Data Reconciliation Automation

A mid-sized financial services firm achieved a £2.1M annual operational saving by replacing three manual reconciliation teams with an ML pipeline integrated into their core banking infrastructure. Payback period: five months.

The Challenge

The firm was running three separate manual reconciliation functions: nostro account reconciliation, inter-entity position reconciliation and client money reconciliation. Combined, these functions employed 14 FTEs working daily across multiple systems, manually matching entries, investigating breaks and escalating unresolved items. Break investigation was the primary time sink: analysts spent an average of 3.2 hours per break tracing entries across fragmented systems. FCA CASS regulations required same-day resolution of client money breaks, creating daily deadline pressure.

The Trovix Approach

Trovix built a unified ML reconciliation pipeline that connected to seven source systems via API, automatically matched entries using a combination of exact matching rules and ML-based fuzzy matching for complex cases, classified break types, and generated investigation packages for human review of unresolved items. The pipeline runs continuously, processing new entries as they arrive rather than in overnight batch runs, enabling same-day CASS compliance without manual intervention.

Technical Architecture

Outcome

The three reconciliation functions were consolidated. Headcount redeployed to higher-value analysis roles. Reconciliation accuracy reached 99.7%, up from an estimated 96.4% under the manual process. The annual operational saving was calculated at £2.1M, with a payback period of five months on the Trovix engagement cost. CASS compliance is now met automatically, removing daily deadline risk.