U.S treasuries trading platform

Building real-time trading for the world’s largest bond market in just three months.

Big market and a big problem

A tier 1 investment bank found its home grown trading platform for US Treasuries was not able to deliver the functionality required of it for complex trades, and had problems rolling out to mobile.

The incumbent platform, which took two years to build, focused processing in its front end, which required it to be installed on the desktop to allow adequate processing power.

This was a challenge for internal clients in the bank and limited use by external customers. Adaptive had six months to turn this system around, coping with the fastest, most liquid bond market in the world, with average daily trading volumes (ADV) reaching US$612 billion in October 2018. Within three months, the bank was trading on a newly built Adaptive system that fully met its needs.

A new approach

The firm wanted a global platform that would show live price streams for cash bond instruments, with the ability to execute limit orders. The architecture had to be flexible to support trading of complex instrument packages, for example, user-definable multi-leg configurations. It would need to integrate with the global order management team, forwarding information for review of trading activity. It was also expected to roll out across all devices, ultimately providing a mobile service offering.

An overarching challenge was working within the technology and time constraints of the existing project. The client wanted to work within the original timeframe, which had taken two years to date, meaning the system would need to be live within six months of engagement, with no loss of functionality.

Adaptive decided to re-architect the system due to the irreconcilable technical issues with the home-grown platform, taking the following approach:

  • Processing conducted on the server side, with architecture based around Java 8, using Nirvana as a messaging bus to support rapid deployment on mobile and desktops.
  • Built with Redux with Reactive Extensions for Java Script (RXJS) for modelling the streams of data, collectively allowing a lighter, more efficient architecture.
  • Employing TypeScript over JavaScript for coding enhancing speed, and feedback from the compiler – which turns the programme into code – when making changes to the platform.
  • Stream-based integration with downstream systems, including regulatory platforms,pricing and execution.

Delivery

In-house deliveries team had used six month and one year delivery schedules; Adaptive’s Agile approach delivered useable tools in fortnightly increments, setting a road map for big feature rollouts using fortnightly iterations.

“Our more agile approach has had many benefits, but one of the most valuable to the client has been the increased engagement with stakeholders” says Mike Stevenson, Principal Engagement Manager at Adaptive.

“Now, they see a robust new iteration every fortnight. They make course corrections, prioritize a clear backlog, and provide quick feedback that makes sure the platform is on the right track. This has completely changed the relationship between tech, and the business.”

Outcome

Once the PoC had been seen by the client and approved, it took three months to go from a blank sheet of paper to a system in production that provided full functionality. Single leg products are in production and multileg products in the pipe. The roadmap for 2019 includes charting, more custom products and the platform is in pilot phase with one external client, transitioning to global clients in Q1