Skip to content
TryHired.
Interview prep · Technical · checked against the official docs

Snowflake interview questions: answers checked against the docs

The cloud data warehouse questions data engineers and analysts get, answered in plain words and checked against the Snowflake documentation.

Updated Nov 16, 2026 · Facts checked against the official documentation · questions: TryHired
Quick answer

Snowflake interviews cover the three-layer architecture, virtual warehouses and cost, micro-partitions and pruning, Time Travel and Fail-safe, zero-copy cloning, clustering keys, concurrency with multi-cluster warehouses, the result cache, streams and tasks, Snowpipe, Secure Data Sharing, access control and semi-structured data. Below are 14 questions; each answer lists the Snowflake documentation pages it was checked against.

Basic

Basic Snowflake interview questions

Advice

Basic1. What is Snowflake's architecture?

Three layers that scale independently: storage (data kept in Snowflake's compressed columnar format in cloud storage), compute (virtual warehouses that run queries) and cloud services (metadata, optimization, security and transactions). Snowflake describes it as a hybrid of shared-disk and shared-nothing designs: all warehouses see the same data, while each warehouse computes on its own. That is why many teams can query the same data without competing for compute.

Checked against: Snowflake docs: key concepts and architecture

Basic2. What is a virtual warehouse, and how is it billed?

A virtual warehouse is a cluster of compute that runs queries and DML. You choose a size, and each step up doubles the credits used per hour. Billing is per second while it runs (with a minimum each time it starts), so auto-suspend after a short idle period and auto-resume are the main cost controls.

SQL snippet · illustrative, not run on this page
CREATE WAREHOUSE reporting_wh
  WAREHOUSE_SIZE = 'SMALL'
  AUTO_SUSPEND = 60
  AUTO_RESUME = TRUE;

Checked against: Snowflake docs: key concepts and architecture, Snowflake docs: virtual warehouses

Basic3. What are micro-partitions, and how does pruning work?

Snowflake automatically splits table data into micro-partitions of 50 to 500 MB of uncompressed data and keeps metadata about each one, such as the range of values in every column. A query with a filter can skip every micro-partition whose range cannot match, which is called pruning. You do not define partitions yourself.

Checked against: Snowflake docs: micro-partitions

Basic4. What is Time Travel?

Time Travel lets you query, clone or restore data as it was at an earlier point within the retention period: by default 1 day, and up to 90 days on Enterprise Edition and higher. You can query AT a timestamp or BEFORE a statement, and UNDROP a dropped table, schema or database.

SQL snippet · illustrative, not run on this page
SELECT * FROM orders AT(OFFSET => -60*30);           -- 30 minutes ago
UNDROP TABLE orders;
CREATE TABLE orders_before_fix CLONE orders BEFORE(STATEMENT => '<query_id>');

Checked against: Snowflake docs: Time Travel

Intermediate

Intermediate Snowflake interview questions

Advice

Intermediate5. What is the difference between Time Travel and Fail-safe?

Time Travel is self-service recovery within the retention period you set. Fail-safe is a further fixed 7-day period after that, during which only Snowflake can recover data, on request; it is not meant for routine access. Transient and temporary tables have no Fail-safe, which saves storage cost for data you can rebuild.

Checked against: Snowflake docs: Fail-safe, Snowflake docs: temporary and transient tables

Intermediate6. What is zero-copy cloning?

CREATE ... CLONE creates a copy of a table, schema or database instantly, because the clone shares the same underlying storage as the source; only data that is later changed in either one takes new storage. Teams use it for development and test copies of production data and for backups before risky changes.

SQL snippet · illustrative, not run on this page
CREATE DATABASE analytics_dev CLONE analytics;

Checked against: Snowflake docs: cloning

Intermediate7. When would you use a clustering key?

Only for very large tables where queries filter on columns that the data's natural load order does not group well, so pruning is poor. A clustering key tells Snowflake which columns to co-locate in micro-partitions, and automatic clustering keeps it that way, at a cost in credits. The docs say clustering keys are not intended for all tables, so check query pruning first.

SQL snippet · illustrative, not run on this page
ALTER TABLE events CLUSTER BY (event_date, customer_id);

Checked against: Snowflake docs: clustering keys

Intermediate8. How does Snowflake handle many concurrent users?

Give different workloads separate warehouses (loading, BI, data science), so they do not slow each other down. For many simultaneous queries on one workload, a multi-cluster warehouse adds clusters automatically as concurrency rises and removes them when it falls. Scaling up (a larger size) helps a single heavy query; scaling out (more clusters) helps concurrency.

Checked against: Snowflake docs: multi-cluster warehouses

Intermediate9. How does the result cache work?

Snowflake keeps query results for 24 hours. If the same query runs again and the underlying data has not changed (along with other conditions), it can return the stored result without using a warehouse. Results can also be post-processed with RESULT_SCAN.

Checked against: Snowflake docs: persisted query results

Advanced

Advanced Snowflake interview questions

Advice

Advanced10. What are streams and tasks?

A stream records the inserts, updates and deletes made to a table since an offset, which is change data capture: consuming the stream in a DML statement advances the offset. A task runs SQL on a schedule or when triggered, for example when a stream has new data, and tasks can be chained into task graphs. Together they build incremental pipelines inside Snowflake.

SQL snippet · illustrative, not run on this page
CREATE STREAM orders_changes ON TABLE raw.orders;

CREATE TASK merge_orders
  WAREHOUSE = etl_wh
  SCHEDULE = '5 MINUTE'
  WHEN SYSTEM$STREAM_HAS_DATA('orders_changes')
AS
  MERGE INTO analytics.orders t USING orders_changes s ON t.id = s.id
  WHEN MATCHED THEN UPDATE SET t.status = s.status
  WHEN NOT MATCHED THEN INSERT (id, status) VALUES (s.id, s.status);

Checked against: Snowflake docs: streams, Snowflake docs: tasks

Advanced11. What is Snowpipe, and when would you use it instead of COPY INTO?

COPY INTO loads files in batches on a warehouse you run. Snowpipe wraps a COPY statement in a pipe and runs it for each new file as it arrives, on serverless compute that Snowflake manages, so it suits files that arrive continuously or at unpredictable times. For streaming rows rather than files, Snowflake has separate streaming ingestion options.

Checked against: Snowflake docs: Snowpipe

Advanced12. How does Secure Data Sharing work?

A provider creates a share and grants it access to specific objects; a consumer account creates a read-only database from the share. No data is copied or moved, so consumers see the provider's live data immediately, and the provider controls access centrally.

Checked against: Snowflake docs: Secure Data Sharing

Advanced13. How does access control work in Snowflake?

Privileges are granted to roles, and roles are granted to users or to other roles, forming a hierarchy (role-based access control). Every object also has an owner role that can grant access to it (discretionary access control). Good practice is to grant privileges to functional roles and never to use ACCOUNTADMIN for daily work.

SQL snippet · illustrative, not run on this page
GRANT USAGE ON DATABASE analytics TO ROLE analyst;
GRANT SELECT ON ALL TABLES IN SCHEMA analytics.marts TO ROLE analyst;
GRANT ROLE analyst TO USER ava;

Checked against: Snowflake docs: access control

Advanced14. How do you query semi-structured data such as JSON?

Load it into a VARIANT column, which can hold any value, including objects and arrays. Read fields with path notation and cast them, and use LATERAL FLATTEN to turn an array into rows.

SQL snippet · illustrative, not run on this page
SELECT raw:customer.id::NUMBER AS customer_id,
       item.value:sku::STRING  AS sku
FROM orders_json,
     LATERAL FLATTEN(input => raw:items) item;

Checked against: Snowflake docs: semi-structured data

Practice

How to practice with this page

Advice
  • Explain each cost question in credits: what runs, for how long and on what size.
  • Practice a stream-and-task pipeline in a trial account.
  • Write a query on JSON with LATERAL FLATTEN.
  • Review SQL: day-to-day Snowflake work is largely SQL.
Next

Related interview prep and tools

Disclosure: TryHired makes HIRED, a paid job-search program, so we have a conflict of interest when we mention it. This question bank is free, with no TryHired signup.

FAQ

Questions people ask

What Snowflake topics come up often in interviews?
Architecture and virtual warehouses, micro-partitions and pruning, clustering, Time Travel and Fail-safe, cloning, streams and tasks, Snowpipe, data sharing, caching, access control and cost management.
Do I need SQL skills for a Snowflake interview?
Yes. Snowflake work is largely SQL, so expect joins, window functions and query tuning alongside the platform questions; the SQL interview page covers those.
Were these snippets run?
No. No Snowflake account was used, so the SQL snippets are illustrations and no output is shown. Object names are examples, and features can depend on your edition.
Method

How this page is built

TryHired wrote these questions and answers. Statements about the technology were checked against the official documentation pages listed below, saved when the page was built (2026-11-16). The code snippets are short illustrations and were not run on this page. The question list is general practice: it does not predict what a particular employer will ask.

What this page can't tell you
  • Interviewers ask different questions for different roles and levels.
  • Answers are short explanations, not complete references.
  • No Snowflake account was used; features and retention limits depend on the edition, as the answers note.
Sources
  1. Snowflake docs: access control (read 2026-11-15)
  2. Snowflake docs: key concepts and architecture (read 2026-11-15)
  3. Snowflake docs: persisted query results (read 2026-11-15)
  4. Snowflake docs: cloning (read 2026-11-15)
  5. Snowflake docs: clustering keys (read 2026-11-15)
  6. Snowflake docs: Fail-safe (read 2026-11-15)
  7. Snowflake docs: micro-partitions (read 2026-11-15)
  8. Snowflake docs: multi-cluster warehouses (read 2026-11-15)
  9. Snowflake docs: semi-structured data (read 2026-11-15)
  10. Snowflake docs: Secure Data Sharing (read 2026-11-15)
  11. Snowflake docs: Snowpipe (read 2026-11-15)
  12. Snowflake docs: streams (read 2026-11-15)
  13. Snowflake docs: tasks (read 2026-11-15)
  14. Snowflake docs: temporary and transient tables (read 2026-11-15)
  15. Snowflake docs: Time Travel (read 2026-11-15)
  16. Snowflake docs: virtual warehouses (read 2026-11-15)