When good intentions go awry: the unintended consequences of CFTC record retention rules

Juan Diego Martin, COO at Fonetic, discusses how the CFTC’s different record retention requirements for electronic and voice communications are creating a conundrum for firms subject to these rules.

CFTC Rule 1.31 states that electronic communications must be retained for a period of five years, while voice communications must be retained for only one year. This has led to a school of thought in the financial community that if a firm transcribes its voice communications it creates an electronic record that becomes subject to longer retention requirements. They believe this will not only increase storage costs, but may also present a legal risk because the transcription is discoverable during litigation.

It seems that the rule is having the opposite effect to its intentions, as this has led to little or no monitoring of voice communications. In light of this, we could see a re-examination by the CFTC further down the line, meaning that firms who choose to follow the letter rather than the spirit of the rule could find themselves in a sticky situation in the future.

However, it’s not just the regulator firms should be worrying about. There are benefits beyond compliance to transcribing voice communications, as the knowledge of what’s being discussed on the trading floor can be leveraged to reduce risk and improve business practices. So, why are firms slow to take this on?

The human element of manual sampling

Instead of using the latest technology to analyze all voice communications, many firms are still using a sampling approach. Unfortunately, this approach has many flaws.

Audio sampling involves teams of people listening to a sampling of voice communications. Not only is this approach extremely resource intensive, but it is also error-prone and not very effective in identifying suspicious behavior.

A large reason behind this is human error. Humans are prone to distraction and listening to hours of calls can be exhausting. And with different resources listening to different communications, they may miss common patterns of behavior and each employee will have their own inherent biases, causing them to interpret the communications in a different way.

In addition, these resources must be well-trained in the business and securities laws to identify inappropriate behavior. This results in high turnover in these positions as the resources move on to more attractive roles.

To transcribe or not to transcribe, that is the question

Modern voice technology has come on leaps and bounds when it comes to aiding better transcription.

Advances in the technology to transcribe voice to text have dramatically improved the accuracy of the transcription process. Artificial Intelligence can be used to categorize and separate calls, meaning we can automatically sort calls related to trading activity from personal or non-business-related calls.

Machine learning and natural language processing can be applied to understand the context of the conversation and identify potential cases of inappropriate behavior. Alerts can be raised and prioritized as low to high-risk calls, so surveillance analysts concentrate on only the relevant behavior and confirm or close suspicious alerts in a timely and manageable manner.

This approach not only dramatically increases the coverage and efficiency of the surveillance, but also reduces the costs incurred during the surveillance process, which up until now remains a predominantly manual process.

The business benefits of audio transcription

In addition to reducing risk by improving the surveillance capabilities, there are many other business benefits to transcribing voice communications.

Firstly, it means that trades are much easier to reconstruct. This is because the content of communications, and not just metadata, can be analyzed to determine whether the communication is related to a trade. This means that when the regulator comes knocking, firms can have all the relevant information at the touch of a button.

Secondly, it improves searchability. Supervisors can easily search communications to monitor employee behavior and identify areas for improvement. For example, if a customer files a complaint, the supervisor can easily search the communications to identify the conversation and address it with the employee. They could also search all communications to see if other customers may have experienced the same issue.

Finally, and most importantly, not transcribing voice communications is the regulatory equivalent of burying your head in the sand. Regulators have already demonstrated their ability to analyze voice communications in cases involving the manipulation of ISDAfix and LIBOR. Companies that think they can reduce their legal risk by not transcribing voice communications are increasing their conduct risk by not understanding what’s in their voice communications.

Firms who can get their audio transcription right will not only stay one step ahead of the regulator, but could also expect more lenient treatment should an incident occur, as they can self-report issues rather than waiting to be identified and reported by someone else.