Export CRM Data: Best Practices for Safe, Compliant, Migration-Ready Files
A CRM export looks like a one-click button. It is actually the moment your relational data gets flattened into rows, your encoding can silently break, and your PII leaves a controlled system. Here is how to export cleanly the first time, whether you are prepping a migration or feeding parallel reporting.
What this covers8
- Decide what the export is for before you click Export
- Export object by object, in dependency order
- Keep the IDs, or you will rebuild relationships by hand
- Get the CSV format right: encoding, delimiters, and leading zeros
- Respect platform export limits and windows
- Treat the export as a PII event, because it is
- Validate before you trust the file
- Where Pineido fits
Decide what the export is for before you click Export
The right export depends on the job. A pre-migration extract needs every object, every relationship link, and a frozen point in time. A parallel-reporting feed needs a clean, repeatable subset you can pull on a schedule without dragging along attachments and dead fields. Treat them differently. The fastest way to create chaos is letting reps keep editing the old system while you export for a migration, so set a clear data cut-off date: a moment after which no new records or edits land in the legacy CRM, so your files are not stale before they load. For reporting feeds, the opposite is true. You want incremental pulls, so capture a last-modified timestamp you can filter on next time.
Export object by object, in dependency order
A CSV is a flat grid. Your CRM is a graph of accounts linked to contacts linked to deals linked to activities. Export everything into one mega-file and you destroy those links. Instead, export one file per object and follow the standard sequence: users and roles, then accounts and companies, then contacts and leads, then deals and opportunities, then activities and notes, then attachments, then any custom objects last. Custom objects usually reference several standard ones, so they have to come after their parents exist. This order is not bureaucracy. A single misordered import on the way back in is how deals end up orphaned, sitting in no pipeline and invisible to your sales team.
Keep the IDs, or you will rebuild relationships by hand
This is the single most overlooked step. Every exported row should carry the source-system record ID, plus the IDs of its parents (the account ID on a contact, the contact and account ID on a deal). When you re-import, each parent lands with a new ID in the target system, and you record a mapping from old ID to new ID, then feed that map into the next object's load so children attach correctly. Skip the IDs and you are reconstructing associations manually across tens of thousands of records, which is the most common reason a migration that 'completed' turns out to be operationally broken. Validating relationships, not just row counts, is the whole game.
Get operations insights weekly
Join service-business operators who get one actionable insight on ops, automation, and growth every week.
No spam. Unsubscribe anytime.
You’re subscribed! Look for our next issue in your inbox.
That did not go through. Please try again, or email [email protected].
Get the CSV format right: encoding, delimiters, and leading zeros
Most export damage is silent and lives in the file format. Export as UTF-8 so names with diacritics, Cyrillic, or emoji survive instead of turning into garbled symbols. Be aware of a classic trap: Salesforce report CSVs are UTF-8 without a byte-order mark, and Excel on Windows opens them assuming Windows-1252, so characters break the moment someone double-clicks the file. Open it through Data, then From Text/CSV, and set the encoding to 65001 (UTF-8) instead. HubSpot likewise requires UTF-8 when files contain foreign-language characters, and a UTF-16 or Windows-1252 file is a common mapping failure. Watch for other quiet corruptions too: Excel strips leading zeros from ZIP codes, phone numbers, and external IDs unless those columns are treated as text, free-text fields get truncated at character limits, and mixed date locales (MM/DD versus DD/MM) and currencies without ISO codes break validation downstream. If a value is an identifier, it is text, not a number.
Respect platform export limits and windows
Native exports have rules that bite at scale. Salesforce keeps an export available for only 48 hours, and starting a new manual or scheduled export immediately deletes the previous one, even if it is still inside that window, because the org retains a single export set at a time. Download promptly. Salesforce packages exports into ZIP files of roughly 512 MB each and auto-splits large exports across multiple ZIPs. For loads back in over a couple thousand records, use the Bulk API rather than the UI. HubSpot's import and concurrency limits are tiered by plan: free and basic accounts allow files up to 20 MB and up to 500,000 rows per day, while Starter, Professional, and Enterprise allow files up to 512 MB. Plan your batches around the tier you actually have, not the one in the marketing page.
Treat the export as a PII event, because it is
The moment data leaves the CRM, it leaves the system's access controls and audit trail. That is a compliance event, not just an IT task. Apply data minimization: export only the fields and records the job needs, since every record you carry forward is data you are actively processing under GDPR with all the obligations that implies. Encrypt the files in transit and at rest, and never leave an unencrypted CSV sitting in shared cloud storage or an email attachment. Keep an audit trail of who exported what and when. And mind retention and deletion rights. If you are exporting years of inactive contacts 'just in case,' you are likely working against your own retention policy, and CCPA gives California residents a private right of action specifically when unencrypted personal information is breached. For test runs, mask or pseudonymize real PII rather than handing production data to a sandbox.
Validate before you trust the file
A finished export is a hypothesis until you check it. Start with record-count reconciliation: total accounts, contacts, deals, and activities in the file versus the live system, per object type, and investigate every discrepancy rather than rounding it away. Then go deeper than counts, because counts alone hide the worst failures. Spot-check 50 to 100 records per object for field-level accuracy, including custom fields, and verify parent-child links actually resolve. Deduplicate before you map, not after, since duplicates carry forward as duplicate associations and consent records that are far more painful to untangle on the far side, and real duplicates are fuzzy (abbreviated names, transposed email domains, varying company suffixes), so exact-match dedupe is not enough. If this export feeds a migration, run a small test load of a few hundred records per object first. Catching a broken mapping costs 200 rows, not 50,000.
Where Pineido fits
Most teams can run a clean export with the checklist above. Where it gets hard is custom objects, fuzzy deduplication at scale, and proving the re-import preserved relationships rather than just counts. That is the work in our CRM and migration sprints: a fixed 30-day scope at $8,500, with an export plan, a documented field-and-ID map, a sandbox dry run, and a reconciliation you can sign off on. If the export is for reporting rather than a move, our reporting-trust sprint builds the repeatable, encoding-safe pull so your dashboards stop disagreeing with the CRM.
Ready to ship?
Tell us where you are stuck. We will review it and send a focused recommendation within one business day.
Frequently asked questions
Should I export to CSV or use the API?
CSV is fine for small-to-mid datasets, one-time pulls, and budget-conscious teams. It is accessible and works everywhere. For large instances, recurring delta pulls, or anything over a couple thousand records on re-import, use the platform's bulk or batch API instead, since it handles pagination, rate limits, and large volumes that the UI export will choke on. Many teams do both: a native UI export for the bulk historical load, then API delta updates for the records that changed after.
Why do special characters break after I export CRM data?
The file is almost certainly not being read as UTF-8. CRMs store data as UTF-8 internally, but Salesforce report CSVs ship without a byte-order mark, so Excel on Windows assumes Windows-1252 and garbles accented or non-Latin characters the instant you open the file. Open the CSV through Excel's Data, then From Text/CSV path and set encoding to 65001 (UTF-8). When saving from Excel, be aware it defaults to machine ANSI encoding, so use a tool or setting that forces UTF-8 on write.
Do I really need to keep the record IDs in the export?
Yes, for any export that will be re-imported. The source-system ID on each row, plus the parent IDs, is what lets you map old records to their new IDs in the target and reattach contacts to accounts and deals to both. Without that mapping you end up rebuilding relationships by hand, which is the most common reason a migration looks done but is operationally broken with orphaned deals and detached activities.
What are the compliance risks of exporting CRM data?
An export pulls PII out of the CRM's access controls and audit trail, so it is a privacy event. Export only the fields and records you need (data minimization), encrypt the files in transit and at rest, never leave unencrypted CSVs in shared storage or email, and log who exported what and when. Under GDPR every record you carry forward is data you are processing, and under CCPA California residents have a private right of action specifically when unencrypted personal information is breached, so encryption is not optional.
How do I know the export is complete and correct?
Reconcile record counts per object type between the file and the live system, and chase every discrepancy. Counts alone are not enough, so spot-check 50 to 100 records per object for field accuracy including custom fields, and confirm parent-child links resolve. If the export feeds a migration, run a small test load of a few hundred records per object and validate relationships, not just rows, before committing the full load.
What is a data cut-off date and why does it matter?
A cut-off date is the moment you freeze edits and new records in the source CRM so the export reflects a single consistent point in time. Without it, reps keep adding and changing records while you export, and your files are outdated before they finish loading. For migrations, freeze the source and keep it read-only until the new system is verified. For recurring reporting feeds you do the opposite and pull incrementally using a last-modified timestamp.