Market Fixing Scandals Drive Banks to Ramp Up Voice Forensic Tech
Market Fixing Scandals Drive Banks to Ramp Up Voice Forensic TechReuters

The world's biggest banks and brokers have stumped up billions of dollars in global fines after rafts of traders were found to have been fixing Libor, foreign exchange, and commodities prices.

While allegations over market manipulation take years to investigate, phone calls and instant messenger transcripts have largely accounted for watchdogs' successes in nailing down would-be price fixers with a 'smoking gun'.

However, one of the largest emerging trends from the scandals is that the banks are adopting voice forensic technology to help monitor and help regulators' investigations and reviews into allegations over market fixing.

So, IBTimes UK exclusively caught up with Simon Richards, CEO at Fonetic USA to get the behind the scenes info on how banks are planning to tackle market rigging through tech.

Q: How have you seen uptake grow in voice analytics since the spate of Libor fixing settlements since 2012?

A: Fonetic has experienced overwhelming demand and interest in its voice analytics software since launching outside the Spanish market in April last year.

A big reason for this has been the need for compliance officers at financial services companies to do more to proactively monitor and manage their trading floors, both for regulatory compliance purposes and to prevent abuses such as attempts to manipulate Libor.

By capturing and indexing data from all voice, email and chat interactions on the trading floor, our voice analytics software flags trader malpractice in real-time and can reconstruct the history of any trade in roughly 30 seconds. Existing approaches to trade reconstruction can be somewhat akin to finding a needle in a haystack, taking weeks if not months to fully piece together.

Q: How does the technology work in terms of flagging up conversations that may be of interest in supervisors looking to monitor potentially egregious behaviour?

A: Fonetic processing is based on direct phrase recognition which means that it converts speech directly from the audio by looking for specific phrases and strings of words that have been pre-defined as 'watch' words.

It does this in two ways. Keyword searching is the easy part with the software simply monitoring for a word or words that have been flagged. Category or Topic Searching is the really interesting part as Fonetic's linguistics software allows the user to look for an irregularity when a word hasn't even been mentioned.

For instance, if a trader were to use the words "fruit, yellow, monkey and long", the system would automatically identify this as being a "banana". The same works for standard trader terminology and associated lingo.

Q: How does it work around people using code – surely more people (who do want to rig markets) will try and evade new systems like these?

A: Trading terminology is evolving and changing all the time and we know that for software such as ours to be effective, it also needs to evolve. If a compliance officer detects a new code that they want to track, it's very easy for them to set up new market abuse categories that will look for these risk patterns and codes.

Libor Fixing Scandal: 'Vast Amount' of Evidence Stacks Up Against 3 ex-Barclays Traders (L-R) Jonathan Mathew, Peter Johnson, and Stylianos Contogoulas
Libor Fixing Scandal: 'Vast Amount' of Evidence Stacks Up Against 3 ex-Barclays Traders (L-R) Jonathan Mathew, Peter Johnson, and Stylianos ContogoulasReuters

For people who set out to specifically rig markets, the software builds 'Trader Trends' reports that can detect unusual, or even slightly different trader behaviour at any given moment. Once a flag has been raised, closer scrutiny can be applied."

Q: Are there some institutions that adopted this software ahead of others and are therefore a bit more ahead of the curve?

A: Our software has been developed over the past five years by some of the most advanced computer scientists, mathematicians and linguistics in Spain. The success of our software with financial institutions such as Santander Group and Banco Bilbao Vizcaya Argentaria (BBVA) led us to launching the company globally midway through last year.

For institutions that have adopted our software, we have enabled the detection of unwanted behaviour in near real-time and enabled compliance with regulation such as the Dodd-Frank Act and the Markets in Financial Instruments Directive, which requires them to provide the complete history of trades within a strict time frame.

The penalty for non-compliance or fraudulent trader behaviour is severe enough to suggest that financial institutions who have not adopted voice analytics software such as ours ought to think about doing so very quickly.

Fonetic is a voice analytics firm that integrates unstructured voice data into business processes for employers to better monitor staff.