How technology principles developed in the medical industry are being used to detect and prevent market abuse in financial markets.
Research Background
The World Health Organisation estimates that 39% of people will receive a cancer diagnosis in their lifetime but with early diagnosis, many cancers can be successfully treated and cured. Our first foray into software development was in this important medical area. The principles of our technology were initially developed as a breakthrough technology that analyses cancer ultrasound scans to identify malignant tissue and help to provide an early diagnosis. This analysis needed to be performed in an unbiased way, with very high levels of sensitivity and specificity therefore with as few false negatives and false positives as possible. The implications of incorrect analysis, in particular a false negative, would of course have profound consequences for the patient.
From Cancer Detection to Financial Markets
An ultrasound works by sending out sound waves from a transducer which are bounced back from within the area of the body being examined and translated into electrical impulses. These are interpreted and an image formed. This means that a cancer tumour detected by ultrasound technology would send back different waves than the normal tissue. The issue is how to correctly differentiate the malignant tissue from normal, healthy tissue, and display this on the ultrasound (US) machine monitor, based solely on the RF (radio frequency) backscattered ultrasound data and without conducting more invasive medical procedures.
The research led by the founder of Features Analytics, Cristina Soviany, was aimed at determining statistical methods to better interpret this flow of data from the transducer mechanism. The importance of this is as stated before. Incorrect analysis or what we term a 'false positive' or 'false negative' can either result in causing undue concern for the patient in case of 'false positive' or even worse, a life-threatening outcome in case of a 'false negative' patient who would remain untreated.
The techniques pioneered involved the analysis of hundreds of millions of data points in the 3D ultrasound scan coming from a body organ being investigated and identifying with high sensitivity and specificity any abnormal tissue in the area under investigation. The process then provided a diagnosis to the medical specialist and patient within two minutes with great accuracy and avoiding false alerts.
This need for absolute accuracy in identifying malignant tissue in real-time ultrasound data coming from any US machine (thus very heterogeneous data) lay the basis for what we now know as the eyeDES Trade Surveillance technology.
In financial markets we deal with large amounts of trades, orders, quotes, market data and news events. There are fragmented markets where trading can happen in milliseconds across multiple venues. Many different client types and trading styles are present. For a tier 1 bank, managing this amount of data becomes a huge task. But even for a smaller bank, broker, asset manager or hedge fund, the data is complex and cannot possibly be effectively analysed by a human. You need effective technology to detect market abuse, the cancer of financial markets.
The problems of 1st Generation Solutions
Technology to detect market abuse has been around for many years and most market participants have implemented vendor or in house built solutions. But the effectiveness of these solutions is constantly being questioned. The last survey by leading industry consultants PwC revealed that 50% of banks surveyed said they were dissatisfied with their surveillance solution. From our conversations with financial institutions, we derive that this figure is even higher.
Why is this? Ostensibly, frustrations rise from poor quality alerting, high volumes of false positives and the lack of flexibility that first generation surveillance tools provide. Almost all first-generation solutions provide users with a complex set of parameters that have to be setup and maintained for each market abuse scenario. Often parameters are set "to keep alerts to a manageable level". This is bad practice. The trouble with parameters is that you are constantly fighting the battle of judging where your parameter thresholds should be so as to minimise false positives but be able to catch bad actors. Set stringent parameters and you run the risk of missing abuse, set them too loose and you are flooded with alerts that must be reviewed manually and a very large percentage of them are false positives. It is not uncommon for these systems to be generating 1000s of alerts per day and even at these rates being unable to spot all the suspicious cases.
A New Approach
Features Analytics developed a unique Zero-Parameter technology based on principles that we developed when working for cancer detection. We are now applying these principles to detecting market abuse, the cancer of the financial markets. Our technology constantly measures "normal" activity and identifies anomalies linked to suspicious behaviour. It performs this in a completely unbiased way (i.e. it is not influenced by historical activity) and without the need for users having to set any parameters. The results are a major step change for users. Not only is alert quality far higher than with parameter driven solutions, but we also provide full transparency and explanations on each alert and additionally insight into emerging patterns of abuse that have previously gone undetected. And with no parameters to manage, the user experience is much more efficient allowing them to focus on the real problem areas rather than wrestling with a mountain of false positives. The result is a far more robust surveillance environment to meet the risks posed by errant traders and to meet the regulatory concerns and naturally, a vast reduction in cost of compliance.
To find out more about our approach visit out website at https://www.features-analytics.com/solutions/trade-surveillance