Towards Error Detection in Bond Market Reporting Data

Fixed income (bond) markets in the UK and EU are fragmented across multiple trading venues, with no consolidated view of activity. While venues must report transactions to regulators, the reporting data is vast, inconsistent, and non-standardised, making it difficult to analyse. This obscures understanding of market dynamics and makes it challenging to detect anomalous behaviour or potential misconduct. In this study, we analyse a dataset of 17.7 million sovereign bond transactions and 11.6 million corporate bond transactions, reported from multiple venues between 2018 and 2025. The data, standardised by Propellant.digital, a Software as a Service (SaaS) provider for fixed income markets, enables us to develop an automated big data methodology to detect price reporting errors. We identify $4,865$ errors (approximately $0.017\%$ of the data), which would otherwise go unnoticed, and find they are more common in off-venue trades, where regulatory oversight is limited. Detecting errors is the first step of our broader research agenda to increase market transparency and support retail participation in fixed-income markets. Future work will extend our framework to detect subtler anomalies, such as manipulation or shifts in market behaviour, and to build AI-based models that forecast bond market movements using transaction data and external signals. Ultimately, we aim to develop personalised AI investment advisors (smart agents) that offer retail investors tailored, practical guidance to maximise returns within defined risk and time constraints.