6.5 Choosing the Right Sharing Method and Readiness Drills

The Decision Framework

Nearly every Data Collaboration question reduces to which collaboration object fits this scenario? Read the stem for three cues — the region/cloud, the audience, and the recipient type — then choose:

All of these are built on Secure Data Sharing, so the invariants hold everywhere: no data is copied, shared objects are read-only, the provider pays storage, and live data is reflected instantly. When two answers seem plausible, eliminate the one that violates an invariant (for example, any answer implying the consumer is billed for storage, or that a direct share crosses regions, is wrong).

Worked Scenarios

Scenario A. A retailer wants to give its analytics partner — a Snowflake customer in the same AWS region — live access to a curated sales view. Answer: create a secure view, add it to a share, and add the partner's account. A direct share is correct because the partner is a known account in the same region.

Scenario B. A data vendor wants to sell a dataset to hundreds of unknown customers across multiple clouds. Answer: publish a paid Marketplace listing with auto-fulfillment, which monetizes the data and replicates it to consumer regions automatically.

Scenario C. A marketing agency must share dashboards with a small client that is not a Snowflake customer. Answer: create a reader account; the agency manages it and pays its compute, and the client consumes data without buying Snowflake.

Scenario D. A bank and an advertiser want to measure audience overlap without revealing customer identities. Answer: a Data Clean Room, which permits approved aggregate queries over combined data while hiding raw rows.

Practicing these mappings out loud — cue, mechanism, why the alternatives fail — converts recognition into reliable recall.

Readiness Markers and Common Traps

Use these markers to judge whether the domain is exam-ready:

• Recall: You can state the Secure Data Sharing invariants (no copy, read-only, provider pays storage, live data) without notes.
• Recognition: You can identify the right object from a scenario even when the stem never says "share" or "listing."
• Application: You can map region + audience + recipient type to direct share, listing, Data Exchange, or reader account.
• Distractor control: You can explain why a tempting answer fails — e.g., it bills the consumer for storage, or claims a direct share reaches another cloud.

High-frequency traps

1. Assuming shared data is copied into the consumer — it is not.
2. Billing storage to the consumer — storage stays with the provider.
3. Using a direct share across regions — that needs replication or a listing.
4. Sharing a standard (non-secure) view — only secure views/UDFs are shareable.
5. Thinking a reader account can consume from multiple providers — it serves only its creator.

When mixed practice stays stable after a day away and you can defend every distractor, the Data Collaboration domain is ready. Remember that the same five mechanisms reappear under many disguises, so train yourself to translate the surface story into region, audience, and recipient type before you even read the options.

Drills That Build Real Recall

Passive rereading creates a false sense of mastery in this domain because the terms are intuitive. Convert study into active retrieval with these drills:

1. Invariant recitation. Without looking, write the four Secure Data Sharing invariants: no data copy, read-only for the consumer, provider pays storage, data is always live. Any scenario answer that breaks one of these is wrong.
2. Mechanism mapping. Given a one-line scenario, name the object (direct share, free/personalized/paid listing, Data Exchange, reader account, or clean room) and the single cue that decided it.
3. Billing flashcards. For each pairing — standard consumer, reader account, paid listing — state who pays storage and who pays compute. The reader-account flip (provider pays compute) is the one most often missed.
4. Secure-object check. For a share that must include a view or function, restate that only secure views and secure UDFs are shareable, and why (definition hiding + leak-safe optimizations).
5. Cross-region trigger. Whenever a scenario names two regions or clouds, immediately reach for replication or a listing, never a plain direct share.

Self-test ledger

Keep a two-column ledger: on the left, the cue from a missed question ("different cloud," "no Snowflake account," "hide tax IDs"); on the right, the correct object or control ("listing," "reader account," "secure view + masking policy"). Review the ledger before test day. If you can answer mixed questions cold after a one-day break, explain why each distractor fails, and recite the invariants without notes, the Data Collaboration domain is genuinely ready rather than merely familiar.