Sync Salesforce to Snowflake

Salesforce is where the pipeline lives. Snowflake is where the company measures it against everything else: product usage, billing, support, marketing spend. Getting Salesforce into the warehouse usually means a managed connector you cannot see into, a nightly job that is hours stale, or a bulk export someone schedules and forgets.

• Accounts, opportunities, and custom objects land in Snowflake as they change, through Salesforce Change Data Capture, not on a nightly cycle
• Every load is a run you can inspect, so a number that looks off in a board can be traced to the sync that wrote it
• Deleted and merged Salesforce records become soft deletes, so a historical report does not silently change
• Model output, a fit score or a health segment, is written back to Salesforce on a schedule so reps see the same number as the analyst

The load flow most teams turn on first.

Salesforce Opportunity changes into ANALYTICS.CRM.OPPORTUNITY, streamed

Change-driven through CDC, with a scheduled full reconcile so nothing drifts.

1. 1
   Trigger: a Live stream trigger with the Salesforce Change Data Capture source on /data/OpportunityChangeEvent, which starts a run for each change as it arrives.
2. 2
   Shape: map the CDC payload to the target column names and types using the table below, keeping the full change event in a _RAW column.
3. 3
   Merge: Execute Snowflake SQL with MERGE INTO ANALYTICS.CRM.OPPORTUNITY t USING (SELECT ...) s ON t.ID = s.ID, update on match, insert when not matched.
4. 4
   Deletes: a delete change event sets IS_DELETED = TRUE and DELETED_AT = CURRENT_TIMESTAMP() rather than removing the row.
5. 5
   Watermark: write the max SystemModstamp seen to a control table, for the reconcile to compare against.
6. 6
   Nightly reconcile: a scheduled flow runs a bounded SOQL query and merges the result, catching anything a gap in the CDC stream dropped.

The starting map for the Opportunity to ANALYTICS.CRM.OPPORTUNITY step.

• Core and custom objects. Account, Contact, Opportunity, Lead, Case, and any custom object land as typed tables plus a raw VARIANT column. (Salesforce to Snowflake)
• Field history and activity. Field history, tasks, and events are queried on a schedule into event tables for stage-velocity and activity modeling. (Salesforce to Snowflake)
• Scores and segments. A model table of account fit, health, or propensity is written back to Salesforce custom fields on a schedule. (Snowflake to Salesforce)
• Enrichment and hygiene. Firmographic fields, deduped account hierarchies, and do-not-contact flags computed in the warehouse update the matching Salesforce records. (Snowflake to Salesforce)

The same controls apply to this sync as to every other Neblex workload.

• Run history. Every load is a run with the change, the merge statement, and row counts. A failed run can be replayed from the step that failed.
• Approval on write-back. Reverse-ETL flows that change Salesforce data can include a Wait for form step, so a person reviews the changes before they are written.
• Scoped access. Salesforce uses a connected app and a permission set limited to the synced objects. Snowflake uses a role with grants on the target schema only.
• Environments. Build against a Salesforce sandbox and a Snowflake scratch schema, then promote the same flow to production.