How to Migrate From Confluence to Slab Safely
Direct answer: Migrating from Confluence to Slab works in three phases: audit and delete, import survivors as unpublished drafts, then rewrite titles and publish. Expect 3-6 weeks for a 500-page space.
Tedious migrations usually fail the same way. Someone clicks export, imports 1,800 pages, takes a screenshot, and then the new wiki sits there full of 2019 meeting notes while everyone keeps using Confluence because that’s where the links point.
We moved an 1,100-page Confluence instance to Slab in 5 weeks and ended with 240 posts. The deletions were more valuable than the migration.
Why should you delete before you import?
Because Confluence’s page count is not your knowledge count. In a mature instance, a large share of pages are superseded, duplicate, or were never read after the week they were written.
Confluence’s built-in analytics, available on Premium and above, give you views and last-updated. If you don’t have it, there’s a cheaper proxy: search Confluence for lastmodified: operators or export a CSV of page metadata through the REST API.
Our audit: 1,100 pages down to 410 that had been read in the past year, and 190 that had been read in the past 90 days and weren’t clearly superseded.
How do you decide what survives?
Three questions per page. Answer them once, in a spreadsheet, and don’t revisit.
- Was it read in the last 12 months? No → delete.
- Does a newer page supersede it? Yes → delete, and log the supersession so nobody recreates it.
- Would a new hire need it in month one? No → archive outside the wiki, usually a Drive folder nobody links to.
That leaves you with roughly 20-25% of pages in a typical instance. If your number comes out at 70%, you skipped question two.
How does the actual export work?
Slab imports Confluence spaces through its migration tool, which handles space export files and the Confluence Cloud API.
- Export the space as HTML from Confluence’s Space Tools, or grant Slab API access for Cloud instances.
- Import into a staging topic marked private. Never import straight into your live structure.
- Check the attachments. Images usually transfer; embedded macros and Jira filters don’t, and those pages will look broken.
- Verify formatting on tables. Confluence tables with merged cells frequently flatten on import.
Pages with heavy macro usage — especially Jira query macros and Roadmap Planner — will need to be rebuilt or dropped. Count them before you start; that count tells you whether 5 weeks is realistic.
What about permissions and history?
Page history does not migrate. Slab imports the current version and its content, and the version history stays in Confluence. If you’re in any regulated context, export PDFs of the final versions of pages you’ll need to defend later.
Permissions map imperfectly. Confluence space permissions map to Slab workspace roles, but page-level restrictions don’t carry over. Expect to lose a small number of page-level restrictions and to have to re-create the important ones manually. Audit those before decommissioning.
How do you handle the broken links problem?
This is the part that decides whether the migration sticks. Confluence URLs are everywhere: Slack history, Jira tickets, old emails, onboarding docs.
- Keep Confluence read-only for 90 days minimum. Decommissioning early means every old link fails and people revert.
- Don’t aim for 100% redirect coverage. Slab doesn’t offer a bulk redirect import from Confluence. Prioritize the top 50 pages by inbound links.
- Put a note in your Confluence space home pointing to Slab, and pin it.
Old Slack links will break no matter what you do. Accept it.
What timeline should you plan for?
| Phase | Duration (500-page space) | Owner |
|---|---|---|
| Audit and delete list | 1 week | Wiki owner + one person per team |
| Export and staging import | 3 days | Wiki owner |
| Title rewrite and topic assignment | 2-3 weeks | Team contributors |
| Publish and announce | 2 days | Comms |
| Read-only Confluence window | 90 days | IT |
The rewrite phase is where projects die, because it looks like content work and content work gets deprioritized. Give each team a quota of pages and a deadline, and make one person accountable.
What are the most common regrets?
Importing too much. Everyone who did a full import regrets it. Slab search quality drops immediately and there’s no quick fix.
Not rewriting titles. An imported page called “Q3 Process Update (FINAL v2)” is unfindable and unchanged after migration. Title rewrites are 60% of the value.
Decommissioning Confluence too early. Give it a quarter. The old system costs you a few dollars a seat; broken links cost you trust in the new one.
Key Takeaways
- Delete 70-80% of Confluence pages before importing; deletion is the real value.
- Page history and page-level permissions don’t migrate — export PDFs if you need an audit trail.
- Keep Confluence read-only for at least 90 days to protect existing links.
- Budget 2-3 weeks for title rewrites; skipping it produces an unfindable wiki.
- Rebuild pages relying on Jira macros manually; they won’t survive import.
