You Can Pass This Exam For Free
Choose Your Study Path
Limited data platform experience. You need to build foundational knowledge of cloud data warehousing and Snowflake concepts from scratch.
Week 1Sign up for a Snowflake free trial. Learn the three-layer architecture: cloud services, compute (virtual warehouses), and storage. Understand how Snowflake differs from traditional databases
Week 2Study micro-partitions, columnar storage, and automatic clustering. Learn about table types: permanent, transient, temporary, external, and Iceberg tables
Week 3Hands-on with virtual warehouses: create, resize, configure auto-suspend and auto-resume. Understand multi-cluster warehouses and scaling policies (standard vs economy)
Week 4Learn data loading end-to-end: stages (internal, external, table), file formats (CSV, JSON, Parquet, Avro, ORC), COPY INTO command, Snowpipe for continuous loading
Week 5Study account management: roles hierarchy (ACCOUNTADMIN, SYSADMIN, SECURITYADMIN, USERADMIN, PUBLIC), RBAC, dynamic data masking, row access policies, and resource monitors
Week 6Cover data protection features: Time Travel, Fail-safe, zero-copy cloning, data encryption. Learn Snowflake editions and which features each includes
Week 7Study data collaboration: Secure Data Sharing, Snowflake Marketplace, Data Exchange, data clean rooms, and native apps. Cover Cortex AI basics and Snowflake Notebooks
Week 8Learn performance optimization: query profiling, caching layers (metadata, result, warehouse), clustering keys, and materialized views. Study streams and tasks for ETL
Week 9Take full-length practice exams. Review all incorrect answers. Focus extra time on Domain 1 (31%) and Domain 4 (21%) — they carry the most weight
Week 10Final review: revisit confusing concepts, re-test weak domains. Ensure you can explain warehouse sizing, caching, Time Travel vs Fail-safe, and role hierarchy
Exam Overview
Format
100 multiple-choice and multiple-select questions, 115 minutes. No performance-based questions.
Scoring
Scaled score 0-1000. Passing: 750. No penalty for wrong answers — always answer every question.
Domains & Weights
- Snowflake AI Data Cloud Features and Architecture31%
- Account Management and Data Governance20%
- Data Loading, Unloading, and Connectivity18%
- Performance Optimization, Querying, and Transformation21%
- Data Collaboration10%
Registration
$175 USD. Available at Pearson VUE testing centers or online proctored from home. Exam fee is $175 USD. Certification valid for 2 years.
Topic Priority Table
Not all topics are tested equally. Focus your study time on Tier 1 first, then Tier 2. Tier 3 topics rarely appear — just recognize what they do.
Domain 131% of exam
Snowflake AI Data Cloud Features and Architecture
The heaviest domain at 31%. Covers Snowflake's unique architecture, the three-layer separation of concerns, compute and storage concepts, table types, data protection features, and the new COF-C03 topics including Cortex AI, Iceberg tables, and Notebooks. This domain tests your fundamental understanding of how Snowflake works under the hood.
Key Topics
Three-Layer ArchitectureVirtual WarehousesMulti-Cluster WarehousesMicro-PartitionsTime TravelFail-safeZero-Copy CloningSnowflake EditionsCortex AIIceberg TablesSnowflake Notebooks
Must-Know Concepts
- Three-layer architecture: cloud services (metadata, security, optimization), compute (virtual warehouses), and storage (micro-partitions in cloud object storage) — each scales independently
- Virtual warehouse sizing from XS to 6XL: credits double with each size increase (XS=1, S=2, M=4, L=8, XL=16). Auto-suspend and auto-resume behavior and timing
- Multi-cluster warehouses: Standard scaling (prioritize performance) vs Economy scaling (conserve credits). Requires Enterprise edition or higher
- Micro-partitions: 50-500MB immutable columnar units, automatically managed. Metadata tracks min/max values for partition pruning
- Table types: permanent (full TT + 7-day FS), transient (TT 0-1 day, 0-day FS), temporary (session-scoped, TT 0-1 day, 0-day FS), external (read-only, no TT/FS)
- Time Travel retention: Standard edition max 1 day, Enterprise+ up to 90 days. Use AT/BEFORE clauses and UNDROP for recovery
- Fail-safe: 7-day non-configurable period AFTER Time Travel. Only Snowflake support can access it. Not available for transient/temporary tables
- Zero-copy cloning: metadata-only copy, no storage cost until data diverges. Can clone databases, schemas, tables, streams, tasks, and sequences
- Snowflake editions: Standard, Enterprise, Business Critical, VPS. Know which features require which edition (e.g., multi-cluster = Enterprise, customer-managed keys = Business Critical)
- Cortex AI: SQL-callable AI functions, Cortex Search, Cortex Analyst. Know what it does and where it fits, not the API syntax
- Iceberg tables: open table format for multi-engine interoperability. Know why they exist vs native tables, not operational mechanics
- Data encryption: always-on encryption at rest (AES-256) and in transit (TLS 1.2). Tri-Secret Secure and customer-managed keys available on Business Critical+
Common Exam Traps
Fail-safe is NOT user-accessible. Only Snowflake support can recover data from Fail-safe. If a question asks about user-initiated recovery, the answer is Time Travel, not Fail-safe
Transient tables have 0-day Fail-safe. Temporary tables ALSO have 0-day Fail-safe AND are session-scoped. Do not confuse their data protection levels
Zero-copy cloning does NOT consume additional storage until data in the clone diverges from the source. But clones are independent objects — changes to one do not affect the other
Virtual warehouse credits double with each size increase, but query performance does NOT always double. Larger warehouses help with complex queries, not simple ones
Auto-suspend timer resets with each query. A warehouse set to auto-suspend after 5 minutes will stay running as long as queries arrive within 5-minute intervals
Domain 220% of exam
Account Management and Data Governance
Covers account setup, role-based access control, system-defined and custom roles, user management, security features like network policies and MFA, and data governance capabilities including masking policies, row access policies, tags, and object tagging. Understanding the role hierarchy and governance features is critical for this domain.
Key Topics
RBAC Role HierarchyACCOUNTADMINSYSADMINSECURITYADMINDynamic Data MaskingRow Access PoliciesResource MonitorsNetwork PoliciesObject TagsData Classification
Must-Know Concepts
- System-defined roles hierarchy: ACCOUNTADMIN > SECURITYADMIN > USERADMIN, and ACCOUNTADMIN > SYSADMIN. Custom roles should be granted to SYSADMIN for object access
- ACCOUNTADMIN: top-level role combining SYSADMIN + SECURITYADMIN. Should be assigned to limited users with MFA enabled. Not for daily use
- SECURITYADMIN: manages grants and can revoke access from any role. USERADMIN: creates and manages users and roles. SYSADMIN: creates and manages objects (warehouses, databases)
- PUBLIC role: automatically granted to every user. Use it carefully as any privilege on PUBLIC is available to all users
- Resource monitors: set credit quotas at account or warehouse level. Actions: notify, suspend after current queries finish, or suspend immediately. ACCOUNTADMIN creates them
- Dynamic data masking: column-level security. Define masking policies with conditions based on the executing role. Applied at query time, transparent to the query itself
- Row access policies: row-level security filtering. Define conditions that determine which rows are visible to which roles
- Object tags: metadata labels applied to databases, schemas, tables, columns. Used for governance, classification, and tracking sensitive data
- Network policies: IP allow/block lists at account or user level. Can restrict where users connect from
- Multi-factor authentication (MFA): supported for all users, required best practice for ACCOUNTADMIN. Uses Duo Mobile
- Secondary roles: allow a session to inherit privileges from multiple roles simultaneously without switching. New COF-C03 topic
- Data classification: automatic detection and tagging of sensitive data like PII, using Snowflake's built-in classification functions
Common Exam Traps
ACCOUNTADMIN is NOT a true superuser. It only sees objects if the role hierarchy grants access. But it CAN see all grants and manage all account-level settings
Custom roles must be granted to SYSADMIN (directly or through hierarchy) for SYSADMIN to manage the objects they own. Orphan roles outside the hierarchy are a governance risk
Resource monitors track credit USAGE, not storage costs. They cannot monitor or limit storage expenses
Masking policies apply at QUERY TIME based on the current role, not when data is written. Changing roles can reveal different data in the same column
USERADMIN can create roles but cannot grant privileges on objects. SECURITYADMIN can grant privileges but typically delegates user creation to USERADMIN
Domain 318% of exam
Data Loading, Unloading, and Connectivity
Covers all methods of getting data into and out of Snowflake: COPY INTO for bulk loading/unloading, Snowpipe for continuous loading, stages (internal and external), file formats, data transformation during load, connectors, and drivers. Also covers semi-structured data handling with the VARIANT data type.
Key Topics
COPY INTOSnowpipeInternal StagesExternal StagesFile FormatsPUT/GET CommandsVARIANT Data TypeSnowflake ConnectorsExternal FunctionsStorage Integrations
Must-Know Concepts
- COPY INTO for loading: loads from stages into tables. Key options: ON_ERROR (CONTINUE, SKIP_FILE, ABORT_STATEMENT), VALIDATION_MODE, PURGE, FORCE, FILE_FORMAT
- COPY INTO for unloading: exports data from tables to stages. Supports partitioning output files and specifying file formats
- Stage types: user stage (@~), table stage (@%tablename), named internal stage (@stagename), external stage (S3, GCS, Azure). User and table stages cannot be altered or dropped
- File formats: CSV, JSON, Avro, ORC, Parquet, XML. Can be defined inline in COPY or as named file format objects for reuse
- Snowpipe: serverless continuous loading. Auto-ingest mode uses cloud event notifications. Uses serverless compute (not warehouse credits). Billed per GB ingested (simplified pricing as of December 2025)
- PUT command: uploads files from local machine to an internal stage. Only works from SnowSQL or JDBC/ODBC. Does NOT work in the web UI
- GET command: downloads files from an internal stage to a local machine. Same connectivity requirements as PUT
- VARIANT data type: stores semi-structured data (JSON, Avro, ORC, Parquet, XML). Access nested fields using dot notation or bracket notation
- FLATTEN function: converts nested semi-structured data (arrays and objects) into relational rows for querying
- Storage integrations: named objects that store cloud provider credentials for external stage access. Avoid embedding credentials directly in stage definitions
- Snowflake connectors: Python, Spark, Kafka, JDBC, ODBC, Node.js, .NET, Go. Know which to use for each integration scenario
- Transformation during load: COPY INTO supports SELECT with column reordering, casting, and simple expressions during the load process
Common Exam Traps
PUT/GET commands ONLY work from SnowSQL CLI or JDBC/ODBC connections. They do NOT work in the Snowsight web interface or through the Snowflake REST API
Snowpipe uses SERVERLESS compute, which is billed differently from virtual warehouse compute. As of December 2025, Snowpipe uses a simplified per-GB pricing model (0.0037 credits per GB ingested), not warehouse size or per-file charges
User stages (@~) and table stages (@%table) cannot be dropped or altered. Only named stages can be configured with file formats and other properties
VALIDATION_MODE in COPY INTO does NOT actually load data — it only validates the files. This is a common exam trap where candidates think validation mode performs a partial load
The VARIANT column stores semi-structured data as a single column. Use LATERAL FLATTEN to explode arrays into rows, not regular FLATTEN alone
Domain 421% of exam
Performance Optimization, Querying, and Transformation
Covers query performance analysis using the Query Profile, caching mechanisms, warehouse optimization, clustering strategies, materialized views, streams and tasks for ETL automation, and Snowpark for programmatic data transformation. Understanding how to diagnose and resolve performance issues is the core skill tested.
Key Topics
Query ProfileResult CacheWarehouse CacheMetadata CacheClustering KeysMaterialized ViewsStreamsTasksSnowparkQuery History
Must-Know Concepts
- Three caching layers: metadata cache (cloud services, instant count/min/max), result cache (24-hour reuse of identical queries, no warehouse needed), warehouse cache (SSD on compute nodes, requires running warehouse)
- Result cache conditions: same query text, same role, no underlying data changes, no changed session parameters. Result cache is free — no compute cost
- Query Profile: visual execution plan showing operators, data flow, statistics, and spilling. Use it to identify bottlenecks, exploding joins, and inefficient partition pruning
- Partition pruning: Snowflake uses micro-partition metadata to skip irrelevant partitions. Effective pruning reduces data scanned and improves performance
- Clustering keys: define columns for data co-location within micro-partitions. Best for large tables (multi-TB) with known filter patterns. Automatic Clustering maintains order over time
- When NOT to cluster: small tables, tables with frequent random writes, or tables without consistent filter predicates. Clustering incurs ongoing compute costs
- Materialized views: pre-computed results that Snowflake automatically refreshes. Enterprise+ feature. Best for expensive subqueries on large tables that change infrequently
- Streams: CDC objects that track DML changes (inserts, updates, deletes). Standard streams track all changes, append-only streams track inserts only
- Tasks: scheduled execution of SQL or stored procedures. Can form DAGs with parent-child trees. Support CRON and interval schedules. Can use warehouse or serverless compute
- Streams + Tasks pattern: use a stream to capture changes, a task to process them. The task checks SYSTEM$STREAM_HAS_DATA() before running
- Spilling to disk/remote storage: occurs when warehouse memory is insufficient. Visible in Query Profile. Fix by using a larger warehouse or reducing data volume
- Search optimization service: accelerates point lookup queries and queries with VARIANT fields. Background maintenance service (Enterprise+ feature)
Common Exam Traps
Result cache requires the EXACT same query text AND the same role. Even a minor change in whitespace or query text invalidates the cache
Warehouse cache is lost when the warehouse SUSPENDS. If auto-suspend is too aggressive, you lose the warm cache. Balance cost savings vs cache benefits
Clustering keys add ongoing maintenance cost. Do not cluster tables that are rarely queried or frequently truncated and reloaded
Materialized views have limitations: no joins, no UDFs, no nested views, and the base table must be a single table. The exam tests these limitations
SYSTEM$STREAM_HAS_DATA() is the function to check if a stream has unconsumed changes. Tasks should use this in their WHEN clause to avoid running on empty streams
Domain 510% of exam
Data Collaboration
The smallest domain but still worth 10 questions. Covers Snowflake's data sharing ecosystem: Secure Data Sharing between accounts, the Snowflake Marketplace, Data Exchange, data clean rooms, native apps, and replication. Understanding how Snowflake enables data collaboration without data movement is the key theme.
Key Topics
Secure Data SharingSnowflake MarketplaceData ExchangeData Clean RoomsNative Apps FrameworkDatabase ReplicationSecure ViewsReader Accounts
Must-Know Concepts
- Secure Data Sharing: share live, read-only data between Snowflake accounts without copying or moving data. Provider creates a share, consumer creates a database from it
- Shared data is READ-ONLY for consumers. Consumers cannot modify shared data. No storage cost for consumers — they only pay compute to query it
- Snowflake Marketplace: data providers publish listings (free or paid). Consumers discover and access data directly in their account. Powered by Secure Data Sharing under the hood
- Data Exchange: private, curated data sharing hub within an organization or with select partners. More controlled than public Marketplace
- Data clean rooms: secure environments for multi-party data analysis without exposing raw data. Privacy-preserving overlap analysis
- Native Apps Framework: build and distribute applications on Snowflake. Providers package code, data, and UI. Consumers install and run apps in their own accounts
- Reader accounts: Snowflake accounts created by a provider for consumers who do NOT have their own Snowflake account. The provider pays for compute
- Secure views: views that hide the underlying DDL and data from consumers. Used with data sharing to prevent consumers from seeing the query definition or base tables
- Database replication: replicate databases across Snowflake accounts and regions for DR, data locality, or sharing across cloud providers
- Shares can include tables, secure views, secure UDFs, and secure materialized views. Regular views CANNOT be shared
Common Exam Traps
Shared data is READ-ONLY. Consumers CANNOT modify it. If a question asks about consumers writing to shared data, the answer is no
Only SECURE views, SECURE UDFs, and SECURE materialized views can be added to shares. Regular (non-secure) views cannot be shared
Reader accounts are paid for by the PROVIDER, not the consumer. The provider covers compute costs for reader account usage
Data sharing works across accounts in the SAME cloud region by default. Cross-region sharing requires database replication
Snowflake Marketplace listings can be free or paid. Consumers do not always get data for free — providers can set pricing
Snowflake Concepts You Must Not Confuse
These pairs appear on nearly every exam. Learn the difference and you'll avoid the most common traps.
Top Mistakes to Avoid
Confusing Time Travel (user-accessible, configurable retention) with Fail-safe (Snowflake-only, fixed 7 days) — if you need to recover data yourself, it is Time Travel
Thinking transient tables have Fail-safe protection — they have 0-day Fail-safe, meaning no safety net after Time Travel expires
Confusing scaling UP (bigger warehouse for complex queries) with scaling OUT (more clusters for concurrent queries) — they solve different problems
Assuming zero-copy clones consume storage immediately — clones are free until data diverges through writes
Thinking result cache requires an active warehouse — result cache is served from the cloud services layer with zero compute cost
Confusing Snowpipe billing (serverless, per-GB credits since December 2025) with COPY INTO billing (warehouse credits) — they use different billing models
Not knowing that PUT and GET commands only work from SnowSQL or JDBC/ODBC, not from the web interface
Mixing up masking policies (column-level, hide values) with row access policies (row-level, filter entire rows)
Thinking shared data can be modified by consumers — Secure Data Sharing is always read-only for the consumer
Assuming all table types have the same data protection — permanent tables get 7-day Fail-safe, transient and temporary tables get 0-day Fail-safe
Confusing ACCOUNTADMIN capabilities with true superuser access — ACCOUNTADMIN still requires role hierarchy for object access
Forgetting that multi-cluster warehouses, 90-day Time Travel, and materialized views require Enterprise edition or higher
Exam-Ready Checklist
Can explain the three-layer architecture and how each layer scales independently
Know all virtual warehouse sizes, credit costs (doubling pattern), auto-suspend, and auto-resume behavior
Can distinguish between scaling up vs scaling out and when to use each approach
Understand all table types (permanent, transient, temporary, external) with their Time Travel and Fail-safe periods
Know the complete role hierarchy: ACCOUNTADMIN, SECURITYADMIN, USERADMIN, SYSADMIN, PUBLIC, and custom role best practices
Can explain all three caching layers and the conditions under which each is used or invalidated
Know stage types (user, table, named internal, external) and when to use each
Understand COPY INTO options: ON_ERROR, VALIDATION_MODE, PURGE, FILE_FORMAT, and transformation during load
Can explain Snowpipe vs COPY INTO: use cases, billing models, and when to choose each
Know streams and tasks: how they work together for CDC and automated ETL pipelines
Understand Secure Data Sharing: read-only, no data movement, consumer pays compute only
Can identify which features require which Snowflake edition (Standard vs Enterprise vs Business Critical)
Know COF-C03 new topics at the conceptual level: Cortex AI, Iceberg tables, Snowflake Notebooks, Git integration
Scored 80%+ on at least two full-length practice exams (750/1000 passing score requires strong preparation)
Recommended Resources
Free & Official Resources
Paid Courses & Practice Exams
These are recommended if you prefer a structured learning path. They can save time but are not required to pass.