

Question 13

Which copy option is used to include column headers when unloading data to Parquet files?

  • A. PARSE_HEADER = TRUE
  • B. SKIP_HEADER = <integer>
  • C. HEADER = TRUE
  • D. HEADER = ('<header_1>' = '<value_1>')
Show answer and explanation ▾

Correct answer: C

The HEADER copy option of COPY INTO <location> controls whether table column names are written to the unloaded files. For Parquet output, HEADER = TRUE keeps the actual column names from the query in the file schema. When it is left FALSE, the default, the columns are given generic names such as col1 and col2, so setting it to TRUE is what preserves meaningful headers.

Why the other options are wrong:

  • A. PARSE_HEADER is a CSV file format option used when loading to read column names from the first line.
  • B. SKIP_HEADER tells a CSV load how many leading lines to ignore; it does not write headers on unload.
  • D. This syntax is not how HEADER works; the option takes TRUE or FALSE, and custom HTTP style headers belong to other features.

Question 14

How can a user optimize the performance of a COPY INTO <table> query that is being run on a very large file?

  • A. Split the file into smaller files.
  • B. Increase the virtual warehouse size.
  • C. Set the SIZE_LIMIT option to be larger than the file size.
  • D. Set the COMPRESSION = AUTO option in the file format.
Show answer and explanation ▾

Correct answer: A

COPY INTO <table> achieves its speed by loading many files in parallel across the threads of a warehouse. A single very large file cannot be spread across that parallelism efficiently, so most of the warehouse sits idle. Splitting the file into multiple smaller files, close to the recommended compressed size of roughly 100 to 250 MB each, lets Snowflake distribute the work and greatly improves load throughput.

Why the other options are wrong:

  • B. A bigger warehouse adds parallel capacity, but one large file cannot use it effectively, so splitting the file comes first.
  • C. SIZE_LIMIT caps the amount of data a COPY statement loads; raising it does not make loading faster.
  • D. COMPRESSION = AUTO only detects the compression type of the files; it does not speed up the load.

Question 15

Which tools, services, or features should be used to write code when using the Snowpark API? (Choose two.)

  • A. Snowpark Container Services
  • B. Snowpark for Scala
  • C. Snowpark for Python
  • D. Streamlit in Snowflake
  • E. SQL Scripting
Show answer and explanation ▾

Correct answer: B, C

The Snowpark API is offered as client libraries for Java, Python and Scala. Developers write DataFrame code in one of these languages, and Snowpark translates it into SQL that runs inside Snowflake on a warehouse. Snowpark for Scala and Snowpark for Python are therefore the listed options used to write code against the Snowpark API, and functions and procedures can also be built with them.

Why the other options are wrong:

  • A. Snowpark Container Services runs containerized applications on compute pools; it is not a language library for the Snowpark API.
  • D. Streamlit in Snowflake is an app framework that can call Snowpark, but it is not one of the Snowpark API libraries.
  • E. SQL Scripting is Snowflake's procedural SQL extension, separate from the Snowpark API.

Question 16

Which function is used to retrieve all errors of the past execution of a COPY INTO <table> command?

  • A. LIST
  • B. VALIDATE
  • C. RESULT_SCAN
  • D. COPY_HISTORY
Show answer and explanation ▾

Correct answer: B

The VALIDATE table function checks the files processed by a previous COPY INTO <table> execution and returns every error encountered during that load. It is called with the table name and a job ID, which can be the query ID of the COPY statement or the value _last for the most recent load in the session. This lets users review all rejected rows and their error messages after the load has already finished.

Why the other options are wrong:

  • A. LIST shows the files present in a stage; it does not report load errors.
  • C. RESULT_SCAN returns the result set of a previous query, which for COPY is a summary per file, not all errors.
  • D. COPY_HISTORY shows load activity per file, including only the first error, rather than every error.

Question 17

Which table type is supported by Open Catalog to provide centralized, read and write access to a table?

  • A. Hybrid
  • B. Iceberg
  • C. Dynamic
  • D. Event
Show answer and explanation ▾

Correct answer: B

Snowflake Open Catalog is a managed service for Apache Polaris, an open-source catalog that implements the Apache Iceberg REST catalog specification. It gives centralized governance and read and write access to Apache Iceberg tables from many engines, such as Snowflake, Apache Spark and Trino, so different tools can work on the same table data. Iceberg is therefore the table type Open Catalog supports.

Why the other options are wrong:

  • A. Hybrid tables are Snowflake-native tables for transactional workloads and are not managed by Open Catalog.
  • C. Dynamic tables are Snowflake objects refreshed from a query, not tables cataloged by Open Catalog.
  • D. Event tables store logs and traces from Snowflake code and are unrelated to Open Catalog.

Question 18

Which view or command provides the MOST detailed storage information for a table?

  • A. SHOW TABLES
  • B. DESC TABLE
  • C. ACCOUNT_USAGE.STORAGE_USAGE
  • D. ACCOUNT_USAGE.TABLE_STORAGE_METRICS
Show answer and explanation ▾

Correct answer: D

The ACCOUNT_USAGE.TABLE_STORAGE_METRICS view reports storage for each table broken down into active bytes, Time Travel bytes, Fail-safe bytes and bytes retained for clones, along with related lifecycle details. This breakdown shows how storage is attributed between a table and the history it keeps, which makes it the most detailed table-level storage source among the options listed.

Why the other options are wrong:

  • A. SHOW TABLES gives row counts and total bytes but no breakdown across Time Travel, Fail-safe or clones.
  • B. DESC TABLE describes columns and their properties, not storage usage.
  • C. STORAGE_USAGE reports daily storage totals for the whole account, not per table.

Question 19

Which feature allows a user to review the monthly projected costs of a Snowflake resource against the monthly actual costs?

  • A. Budgets
  • B. Auto-suspend
  • C. Scaling Policy
  • D. Resource monitor
Show answer and explanation ▾

Correct answer: A

Budgets define a monthly spending limit for the account or for a custom group of objects and track credit usage against it. Snowsight shows current spending alongside a forecast of the month's projected spending, and Snowflake can send notifications when projected spending is likely to exceed the limit. That comparison of projected and actual monthly cost is exactly what the stem describes.

Why the other options are wrong:

  • B. Auto-suspend stops an idle warehouse to save credits; it provides no cost reporting or projection.
  • C. A scaling policy controls how a multi-cluster warehouse starts and shuts down clusters, not cost tracking.
  • D. A resource monitor tracks warehouse credit usage against a quota and can suspend warehouses, but it does not show projected spending.

Question 20

How can a user work with multiple assigned roles without the need to frequently switch roles?

  • A. Use secondary roles and primary roles.
  • B. Grant the user the ACCOUNTADMIN role.
  • C. Grant the user the parent role of the roles that have been assigned.
  • D. Create a new role and grant it to the user, then grant the existing roles to the new role.
Show answer and explanation ▾

Correct answer: A

Every session has one primary role, and secondary roles let a user activate additional granted roles at the same time, for example with USE SECONDARY ROLES ALL. Privileges from the secondary roles are then combined with those of the primary role for most operations, while the primary role governs object creation and ownership. This lets the user work with several assigned roles without switching between them repeatedly.

Why the other options are wrong:

  • B. Granting ACCOUNTADMIN gives excessive privileges and breaks least privilege rather than combining the existing roles.
  • C. Granting a parent role changes the hierarchy and grants extra access, rather than using the roles already assigned.
  • D. Building a new wrapper role adds administrative overhead and alters the hierarchy, while secondary roles achieve the goal directly.

Question 21

If a query is being used to unload a 1 TB table into a stage, which keyword will be shown in the Query History?

  • A. INSERT
  • B. UNLOAD
  • C. COPY
  • D. UPDATE
Show answer and explanation ▾

Correct answer: C

Snowflake unloads data with the COPY INTO <location> command, which writes table or query results into an internal or external stage. There is no separate unload statement in Snowflake SQL, so the query text recorded in Query History for unloading a large table begins with the COPY keyword. The same command family is used for loading, distinguished by whether the target is a table or a location.

Why the other options are wrong:

  • A. INSERT adds rows into a table; it does not write data out to a stage.
  • B. Snowflake has no UNLOAD SQL statement; unloading is performed through COPY INTO <location>.
  • D. UPDATE modifies existing table rows and has nothing to do with unloading data.

Question 22

What can load data from an external stage immediately after the data is taken in?

  • A. COPY command scheduled with a Task
  • B. Snowflake Kafka Connector
  • C. Auto-ingest Snowpipe
  • D. External table
Show answer and explanation ▾

Correct answer: C

Snowpipe with auto-ingest enabled loads files from an external stage as soon as they arrive. The cloud provider's event notification service, such as Amazon S3 event notifications delivered through SQS, alerts Snowflake when a new file lands, and Snowpipe queues and loads it using serverless compute. No schedule or manual command is needed, so data is loaded shortly after it is written to the stage.

Why the other options are wrong:

  • A. A task runs COPY on a schedule, so files wait until the next run instead of loading immediately on arrival.
  • B. The Kafka connector reads from Kafka topics, not from files arriving in an external stage.
  • D. An external table queries files in place; it does not load data into Snowflake tables.

Question 23

Which methods for a second factor of authentication does Snowflake support? (Choose two.)

  • A. Email codes
  • B. Authenticator applications
  • C. Social logins
  • D. Passkeys
  • E. Security questions
Show answer and explanation ▾

Correct answer: B, D

Snowflake multi-factor authentication lets a user enroll a second factor after the password. Authenticator applications generate time-based one-time passcodes that the user enters at sign-in, and passkeys use a device-bound credential, such as a platform biometric or a hardware security key, to confirm the user's identity. Both are supported second-factor methods that users can enroll in Snowsight, alongside Duo Mobile.

Why the other options are wrong:

  • A. Snowflake does not offer codes sent by email as an MFA second factor.
  • C. Social logins are not a Snowflake authentication method; federated sign-in uses SAML SSO with an identity provider.
  • E. Security questions are not a Snowflake second factor for authentication.

Question 24

A user needs to load Snowflake files from a stage that contains data from the past 5 years while using the LEAST amount of operational overhead. The file names include the year in this format: SALES_2025_01.gz.

What command should be used to load the data from 2024?

  • A. The COPY INTO command to individually load the needed files.
  • B. The COPY INTO command using the PATTERN parameter to define the needed files.
  • C. The REMOVE command to eliminate unwanted files, then use the COPY INTO command.
  • D. The GET command to download the needed files, then create a new stage and load the files onto the stage using the COPY INTO command.
Show answer and explanation ▾

Correct answer: B

The year is embedded in each file name, such as SALES_2025_01.gz, so a regular expression can select only the 2024 files. The PATTERN option of COPY INTO <table> filters the staged files by name, for example matching names that contain SALES_2024_. A single statement then loads all the required files without listing each one and without altering or moving the stage contents, which keeps operational effort to a minimum.

Why the other options are wrong:

  • A. Naming each 2024 file individually works but requires far more manual effort than one pattern.
  • C. Removing files from the stage is destructive, loses the other years' data and adds steps.
  • D. Downloading and restaging files adds several unnecessary steps and much more overhead.

Question 25

An organization is preparing for a data structure change on a critical table, SALES_BUDGET, and needs to identify which users, dashboards, reports and processes might be affected.

What action should be performed to determine the impact of this change?

  • A. Create a stream on SALES_BUDGET and check for inserted or updated rows to identify the users responsible for data access.
  • B. Use SHOW GRANTS ON SALES_BUDGET to see which roles or users accessed the table and track their query frequency.
  • C. Query SNOWFLAKE.ACCOUNT_USAGE.ACCESS_HISTORY to see who accessed the SALES_BUDGET table and what columns were read.
  • D. Review TABLE_STORAGE_METRICS in the INFORMATION_SCHEMA to estimate read-frequency based on the table size and compression changes.
Show answer and explanation ▾

Correct answer: C

The SNOWFLAKE.ACCOUNT_USAGE.ACCESS_HISTORY view records, for each query, which user ran it and which tables and columns were read or written. Querying it for SALES_BUDGET shows the users, service accounts and the queries behind dashboards, reports and pipelines that depend on the table, down to the specific columns they read. That is exactly what is needed to judge which consumers a structural change would affect.

Why the other options are wrong:

  • A. A stream tracks row changes made to a table, not who read the table or which columns they used.
  • B. SHOW GRANTS lists privileges that have been granted, not actual access or query frequency.
  • D. TABLE_STORAGE_METRICS reports storage bytes and gives no information about who reads the table.

Question 26

Which feature allows users to audit how sensitive data is being accessed over time?

  • A. Column-level security
  • B. Row-level security
  • C. Tag-based masking policy
  • D. Object tagging
Show answer and explanation ▾

Correct answer: D

Object tagging lets data stewards label tables and columns that hold sensitive data and then track that data for compliance and auditing. Because tag assignments are recorded as metadata, they can be joined with access records such as ACCESS_HISTORY to see over time which users queried tagged objects and columns. The tags are what make it possible to focus an audit on sensitive data rather than on every object in the account.

Why the other options are wrong:

  • A. Column-level security masks values when queried; it protects data rather than auditing access.
  • B. Row-level security filters the rows a role can see but does not track usage over time.
  • C. A tag-based masking policy applies masking to tagged columns; its job is protection, not auditing.

Question 27

Which action can be performed by the SYSADMIN role without requiring elevated privileges from the ACCOUNTADMIN or SECURITYADMIN role?

  • A. Granting privileges
  • B. Creating a new user
  • C. Managing virtual warehouses
  • D. Viewing billing and usage information
Show answer and explanation ▾

Correct answer: C

SYSADMIN is the system role intended for creating and managing objects. It holds the privileges to create warehouses, databases and other objects by default, so it can create, resize, suspend and otherwise manage virtual warehouses without help from a higher role. Creating users belongs to USERADMIN, account-wide grant management belongs to SECURITYADMIN through MANAGE GRANTS, and billing and usage data is available to ACCOUNTADMIN unless access is explicitly granted to others.

Why the other options are wrong:

  • A. Broad privilege granting relies on MANAGE GRANTS, held by SECURITYADMIN; SYSADMIN can only grant on objects it owns.
  • B. Creating users requires the CREATE USER privilege, which is held by USERADMIN and roles above it, not SYSADMIN.
  • D. Billing and usage information is visible to ACCOUNTADMIN by default and requires explicit grants for other roles.

Question 28

Which Snowflake feature can be used to identify tables, views, and columns that contain sensitive information by assigning and querying metadata?

  • A. Tags
  • B. External Tokens
  • C. Row access policies
  • D. Tag-based masking policies
Show answer and explanation ▾

Correct answer: A

Tags are schema-level objects that can be assigned as key and value pairs to tables, views and columns, among other objects. Data stewards can tag objects that hold sensitive data, for example with a classification such as PII, and then find them later by querying tag references through functions and Account Usage views. Assigning and querying this metadata is exactly how tags identify where sensitive information lives.

Why the other options are wrong:

  • B. External Tokenization replaces sensitive values with tokens through an external function; it does not label or locate data.
  • C. Row access policies filter which rows a role can see rather than identifying sensitive objects.
  • D. Tag-based masking policies protect columns that carry tags; the identification itself comes from the tags.

Question 29

Which interface or service allow users to connect to Snowflake with Python? (Choose two.)

  • A. Snowpark API
  • B. Snowsight
  • C. Streamlit in Snowflake
  • D. AI_EXTRACT
  • E. Snowflake CLI
Show answer and explanation ▾

Correct answer: A, E

The Snowpark API includes a Python library that opens a session to Snowflake from any Python environment, so code can query and transform data using DataFrames. Snowflake CLI is an open-source command-line tool built in Python and installed with pip, and it connects to Snowflake to run SQL and to deploy Python-based objects such as Snowpark functions, procedures and Streamlit apps. Both give a user a Python-based way to connect to Snowflake from their own environment.

Why the other options are wrong:

  • B. Snowsight is Snowflake's browser interface, not a Python connection method.
  • C. Streamlit in Snowflake runs apps that are hosted inside Snowflake rather than serving as a client that connects to it.
  • D. AI_EXTRACT is a Cortex AI SQL function for extracting information, not a way to connect to Snowflake.

That was 29 of 116.

The full Snowflake SnowPro Core COF-C03 pack has all 116 questions, each with the answer, the explanation and why the other options are wrong, plus a questions-only copy for timed runs. $39.00 USD, paid once, with free monthly updates and a pass-or-your-money-back guarantee.

Get the full pack

Not ready yet? Get the first 58 free as one PDF by email. Get the Half Pack →

About the COF-C03 exam

These 29 Snowflake SnowPro Core COF-C03 questions are free here. The free PDF, the first half of the pack, has 58 free by email, and the full pack has all 116. Each step includes the one before it.

They are drawn from the same bank as the full COF-C03 pack, which has 116 questions in total.

Get the full COF-C03 question bank (116 questions) →

Back to blog