One record per member, across four systems at once
How a health insurer merged thousands of duplicate records spread across its document management system, its management software, its support tool and its database.
The client
A health insurer
The client manages several hundred thousand members. Their information is spread across four tools: a document management system, Cegedim management software, the Zendesk support platform and an Azure SQL database.
Every day, case handlers create and edit records in these tools. The client asked to remain anonymous.
The project
Thousands of duplicates built up over the years
When a case handler cannot find a member, they create a new record: a typo, a double-barrelled name, an address written differently, and there is a duplicate. Over the years, thousands of people existed several times.
Merging two records in one tool was not enough: the other three kept the old one. And each duplicate held personal data for no reason, which is a problem under GDPR.
Records re-created instead of found.
Each with its own copy of people, with no link between them.
No way to see a member's full history.
Personal data kept twice.
What we put in place
A merge process across all four systems
The client's teams spot duplicates and list them. Our script takes this list and applies each merge across the four tools: it updates references, groups histories and removes surplus records.
A list as input
Case handlers decide what counts as a duplicate. The process applies the merges they have approved, and only those.
A merge applied everywhere
For each duplicate, the script connects to the four tools and applies the same merge. The record kept inherits the full history.
A script connected to the four sources, running on the client's cloud infrastructure.
A report for every operation
Every merge is logged: original records, record kept, systems changed. The client can justify each operation in an audit.
How we got there
Start small, check, then scale up
We first ran the process on a few dozen duplicates chosen with the case handlers. Each result was checked by hand in all four tools before increasing the volumes.
The merge log served as evidence at every step: the client knew exactly what had been changed, where and when.
Timeline
Three months from scoping to production
The four tools, their access and the merge rules.
The connections and the merge engine.
A few dozen duplicates, checked by hand.
Large-scale merges, with their log.
Results
A cleaner database, and merges you can justify
Duplicate data in your tools?
A first 30-minute conversation, free and with no commitment.