Trade in “human-traded” markets is more nebulous compared to others, as the associated transactions are not handled electronically. A Fortune 50 customer knew the challenges of this over-the-counter method first-hand. When purchasing oil from third-party brokers, its traders still communicated via online chats and phone calls. These communications were then documented using chat logs and call transcripts.
Because these data sources are all unstructured and stored in multiple locations, this data could not be interpreted easily by a human or a computer. Complicating matters further, most of the dialog recorded was about non-trade-related topics – in other words, general conversations that needed to be weeded out of data analysis. The company’s traders needed a way to analyze this data to improve the efficiency of current and future trading.
Management wanted a solution that would empower traders to make practical use of their historical data, quickly, easily, and independently. The ideal solution would enable traders to view and analyze historical data on demand in order improve current and future trading. This would include allowing them to test their own hypotheses before making decisions, as well as uncover hidden trends that could be leveraged for competitive advantage.
How Maana Helped
To address this need, Maana first captured domain expertise in the form of a comprehensive “Trader Dictionary” that contains the majority of all trade-related information in both its abbreviated form (as found in chat logs, such as “sing 380 330/340 ap”) and its normalized, or full-length form (or Market = 380cst Sing Bid Price = $330/barrel, Offer Price = $340/barrel, Tenor = April, respectively). Business processes were captured as a series of complex regexes (or regular expressions), by which different forms of trade-related information could be captured from a variety of patterns, or as broker mapping data, by which a broker can be associated to a specific region and market.
Maana worked with traders to identify the key data they wanted to be able to analyze: six years of semi-structured historical chat log data from four different Traders, which included CSV files and six years of structured historical data from a third-party clearinghouse. Completed trades were stored in these CSV files – data that could be used to normalize and compare trade-related information found in the chat logs.
The models created for this solution matched patterns from the unstructured chat logs with the previously created trader dictionary. They then compared the two data sources to determine potential trends. The output was sent as a single file to be consumed by traders.
Whereas before, traders lacked key information and insights to make important decisions, now they can use Maana to make key business decisions 100% faster than before. The platform’s structured and normalized output empowers them to review historical trends and investigate their hypotheses at will, quickly and easily.
As far as management is aware, no other competitor was performing analytics on this market data, including the clearinghouse company that stores completed trades. Management expects that by using Maana, they can potentially have a huge advantage over other traders in the 100+ markets in which the company trades.