Use this for last-mile review. DEA-C02 usually gets easier when you classify the Snowflake responsibility first instead of trying to solve everything with one familiar object.
DEA-C02 answer sequence
Use this when the stem mixes loading, streams, tasks, dynamic tables, sharing, or warehouse fit.
flowchart TD
S["Scenario"] --> L["Classify the Snowflake data-engineering lane"]
L --> O["Check object, stream, or task behavior"]
O --> W["Check warehouse or performance fit"]
W --> V["Verify with history, profile, or recovery evidence"]
Fast lane picker
| If the question is mainly about… |
Strongest first lane |
| loading files from stages or handling schema drift in loads |
chapter 1 |
| lower-latency ingest and pipe behavior |
chapter 1 or chapter 3 |
| ELT logic, dynamic tables, Snowpark, UDFs, or procedures |
chapter 2 |
| secure sharing, listings, replication, or failover |
chapter 2 |
| change capture, tasks, and near real-time orchestration |
chapter 3 |
| warehouse sizing, concurrency, or serverless fit |
chapter 4 |
| query history, query profile, and bottleneck evidence |
chapter 5 |
Snowflake object-boundary map
| Object or feature |
Better first reading |
| stage |
where data is referenced or staged for loading |
COPY INTO |
explicit table load from staged data |
| Snowpipe |
automated staged-file ingest |
| Snowpipe Streaming |
lower-latency streaming writes into Snowflake |
| stream |
change data capture on table or view changes |
| task |
scheduled or triggered work execution |
| dynamic table |
managed refresh of query-defined derived data |
| secure share |
live governed data access for another Snowflake account |
| replication or failover |
copy or continuity boundary across accounts or regions |
| warehouse |
compute resource for running queries and data-engineering workloads |
High-confusion pairs
| Pair |
Keep this distinction clear |
| stream vs task |
change capture versus execution scheduling |
| dynamic table vs stream-plus-task |
managed refresh versus explicit CDC plus orchestration pattern |
| Snowpipe vs Snowpipe Streaming |
automated staged-file ingest versus lower-latency streaming write path |
| secure sharing vs copying data |
live governed access versus physical duplication |
| replication vs failover |
copied continuity setup versus recovery or continuity action path |
| SQL ELT vs Snowpark logic |
SQL-native transformation versus code-first programmable path |
| Symptom |
Better first instinct |
| queueing or concurrency pain |
warehouse fit or multi-cluster question |
| slow query with no evidence yet |
query history or query profile first |
| repeated high compute cost |
workload fit and pipeline boundary question before resizing |
| poor pruning or scan behavior |
performance diagnosis, clustering, and filter-pattern question |
Last 15-minute recheck
| Recheck this |
Because the miss often hides here |
| stage vs pipe vs stream vs task |
Snowflake object ownership drives many answers |
| dynamic tables vs explicit orchestration |
managed refresh and explicit CDC are not the same |
| sharing vs replication |
delivery and continuity are different responsibilities |
| warehouse fit before performance tweaking |
wrong compute shape can make every later step look broken |
| history and profile before tuning |
diagnosis beats guessing |
One-sentence memory hooks
- If the question is about change capture, think stream before task.
- If the question is about scheduled execution, think task before stream.
- If the requirement is live governed delivery, think share before copying.
- If the requirement is lower-latency ingest, ask whether staged-file automation is enough or whether Snowpipe Streaming is the better fit.