Collaboration, Marketplace, and Cost Optimization: free practice questions

Aggregate policies (B) are a Snowflake governance feature specifically designed to enforce minimum aggregation thresholds—preventing consumers from querying data at the individual row level. A secure view containing only GROUP BY aggregations (D) structurally prevents row-level access since the view definition forces aggregation, and the SECURE keyword hides the view definition from consumers. Row access policies (A) can block specific rows but cannot prevent a consumer from scanning all rows that pass the filter, so they do not reliably enforce aggregation minimums. Dynamic data masking (C) obfuscates column values but does not prevent row-level enumeration. Tag-based masking (E) has the same limitation as masking policies. Network policies (F) restrict network access by IP, not by query type, and cannot differentiate aggregated from non-aggregated queries.

Correct: Secure views prevent consumers from inspecting the view definition and restrict access to only the columns the provider exposes; since the shared object is the secure view, consumers can never access the underlying PII columns. This is the standard Snowflake pattern for governed data sharing. — Distractor B is wrong because Terms of Service is a legal control, not a technical enforcement mechanism; a standard view also exposes its definition to consumers. — Distractor C is wrong because sharing base tables gives consumers access to all columns; expecting consumers to apply masking policies is not a provider-enforced control. — Distractor D is wrong because cross-cloud auto-fulfillment and network policies operate at the network/account level, not at the column level, and cannot hide specific columns from consumers.

Correct (A): Snowflake Marketplace listings can be edited by the provider, including changing the pricing model (free to paid) and updating the underlying share. The provider updates the listing's pricing configuration in the Provider Studio interface. The underlying share is modified with ALTER SHARE to include the new table. Existing consumers who had free access must go through the new request/approval flow under the paid model—access is not automatically grandfathered. Wrong (B): Deleting and recreating a listing is unnecessary and disruptive; it breaks existing consumer integrations, does not migrate consumers automatically, and wastes the listing's review/approval history. Snowflake does not automatically migrate consumers between separate listings. Wrong (C): CREATE SHARE … CLONE is not a valid Snowflake command—shares cannot be cloned using this syntax. Running two listings (free and paid) for the same product in parallel also creates consumer confusion and governance complexity. Wrong (D): Pricing model changes and listing updates are fully self-service through the Provider Studio in Snowsight. Snowflake Support involvement is not required for routine listing modifications.

Correct (A): By wrapping analysis logic in stored procedures and UDFs within the Native App, the provider can expose only aggregated or derived outputs. The consumer invokes the app's interface but is never granted direct SELECT on the underlying tables—the code acts as a privacy-preserving computation layer. Correct (C): The Native App is installed and runs inside the consumer's Snowflake account, meaning the consumer's compute is used and their data can be joined with provider data. However, the provider fully controls the app code, defining exactly what queries execute and what results surface—this separation of control is the core privacy mechanism. Wrong (B): Differential privacy noise injection is NOT a default or automatic feature of the Native App Framework. It can be manually implemented in custom logic, but it requires explicit coding; it is not built in. Wrong (D): Native Apps do not include automatic HIPAA compliance, PHI detection, or built-in data masking. HIPAA compliance is a contractual and architectural responsibility; Snowflake provides tools (masking policies, etc.) but not automated PHI classification. Wrong (E): Native App versioning allows providers to publish new versions, but consumers must upgrade (or auto-upgrade can be configured); the statement that consumers never need to re-install is not accurate for all update scenarios and is not the primary privacy-preservation mechanism.

Correct A: A low AUTO_SUSPEND value (e.g., 60 seconds) ensures warehouses suspend quickly when idle, directly reducing idle credit consumption — the primary waste identified. Correct C: AUTO_RESUME ensures the warehouse starts automatically when a query arrives, so users experience minimal friction even though the warehouse is frequently suspended. Together, these two settings maximize the suspend/resume cycle efficiency. — Distractor B is wrong because a high AUTO_SUSPEND value (30 minutes) keeps warehouses running during idle periods, which is the exact behavior causing the credit waste. — Distractor D is wrong because defaulting all workloads to X-LARGE wastes credits if the workloads don't require that scale; right-sizing should accompany consolidation. — Distractor E is wrong because the query result cache is free (no warehouse credits consumed) and disabling it would increase warehouse utilization and cost, not decrease it.