Fin doesn't read your knowledge base the way a person does. It reads structure, tags, and outcomes. Get those wrong during migration, and resolution rate suffers from day one — even if every article technically "moved."
Across Intercom's customer base, Fin averages a 67% resolution rate. Top-performing teams hit 80% or higher. The gap between those two numbers usually comes down to migration quality, not the model.
How Fin actually reads your knowledge base
Fin doesn't treat your help center as a flat pile of articles. It uses hierarchy — Collection, Section, Article — to understand scope and context before it answers.
Most help desks only support a three-level structure. Intercom allows a simplified Collection > Article path with no Section in between. That's fine natively. It's a problem in migration: if your source platform's categories don't map cleanly onto that structure, a migration tool has to guess, usually by dropping orphaned articles into a default folder. That default bucket is exactly where Fin loses confidence about what an article covers.
Tags aren't metadata — they're how Fin decides what to trust
Tags matter just as much. Labels and categories from your source platform typically convert into Intercom tags on import. Tags aren't cosmetic — they're what let you filter content and, more importantly, exclude outdated or irrelevant articles from Fin entirely. If your source tagging is inconsistent, that inconsistency migrates too, and Fin ends up drawing from an inconsistently labeled library.
Audit your source hierarchy first. Flatten or reorganize any category without a clean three-level path, so nothing lands in a generic fallback folder.
Clean up your tagging taxonomy on the source side, not after everything's landed in Intercom. It's a much smaller job with a few hundred articles than with a few thousand already imported.
Resolved-only, recent, and highly rated — the same logic applies to tickets
If you're bringing historical tickets over to train Fin or power suggested replies, sequence matters. Open tickets have no outcome, so they add noise, not signal. Bulk-importing years of unfiltered history before any cleanup can drop baseline triage accuracy below 40%.
The fix isn't exclusion, it's order: migrate your most recent, highest-rated, resolved tickets first, confirm Fin's resolution rate holds, then bring in the rest of the archive as a second pass — it still has value for search and reporting, just not as Fin's initial training signal.
Formatting details that quietly break in translation
A few formatting details quietly break in translation between platforms:
- Inline images need to travel as actual image data, not broken links — verify this explicitly, since some export paths drop them silently.
- Multi-language content often transfers only the main language by default. If your help center is multilingual, confirm each language version is set up on the Intercom side before migration, or translations get orphaned from their source articles.
- Internal linking between articles doesn't always carry over automatically and may need to be requested as a specific step, or cross-references turn into dead links.
None of these break the migration outright. They just quietly limit what Fin can do with the content once it's there.
Test before you migrate all records
Run a small sample first — 20 or so articles and tickets — and check exactly how the hierarchy, tags, and formatting land before moving everything. Tools like Data Migration (check it on Intercom Marketplace) offer a free Demo Migration for exactly this reason. Catch a structural problem before your full archive is in and Fin's already live.
Get the structure right before the content moves, and Fin's performance reflects your actual knowledge base — not the parts of it that got flattened into a default folder along the way.
Already prepping a migration for Fin? What's been the hardest part to get right — hierarchy, tagging, or the historical ticket cleanup?