Reputation.
Trust.
A higher STP rate.
Revolutionising Data Analysis: taming massive FX Data sets for a global market leader
One of the largest and most efficient FX trading businesses wanted more: to improve its market leading 98.5% STP rate. Beneficial to improving the STP rate is also cost savings and positioning the business for growth in the dynamic and fast-paced FX market.
Challenges
As a market leading global FX businesses, the data set was large.Three months of data generated a 90+ mm row file from a single system, with several others of 20+ mm.
Multiple trade versions, not all of which would be present in all systems; trade IDs that could be amended and multiple aggregation levels added further complexity in a tens of millions of rows data set.
Overcoming data complexity and size constraints in existing infrastructure: the client had an existing data analysis infrastructure but its functionality had to be throttled back due to the data’s complexity and size.
Data Solution
Designing the analytical approach
Calimere Point’s data analytics tooling allowed us to connect disparate data sources to clean and tag data and easily join across and between aggregation levels.
No data warehouse. No data model required. Or functional specification. Or detailed BRD.
We also used modern data visualisation tools to help the data analysis and data story.
Uncovering trade exception causes through data-backed analysis of 95 million rows
As the FX trading firm was already at the leading edge of efficiency, so benchmarking, extrapolation or work on sample data would have been inadequate.
Any conclusions had to be explicitly backed up by real data as unambiguous proof of the root cause: which events in 95 mm rows led to the trade exceptions?
We looked at the exceptions and, following the data, recreated the chain(s) of events and business rules to find the patterns across the systems.
Data analytics value creation: 20+ headcount reduction from efficiency gains
Our client’s senior management was provided a view of their business which they had not seen before. It allowed them to adjust the operational model to reduce cost and improve process efficiency.
The resulting data model could be implemented in production and run on an automated basis, with alerts based on business rules to detect upstream exception causing events.