What happened
What this says
As at Fri 9 Oct 2026Imran Hussain reported duplicate customer records in search results with slightly different name spellings appearing after a Friday reindex. Olivia Grant escalated to the platform team and identified that the reindex broke duplicate detection, with the brain failing to merge records by site ID.
Problem: After a Friday reindex, multiple customer sites began appearing twice with name variations like Limited versus Ltd. Same site IDs in source data but not being merged by the brain. This confuses account managers running queries.
Handling: Imran provided specific examples. Olivia discovered changes to normalisation rules during the reindex broke duplicate detection. The brain was no longer merging records by site ID. Olivia escalated to platform and committed to testing a fix using customer examples before deployment.
Watch: Data quality issue affecting query accuracy. Several hundred records potentially affected.
Details
Updated 9 Oct 26Trail
- created · Imran Hussain 4 Sep 2025 16:09Hi, we're seeing duplicate customer records appearing in the brain again. We've got a couple of instances where the same customer site has been indexed twice with slightly different spelling in the company name. When our account managers query the system, they're getting both records back, which is causing confusion when they're trying to look up a specific site. I've checked our source data in the CRM and there's only one record per site there, so the duplication is happening somewhere in the indexing. Can you help us identify what's causing this?
- Public reply · Olivia Grant 4 Sep 2025 22:35Hi Imran, thanks for reporting this. Just so I've got this right, you're seeing the same customer site appearing twice in the brain's search results with slightly different company name spellings? Can you share a specific example, the customer name and how the two versions are different, and I can take a look. Do you know roughly when these duplicates first started appearing?
- Public reply · Imran Hussain 8 Sep 2025 14:41Olivia, sorry for the delay. I've got some examples now. We're seeing duplicate entries for several customer sites, where the system has indexed the same site twice with different name variations: using 'Limited' in one entry and 'Ltd' in another, or different hyphenation in place names. Same site ID in both entries in the source data. These only started showing up after we ran a full reindex on Friday. I've tested the CRM export and the data looks clean there, so I'm wondering if something changed in the ingestion process.
- Internal note · Olivia Grant 9 Sep 2025 14:34Looped in platform team to check the reindex logic. These look like name variants not being deduplicated on ingestion, the CRM connector is bringing them in separately and the brain isn't merging them by site ID. Check Friday's changes to the normalization rules.
- Public reply · Olivia Grant 9 Sep 2025 16:18Thanks for those examples, Imran. I've escalated this to our platform team. It looks like the reindex process on Friday may have affected how duplicate detection works, the brain isn't merging records by site ID the way it should be. I've asked them to check the normalization rules and get back to me. Once we've got a fix, I'll test it with your examples before we switch it on. Should have an update by end of the week. Thanks, Olivia
- Status change · 13 Sep 2025 23:59