Big Data is a term that references the 3 V’s: Volume, Velocity and Variety. Whereas structured data in Financial Services often contains very large data sets and real time pricing and transaction data, it’s really the huge Volume and Variety of unstructured data that adds complexity that traditional relational databases cannot support. This is where new - often open source – Big Data technologies and techniques come into play.
PanoVista.co recently conducted a short survey to Financial Services firms that indicates that while structured internal data management is maturing with active projects planned or underway, the state of unstructured internal and external data is less mature with fewer projects—even though there is an awareness that change is needed. So, what are the practical use cases for Big Data that thought leading firms are engaged in?
In many ways, social media has been the biggest driver of the explosion of Big Data, and that is true for retail oriented Financial Services firms such as banks, insurance companies, mutual funds and wealth management firms. Marketers are creating highly segmented campaigns and using predictive analytics to nurture leads to increase conversion rates online and via social media optimization, or to pass “marketing qualified leads” to the sales team to confirm if the leads are “sales qualified leads” in their pipeline.
The ability to search large volumes of data looking for inconsistencies has a natural benefit for evaluating system logs looking for security breaches or curious user behavior. Furthermore, the horizontal scalability of Big Data platforms can overcome the performance hit associated with database encryption, reducing potential financial and reputational risk of the theft of data left unencrypted in place, which has been a shockingly common occurrence over the last 24 months.
Enterprise-wide Business Intelligence
As noted above, Financial Services firms typically manage large data, and more advanced firms have centralized and put quality controls around their structured data. However, much of this data still resides in silos by business unit, and few firms have taken the step to fully integrate internal unstructured data. The Big Data approach is often called a Data Lake, leveraging Hadoop’s ability to read across structured RDBMS and unstructured data such as email, documents and video and treating all of the data as a common logical data platform. Business Intelligence and advanced data analytics can then be used to generate new insights into the inner workings of an organization to improve efficiency, reduce risks and provide better client servicing.
Top tier hedge funds, asset management firms and banks are the early adopters using Big Data in the Investment Research area. This is a little more controversial since most results today are anecdotal versus backed by formal research, however the anecdotes indicate that traders and portfolio managers are indeed gaining a competitive advantage by having broader insight into their investments, often before traditional data providers are aware of news or can update their feeds. Social media activity prompted the SEC to rule on April 2, 2013 that public companies can use social media outlets to disclose material non-public information. One area that is less controversial is that very large research databases have been proven to be managed and mined more effectively on Big Data platforms such as Hadoop and NoSQL than on traditional relational database platforms.
Big Data is gaining rapid acceptance across most industries more quickly than in Financial Services. The technologies are evolving quickly, and often there is a mix of technologies that are needed for to support a given use case. These technologies are designed to scale horizontally across many commodity servers, so before investing in huge data farms it may be worth considering cloud deployments to support scaling only on-demand and to also facilitate mixing technologies as needed to support evolving use cases. As every data scientist will attest, knowing what questions to ask and plumbing the data are critical steps for good analytics. An Agile project approach is best suited to support the iteration that will likely be needed before locking down an operational process. Therefore, planning on the approach, tools, resources, and anticipated results and benefits should remain part of the business case, just as with traditional technologies.