Data Management
Data management spans model design, loading, quality, security and lifecycle. Start with the data model: pick the right object granularity, define relationships (lookup vs. master-detail, which controls ownership, roll-ups and deletion behavior), and create external IDs on objects you will upsert or integrate - an external ID is what makes idempotent loads possible.
Import tooling: the Data Import Wizard handles straightforward loads (commonly up to 50,000 records) with a guided UI and built-in duplicate checking; Data Loader handles larger volumes, all objects, scheduled extracts and command-line automation, with configurable batch sizes and bulk API options. Always load in dependency order, use upsert with external IDs for repeatable loads, and test in a sandbox first.
Quality: matching rules define what a duplicate looks like; duplicate rules define the action (allow with alert, or block) at the record level for users and API; validation rules prevent bad values; required fields and picklist governance keep values clean; periodic de-duplication and merge runs keep the org tidy. Field history tracking, field audit trails and the recycle bin support recovery and audit.
Lifecycle and governance: storage limits and archiving strategy, backup (scheduled exports or a backup vendor), retention policies, data privacy obligations such as GDPR/CCPA (right to be forgotten, consent fields), and documenting data owners and stewardship.
Traps: assuming duplicate rules block Data Loader by default; forgetting that master-detail relationships cascade delete; using text fields as keys instead of external IDs; loading data before users exist; assuming deleted records are gone (recycle bin retention).
Worked example: migrate 2 million legacy accounts/contacts - build external IDs, load accounts with upsert in Data Loader batches, then contacts referencing the account external ID, then verify counts and run duplicate reports before enabling the duplicate rule in blocking mode.