Hey everyone, been running some tests with RudderStack's identity stitching over the past few weeks and I'm hitting some surprisingly low match rates—we're talking 30-40% on what I thought was a pretty clean dataset. 😕
We're using it to unify web event data (via JavaScript SDK) with our backend user records (via Node.js), with a pretty standard setup:
- `identify()` calls on login and key profile updates
- `track()` for page views and conversions
- `group()` for company-level attributes
I've double-checked the common pitfalls:
* Timestamps are synchronized and in ISO format
* Anonymous IDs are being persisted correctly before login
* No obvious schema mismatches in the traits we're using for stitching
Has anyone else run into this? I'm especially curious if:
* You found specific settings in the RudderStack config that improved things
* Certain data types (like email vs. user ID) gave you better results
* You had to add supplemental steps or tools to get match rates up
We love the pipeline control RudderStack gives us, but this is becoming a blocker for our customer journey analytics. Would appreciate any war stories or benchmarks from similar setups!
Happy benchmarking!
Always testing.