Relationship Analytics, Social Network Analysis

Behavioral Science & Deep Content Analytics

Wrapped in Machine Learning

Catelas is the only solution to take a truly Holistic Surveillance approach to Surveillance.  We not only integrate trade events, with communications and security data but we also leverage a combination of sciences to triangulate our understanding of the risk associated with people, their communications and their behavior.  Machine Learning is deployed throughout the application to help create closed loop learning patterns deep within each algorithm.  The net effect is a technology that better understands your world each day and helps the compliance analysis visualize how a firm does business and where the risks lie.

Our patented algorithms automatically uncover how people connect in large networks, the strong relationships within them and who matters. At the core of all our applications are the patented algorithms that look to understand People, the strength and nature of their Relationships, and Human Behavior. Human behavior is universal – how relationships grow and strengthen, the level of trust developed between people, and behavioral patterns are the same the world over. Our technology can uncover behavioral signatures indicative of actions such as fraud, vendor kick-back, and information theft. Our software can even predict when an employee will resign before they resign – irrespective of whether that employee is American, French or Chinese.

To do this we analyze the ‘Meta-data’ of communications such as email, IM, chat rooms etc., and examine the patterns of Interactivity between each person over time. Meta-data includes fields such as the Sender, the Recipients, the date / time stamp etc. and not the content of the message – we don’t care what you are saying or in fact doing. The fact that we don’t use the content to understand relationship strength or other types of behavioral attributes mean we are both language and culture independent. Catelas does not count emails. Instead, the software looks at how employees interact over time and we use this to identify ‘shared experiences’ between people.


Shared experiences are how humans build trust and deepen friendships. Examples range from sharing a dorm room in college to working with a team of colleagues on a new product launch. Some relationships are short and sweet while others last a lifetime. People become connected in many ways – climbing Mt. Kilimanjaro with 10 strangers that quickly become close friends or collaborating with a colleague on a new customer proposal. Catelas looks for these occurrences and accumulates them over time to measure the strength of a relationship. At any particular moment we can uncover and map out the ‘Interconnectedness’ within any network. This allows us to effectively uncover and visualize how employees interact within and outside of their company.

Current Approaches to the Problem

Current monitoring, investigative & eDiscovery solutions today rely excessively on collecting data and keyword searching to find ‘evidence’ to support or refute an issue. Clearly, if you chose the wrong keyword or fail to define it properly (e.g. ‘dog’ versus ‘canine’) you will get a large number of false positives that waste time and money, but more importantly create opportunity cost – what you fail to uncover prevents you from adequately meeting your objectives, whether that’s protecting your IP, negotiating with a regulator or litigating a contract dispute.

Advanced versions of the search engine have been developed which try to address this problem. You can build complex Boolean or linguistic rules to differentiate between ‘parking’ a stock (a bad thing), ‘parking’ a car (an everyday thing) and ‘parking’ on ‘Park’ Avenue (a difficult thing). However, these better mouse traps miss what is the bigger issue – what if you have not identified the right person, or what if people use slang or code (it’s on the shelf). Either way current approaches overload people with messages to read. It’s just too slow.

So why did we get caught into this search trap? Law enforcement for thousands of years never conducted investigations by gathering the usual suspects and asking them all the same 15 questions as before. Law enforcement usually first tries to identify the people and then understand how they connect – what they have in common – and from there digs deeper with interviews. So while search engines are valuable, there is a gap in existing processes. Catelas fills that gap by identifying the people and how they connect first and using relationship & behavioral algorithms to focus the analyst on the key communications to read or search.

Client Relationship Database

Email Message Log Files

The Catelas suite is underpinned by a Relationship Database that is created by ingesting email log files from email message servers WITHOUT INTEGRATING WITH THE EMAIL SERVER. Every email server generates a log or record of every email sent or received by that server. These logs only contain the ‘Meta-data’ of the email and not the content of the email or any attachments. So the logs tell you who sent the mail, who received it and when.

These log files are valuable, because it allows Catelas to map out the entire relationship network of a company’s extended enterprise – internal relationships and external ones, with webmail and 3rd parties, for example. This is highly scalable, is deployed literally in days, is non-disruptive to IT & the business and does not need to integrate to Exchange or Notes.

Additionally, within Surveillance and Monitoring, our advanced behavioral algorithms use these email transaction logs to identify inappropriate relationships, IP theft, HR risk and anomalous behavior as a way to detect & manage high risk events.


In creating the Catelas Relationship Database from log files, all personal message data is stripped away at import and only the name of the person retained along with a Relationship Strength number derived from the interactions themselves, before they are discarded. Essentially, the Catelas Relationship Database is exactly the same as a CRM whereby the names and connections are added automatically rather than manually by someone in the field.

Financial Services

In financial services, where full content needs to be reviewed, all messages are imported directly via Exchange journaling, Bloomberg XML, Bloomberg Chat room data and various IM feeds.
The content is maintained for a rolling 90 day period, while the audit trail itself is kept for up to 7 years depending on the local regulatory requirements.

Intelligent Collection & Monitoring

Only for Security & Legal applications, not for FCPA Compliance, do we retain the email logs, so that we can view the message headers themselves and link those back to the emails stored in an archive (via direct API call) or on the servers/laptops. This allows you to use the network maps and the advanced algorithms to identify key people for preservation and isolate the content that needs to be collected for review.

"Catelas is an interesting concept and really where the future of ECA {Early Case Assessment} is going."

− Barry Murphy, eDiscovery Journal

Other Products