In the March 22 issue of Energy Metro Desk we ran an interview with Juan Manuel Soto, the CEO of Madrid-based trade surveillance and trade reconstruction firm Fonetic. The story was entitled, “Surveillance Holy Grail.” Soto discussed the origins of his company – advanced voice recognition technology for calling centers – and the success he’s been having among some global banks with his product, the ATR Integra platform which seamlessly fuses voice, email, IM and anything else you can think of for trade surveillance and trade reconstruction purposes. We received some decent feedback on the story, which was expected because apparently we weren’t the only people who didn’t know such trade reconstruction was even possible. Thus the title, “Surveillance Holy Grail.” While we didn’t have a chance to demo the system before we ran that initial story, we did speak to an ATR Integra customer, off the record, to make sure what we ran was accurate. So, long story short, we finally had a chance to do a demo of the system, and by and large, we really did get that first headline right. Thus the current headline.
So, pinpoint accuracy in voice reconstruction-to-text output? Check. Multiple languages in the same phone conversation converted to English text? Check. Matching IMs with emails with connected voice calls and trade blotter? Check. In English? French? German? Or most other languages, including Arabic? Check. So, yeah, it worked. It worked quickly, and it worked simply. The system featured preset triggers and red flags, alerts and the like, to keep compliance managers in the loop, in essentially real time. Fonetic’s Ana Santana walked us through.
Once you fire up the system, you may choose to view a set of reports (preset or custom) that, for example, show voice or other communication activity between a particular trader, desk or group, and the various counterparties or otherwise. A second report may look specifically at particular desks in different locales and hunt for different types of market manipulation activities – one field on the dashboard was actually marked “misconduct or manipulation.” Things like spoofing or wash trading activities have already been preprogrammed into the system. If certain activity was flagged automatiche by the system or compliance staff operator, all possible communications related to the transactions are tagged, aggregated and organized. In English. Or, as we noted earlier, most other languages. So, not only is the mischievous activity flagged, it’s also accurately reconstructed, in the blink of an eye. We watched the system translate and transcribe a recorded conversation, in near-real time, very accurately. It even picked up all the charming trader vernacular.
The system is set up to score each trader or desk by misconduct frequency. Huh? So what’s meant by misconduct, you might ask. It could be spoofing-like trade strategies or simply just foul language. Santana tells us you can create any scenario or framework for misconduct you like. The platform has preset stuff and you are free to create more, as you like, in real time. You can thus compare misconduct averages among staff and desks, so the results could be part of your desk’s bonus process, or employment agreements generally. Kind of simplifies the HR process somewhat, eh?
She tells us the many reports that can be constructed is basically limitless. While we were introduced to some very basic stuff, operators may choose a gazillion templates which may be manipulated to your heart’s desire, Santana says. Bubble charts, bar charts and comparisons of every possible design.
The reports are also dynamically designed; if one report is real-time hunting for particular speech or word patterns in email or voice communications, on top of particular trade patterns, across markets, across portfolios, an operator can add a word or trade or commodity into the mix, and the system instantly update the search and surround process automatically. It really was quite impressive.
One screen on the dashboard involved all the communication playback; you pick and choose what you want to see. Email, check. IM, check. Voice, check. Demarcate which dates and time range. Add some additional key words to the search. The system is constantly being updated with slang and vernacular unique to the trading community. Santana keyed in the word ”sell.” A voice clip was played, the text appeared, and the word sell was automatically highlighted, but so was the word ”done.” “Traders tend to have their own language, and different words have possibly different meanings. Our system is aware of this and identifies these different meanings and contexts. In the case of the word sell, we selected a category that includes all words that mean sell. And while the search is in English, it will also pull all the instances of the word meaning to sell, in any language.” So, it would show done, “yours” ”mine” as wellas “verkaufen.”
Santana says they have many different linguistic models for different desks, different countries and markets, asset classes and so on. The user plugs in all the relevant market or portfolio data and the platform matches markets data with linguistics models and report screens are set. Press go.
So, how simple is performing a request for information from a regulator? “Simple,” she says. And so it was. Pick the trader, products, trades, markets, click on the dates and comms and click any number of additional fields, and you can reconstruct the trade almost instantaneously, in one neat package. No more regulator fishing expeditions to deal with. Bam!
Fonetic also includes known market rules for the solution, and is constantly updating the system. Dodd-Frank, MiFID, or anything else you want, it’s in there. And the good news is, it’s now available for our sector, good to go. Implementation times vary and so do costs, but neither seemed particularly outlandish, according to Fonetic’s top guy in the US, Simon Richards. Ping him at firstname.lastname@example.org. For more on the system go to http://fonetic.com/atr-integra/.
Energy Metro Desk, March 25, 2016