Study SnowPro DEA-C02 SQL ELT and Snowpark: key concepts, common traps, and exam decision cues.
On this page
Transformation questions usually become manageable once you identify whether the workload is best served by SQL-native ELT, managed refresh, or code-first logic.
Transformation-choice map
Requirement
Better first instinct
declarative derived table with managed refresh
dynamic table
straightforward set-based transformation
SQL-native ELT
code-first transformation workflow that benefits from programmatic APIs
Snowpark
reusable scalar logic inside SQL expressions
UDF
multi-step orchestration or procedural control
stored procedure
Start with the lightest abstraction that fits
If the requirement is…
Stronger first answer
simple set-based transform
SQL-native ELT
managed refresh of derived data
dynamic table
programmatic API-rich transform flow
Snowpark
inline reusable expression logic
UDF
explicit control flow with multiple steps
stored procedure
What the exam is really testing
If the stem says…
Strong reading
“managed refreshed transformation”
dynamic table is likely the first lane
“inline reusable logic in SQL”
UDF boundary matters
“multi-step control flow”
procedure or orchestration layer may be more appropriate
“developer workflow and programmability”
Snowpark may fit where SQL alone becomes awkward
Why over-programming is often a miss
DEA-C02 usually prefers the most Snowflake-native, lowest-friction abstraction that still solves the job. If SQL-native ELT or a dynamic table already fits cleanly, a heavier Snowpark or procedural answer can be weaker because it adds complexity without owning a different responsibility.
Common traps
Trap
Better rule
using Snowpark when SQL-native ELT already fits directly
choose the lightest abstraction that solves the problem
calling procedures where a scalar SQL expression is needed
procedures and UDFs have different output shapes
treating dynamic tables as a generic replacement for every pipeline object
managed refresh has a specific boundary
Scenario triage
Scenario clue
Stronger answer shape
“managed refreshed derived table”
dynamic table
“simple set-based transform”
SQL-native ELT
“inline reusable expression logic”
UDF
“developer-oriented programmable workflow”
Snowpark
“multi-step procedure with control flow”
stored procedure
Decision order that usually wins
Transformation questions in Snowflake usually reward choosing the lightest abstraction that fits. If a derived table needs managed refresh from an upstream query, think dynamic table. If ordinary SQL ELT already solves the transformation clearly, do not reach for Snowpark just because it is more programmable. If reusable inline expression logic is enough, think UDF rather than stored procedure. DEA-C02 usually penalizes code-first complexity without a real need.