CLIENT STORY
A top-tier European bank serving millions of retail and commercial customers had grown the way most large banks do: one system at a time. A CRM tracked relationship-manager conversations, a core banking platform held transactions, a servicing system logged account changes, and a support desk ran calls and tickets. Leadership wanted that data feeding AI-driven scoring, personalization, and prediction, but every one of those initiatives needed something the bank didn’t have yet: a single Customer 360 profile per client.
| Service: | Data Engineering |
|---|---|
| Industry: | Finance |
| Region: | EU |
Read summarized version with
We had four versions of the same customer and no way to tell which one was right. We needed a team to own that problem end to end, not another report comparing the four. — Director of Customer Data
Our relationship managers stopped asking which system has the real picture. There's one profile now, and it's the one everyone trusts. — Head of Digital Banking
The Challenge: No Single Source of Truth
The bank stored customer data across multiple disconnected systems: a CRM for relationship managers, a core banking platform for transactions, a servicing system for account changes, and a support desk for calls and tickets. None of those systems were built to reconcile with one another, or to agree on which records belonged to the same customer, even when the difference was obvious to a human reading both screens.
As a result:
- The same customer could appear under multiple identifiers across systems.
- Customer information was inconsistent across departments.
- Manual reconciliation became a routine operational task.
- Customer analytics lacked a trusted single source of truth.
- Planned AI initiatives were blocked by poor entity resolution.
Before introducing predictive models, personalization engines, or customer scoring, the bank needed a Customer 360 layer.
Building the Customer 360 Layer Across the Bank's Core Systems
GroupBWT designed and implemented the bank’s Customer 360 layer, an identity resolution system that continuously reconciles records across core systems.
Deterministic and probabilistic matching. The layer matches customer records two ways: deterministic rules that check exact IDs or account numbers, and probabilistic scoring that weighs name, address, and behavioral similarity. That combination avoids choosing between a rigid rules engine and a model that’s usually right but never certain.
Household-level relationship mapping. Beyond matching individual customers, the system maps relationships between accounts: a joint mortgage, a shared address, an authorized signer, into a household view. This allows relationship managers and risk teams to analyze customer relationships at the household level rather than as isolated accounts.
Confidence scoring for ambiguous matches. Every match carries a confidence score; high-confidence matches merge automatically, and low-confidence ones route to a data steward to confirm, so higher accuracy never comes at the cost of quietly merging two different customers.
Built for downstream AI and analytics. The profile was designed to support future use cases including next-best-action, segmentation, fraud detection, and personalization, without a new integration project per initiative. GroupBWT delivered the first of these: a churn-risk model running on the unified profile.
Tech stack: Python, Apache Spark, Databricks, Delta Lake, Apache Kafka, Airflow, AWS.
Matching customer records across systems is never really a tooling problem. It's a judgment problem. The hard part is deciding what 'the same customer' means when two systems disagree, which is why every ambiguous match gets a confidence score instead of a forced merge.
95%+ Match Accuracy Behind the Bank's New Customer 360 Profile
- By building a continuous identity resolution layer across CRM, core banking, servicing, and support, GroupBWT achieved 95%+ customer-matching accuracy across the bank’s four core systems.
- GroupBWT replaced manual matching with deterministic and probabilistic resolution rules, cutting the bank’s reconciliation effort by 80%.
- Relationship management, servicing, and risk teams now work from one Customer 360 profile instead of disconnected customer records spread across four core systems.
- The unified profile already powers a live churn-risk model for the risk team, with next-best-action, segmentation, and fraud detection use cases queued up next.
- Profile updates now propagate near real-time across systems, replacing the batch refreshes that left teams working from stale records.
Ready to Build a Customer 360 for Your Bank?
We help top-tier banks build a Customer 360 — one trusted profile across CRM, core banking, servicing, and support their AI, personalization, and risk teams can build on.
You have an idea?
We handle all the rest.
How can we help you?
