Bulk Operations Template for Product Planning

Use synchronous processing when the batch is small (under 50 records) and the operation is fast (under 10 seconds total). Use asynchronous for everything else. The threshold should be based on user experience: if the operation takes long enough that a user might navigate away, it must be async with notifications. The [glossary entry on asynchronous processing](/glossary/prioritization) covers the architectural patterns.

Give the user three options at import time: Skip (keep existing, ignore duplicate), Update (merge new data into existing record), or Create duplicate (insert regardless). Default to Skip. Track which option was chosen in the import job metadata so the behavior can be audited. For CRM and contact data, also offer a "review duplicates" step that shows potential matches before processing.

There is no universal threshold. The right question is: can the user fix and retry the failures efficiently? A 2% failure rate on a 50,000-record import (1,000 failures) is fine if you provide a downloadable error CSV and a one-click retry. A 50% failure rate on a 100-record update suggests a systemic issue that should block the operation entirely.

Bulk operations must be scoped to a single organization. Cross-org bulk operations are an internal-only capability restricted to platform operators with elevated permissions. For multi-org setups, see the [Multi-Tenant Design Template](/templates/multi-tenant-design-template) for isolation patterns that apply to batch processing. ---