If you are searching for a SnowPro Core certification study plan right now, there is a good chance the guide you land on is describing an exam that no longer exists. Snowflake released COF-C03 on 16 February 2026 and retired the English COF-C02 exam on 14 May 2026. The six-domain blueprint that almost every ranking study guide, YouTube playlist and dumps site still teaches has been replaced by a five-domain one, with entirely new topic areas bolted on.
That is the short answer to the question most candidates are actually asking: yes, the SnowPro Core certification is still absolutely worth sitting, and no, the study material you bookmarked last year is not going to cover it. This guide gives you the current COF-C03 blueprint with real domain weights, the specific new topics that catch people out, and a 6-week plan that allocates your time in proportion to what the exam actually tests.
What Changed in COF-C03 (And Why Most Study Guides Are Out of Date)
Snowflake did not simply refresh a few questions. It restructured the exam around how the platform is used now, which means AI and open table formats are no longer footnotes.
The headline changes:
- Six domains became five. Data Pipelines was dissolved and its content redistributed, mostly into loading and connectivity.
- Architecture got much heavier, rising from roughly 25% to about 31% of the exam. That single domain is now nearly a third of your score.
- Whole new topic areas appeared: Snowflake Cortex AI, Snowpipe Streaming, Dynamic Tables, Apache Iceberg tables, Data Clean Rooms, Trust Center, Git integration and Data Lineage.
- Security was folded into governance. The old standalone Account Access and Security domain is now Account Management and Data Governance.
Exam Tip: Architecture is heavier in COF-C03 precisely because Iceberg, Cortex, Snowpark and Streamlit are all new architectural primitives. If you treat architecture as a warm-up topic the way COF-C02 candidates did, you will lose the domain that carries the most marks.
The translated versions of COF-C03 landed on 15 April 2026, and the COF-C02 translations retired on 31 July 2026. There is no transition window left. Whatever region or language you sit in, COF-C03 is the exam.
SnowPro Core Certification Exam Facts at a Glance
These are the numbers to plan around. The SnowPro Core exam has 100 questions and a 115 minute time limit, which works out at roughly 69 seconds per question.
| Detail | COF-C03 |
|---|---|
| Questions | 100 |
| Time limit | 115 minutes |
| Passing score | 750 out of 1000 (scaled) |
| Question formats | Multiple choice and multiple select |
| Registration fee | $175 per attempt |
| Delivery | Pearson VUE, test centre or online proctored |
| Prerequisites | None (6+ months of Snowflake experience recommended) |
| Validity | 2 years |
| Retake cost | Full $175 again, no free second attempt |
Two of those deserve emphasis. The passing score is scaled, so 750 out of 1000 is not the same as answering 75% of questions correctly, and Snowflake does not publish the raw conversion. And there is no free retake on SnowPro Core, so every attempt is a real $175. Both facts argue for over-preparing rather than booking optimistically.
If you are sitting the exam from home, the check-in process is standard Pearson VUE, and it trips more people up than the content does. Our guide to OnVUE online proctored exams covers the room scan and the behaviour that gets sessions terminated.
The COF-C03 Five-Domain Blueprint
Here is the current blueprint next to the one it replaced, so you can see exactly where your study time needs to move.
| COF-C02 domain (retired) | Weight | COF-C03 domain (current) | Weight |
|---|---|---|---|
| Snowflake AI Data Cloud Capabilities and Architecture | 25% | Snowflake AI Data Cloud Features and Architecture | 31% |
| Account Access and Security | 20% | Account Management and Data Governance | 20% |
| Performance Concepts | 15% | Performance Optimization, Querying and Transformation | 21% |
| Data Loading and Transformation | 20% | Data Loading, Unloading and Connectivity | 18% |
| Data Protection and Data Sharing | 10% | Data Collaboration | 10% |
| Data Pipelines | 10% | (dissolved and redistributed) | - |
The practical read on this table: architecture plus performance now account for 52% of the exam between them. If you get those two domains solid, you are more than halfway to a pass before you touch governance or loading.
Domain 1: Features and Architecture (31%)
Multi-cluster shared data architecture, the three layers (storage, compute, cloud services), micro-partitions and clustering, virtual warehouse sizing and scaling policies, caching layers, editions and their feature gates, and the newer primitives: Snowpark, Streamlit, Iceberg tables and Cortex AI.
Domain 2: Performance Optimization, Querying and Transformation (21%)
Query profile interpretation, pruning and clustering keys, search optimisation service, materialised views, query acceleration, result cache versus warehouse cache, and the transformation work that moved here from the old Data Pipelines domain: streams, tasks and Dynamic Tables.
Domain 3: Account Management and Data Governance (20%)
Role-based access control and the system-defined roles, secure views, masking and row access policies, object tagging, network policies, authentication methods, Trust Center, Data Lineage and access history.
Domain 4: Data Loading, Unloading and Connectivity (18%)
Stages, file formats, COPY INTO in both directions, Snowpipe and Snowpipe Streaming, semi-structured data handling with VARIANT and FLATTEN, external tables, drivers and connectors, and Git integration.
Domain 5: Data Collaboration (10%)
Secure Data Sharing, reader accounts, the Snowflake Marketplace, listings, and Data Clean Rooms.
The 6-Week SnowPro Core Study Plan
This plan assumes 8 to 10 hours a week and access to a Snowflake trial account. Hands-on time is not optional here. A large share of COF-C03 questions describe a scenario and ask what actually happens, and you cannot reliably answer those from reading alone.
Week 1: Architecture Foundations
Cover the three-layer architecture, micro-partitions, clustering, and virtual warehouses end to end. Spin up warehouses at different sizes and watch the credit consumption. Run the same query twice and identify which cache served the second run.
Deliverable by Sunday: you can explain, without notes, what happens between submitting a query and receiving results, and which layer does what.
Week 2: The Rest of Architecture, Including the New Primitives
Editions and feature gating, Snowpark, Streamlit, and the two big newcomers. Create an Iceberg table and a native table holding the same data, and write down when you would choose each. Call at least three Cortex AI functions in SQL.
Exam Tip: For Cortex AI, know the SQL-callable functions (COMPLETE, EXTRACT_ANSWER, SENTIMENT, SUMMARIZE, TRANSLATE), what each returns, and the fact that they run inside Snowflake with no external API call. That last point is a favourite distractor.
Week 3: Performance Optimization and Transformation
Live in the Query Profile. Build a deliberately badly clustered table, query it, read the pruning statistics, then fix it and compare. Cover search optimisation service, materialised views and query acceleration, then move on to streams, tasks and Dynamic Tables.
Dynamic Tables are the topic most likely to appear in a form you have not seen if you studied for COF-C02. Understand how they differ from materialised views and from a stream plus task pattern.
Week 4: Account Management and Data Governance
Build the RBAC hierarchy by hand: create custom roles, grant them, and trace an inherited privilege from SYSADMIN downwards. Apply a masking policy and a row access policy and confirm the behaviour changes per role. Then cover tagging, network policies, Trust Center and Data Lineage.
Week 5: Loading, Unloading, Connectivity and Collaboration
Load from an internal stage and an external stage. Use COPY INTO in both directions. Ingest semi-structured JSON and query it with FLATTEN. Set up Snowpipe, then read up on how Snowpipe Streaming differs from it, because the comparison is prime question material.
Finish the week on Data Collaboration: shares, reader accounts, Marketplace listings and Data Clean Rooms. It is only 10% of the exam, so cap it at a day and a half.
Week 6: Practice, Diagnose and Fix
Stop learning new material. Sit full 100-question timed practice exams, then work through every question you got wrong and write down why. Retake until you are consistently clearing your target with time to spare.
Book the exam for the end of this week. An open-ended study period is how the SnowPro Core certification quietly turns into a six-month project. Our breakdown of why most people fail certification exams covers that failure mode in more detail.
The New COF-C03 Topics That Catch People Out
If you are short on time, these are the highest-yield areas precisely because the older study material does not cover them.
- Cortex AI functions. Know them by name and by what they return, and know they execute inside Snowflake.
- Apache Iceberg tables. The exam cares about interoperability. Iceberg lets other engines such as Spark and Trino read and write the same data. Know when that matters and when a native table is the better call.
- Dynamic Tables. Declarative, target-lag-driven refresh, distinct from materialised views.
- Snowpipe Streaming. Row-level ingestion via an API, not file-based like classic Snowpipe.
- Data Clean Rooms. Collaboration on sensitive data without exposing the underlying rows.
- Trust Center. Security posture scanning and recommendations at account level.
- Git integration. Repository objects inside Snowflake for version-controlled code.
- Data Lineage. Tracing how data moved and transformed across objects.
Every one of these is new to the blueprint. Any guide that does not mention them was written for the retired exam.
How Hard Is the SnowPro Core Certification, Really?
SnowPro Core is a genuine associate-level exam, not a fundamentals badge. The difficulty is not conceptual depth so much as breadth plus pace. You have 115 minutes for 100 questions, several of which are multiple-select with no partial credit, and the scenarios are worded to punish surface-level familiarity.
Snowflake recommends six or more months of hands-on experience. If you have that, six weeks of structured study is realistic. If you are coming in cold from another platform, budget eight to ten weeks and spend the extra time in the console rather than in slides.
The most common failure pattern is candidates who can define every feature but cannot predict behaviour. The exam rarely asks what a clustering key is. It asks what happens to pruning when you cluster on a high-cardinality column.
Cost, Recertification and What Happens After You Pass
SnowPro Core costs $175 per attempt and the certification is valid for two years. There are no free retakes, so a failed attempt costs another $175 in full.
Recertification changed. Snowflake retired the standalone COF-R02 recertification voucher and moved renewal into its Continuing Education programme. The two realistic paths before your two-year expiry are completing an eligible Snowflake instructor-led training course, or earning an equivalent or higher SnowPro certification such as a SnowPro Advanced or a Specialty exam. You can also retake the full SnowPro Core up to six months before expiry to reset the clock.
SnowPro Core is the gateway credential. Everything above it, the Advanced tier and the Specialty exams, assumes you hold it. If your goal is a data engineering role, pairing it with a platform-neutral or adjacent credential is the stronger play. Our Databricks Certified Data Engineer Associate study plan and the Microsoft DP-700 Fabric Data Engineer plan both pair well with SnowPro Core on a CV.
Frequently Asked Questions
Is the SnowPro Core certification worth it?
Yes, if you work with or want to work with Snowflake. It is the prerequisite-in-practice for every Advanced and Specialty SnowPro exam, it is widely recognised in data engineering and analytics job adverts, and at $175 it is one of the cheaper vendor certifications relative to the platform's market position. It is less useful as a first-ever IT certification if you have no data background.
How much does the SnowPro Core certification cost?
The exam is $175 per attempt. There are no free retakes, so a second attempt is another $175. Recertification through the Continuing Education programme carries its own costs, either an instructor-led training course or another SnowPro exam such as a Specialty at $225 or an Advanced at $375.
How long is the SnowPro Core certification exam?
The exam is 115 minutes long and contains 100 questions, giving you roughly 69 seconds per question. The certification itself remains valid for two years from the date you pass.
What is the pass mark for the SnowPro Core certification?
You need a scaled score of 750 out of 1000 to pass. Because the score is scaled rather than a raw percentage, Snowflake does not publish exactly how many questions that equates to, so aim comfortably above 75% in practice exams rather than treating it as the target.
Is SnowPro Core harder than other data certifications?
It sits at a similar level to the Databricks Certified Data Engineer Associate and Microsoft DP-700, and it is noticeably harder than fundamentals-level exams such as AZ-900 or DP-900. The pace is what separates it. Multiple-select questions with no partial credit and roughly a minute per question leave little room for deliberation.
Ready to Start Practising?
Reading about micro-partitions will not tell you whether you can answer a scenario question about them under time pressure. Practice questions will.
CertCrush gives you exam-style practice with full explanations for every answer, so you find your weak domains in week two rather than on exam day. Browse the full course catalogue to see what is available, check the pricing options, or create a free account and start working through questions today.
Six weeks, five domains, one exam. Book the date, then work backwards.