We’d like to introduce you to a new member of our industry advisory council: Carl Reed, formerly managing director at global financial services company Goldman Sachs and Credit Suisse. Carl is an expert in data discovery and analytics for financial institutions, and recently participated in our Tweet Chat earlier this month on “An Insider’s View: Finding Value with Data Engineering & Semantic Standards in Finance.”
The following is a brief Q&A with Carl.
Q: Tell us about your background in the financial sector.
My experience in the fintech space spans more than 25 years as a developer, engineer, strategist and enterprise architect. More recently, over the last six to seven years, my focus has been on data, ontology and graph semantics.
I have spent a significant part of my career in the financial industry with 20 plus years at Goldman Sachs and then, more recently, Credit Suisse. I didn’t start in the Financial sector, I was originally a physicist in Research and Development sponsored by Rolls Royce, then Oil and Gas exploration with 3 years as an engineer working offshore in the North Sea followed by 3 years in management and technology consulting with Price Waterhouse. I only mention these others because that's where I learned the importance of multi-disciplinary teams and the value of combining technology, engineering, analytical and business expertise with a sum of the blend always being more effective than any one particular ingredient.
Q: What would you say are the top two or three applications or ways that financial institutions can extract value?
For the financial sector, I would say the general goal is actionable business intelligence. I’d suggest this falls into four key “directorates” namely: Client, Market, Operational and Risk & Reputation. The analytical focus for each is obviously different, although much of the raw data signal they require is common across all (i.e. they have a common interest in similar “things”) but more specialized interest in particular “relationships” between these things and how they navigate them to get from a relevant event to some kind of actionable outcome.
Q: Where do the compliance standards fall in? Is that under FIBO or a tighter authority?
I think compliance is an important sub-domain of the Risk & Reputation directorate. It can be architected as a parochial vertical acquiring authoritative source data itself, shaping it for its own needs and then managing that shape parochially as one of many shapes across the enterprise. Alternatively, you could architect the acquisition and shaping of the required authoritative data at the enterprise level and make compliance an enterprise data client. Compliance can focus on what it needs to do with harmonized data such as surveillance analytics. The enterprise gets to demonstrate how it is using an authoritative source registry, enterprise business glossary, managing data lineage and governance. All of which are regulatory requirements across current and pending regulatory deliverables such as BCBS 239, CCAR, FRTB, etc. If you go down the semantic route FIBO can provide a good starting point to seed a common enterprise lingua franca your organization can build on top off.
Q: What are your thoughts on semantics and graph-based analytics?
I think there is enormous value for any organization if it can apply common semantics at the enterprise level, i.e. create harmonized high fidelity data using a demonstrable process and governance at the organizational “center”. Such data as an enterprise asset can then be leveraged by multiple analytical, modeling and reporting teams on the organizational “edge”. Unfortunately, organizations don’t have the luxury of starting with a blank sheet of paper. Everyone has degrees of a plethora of disjointed, ambiguous and duplicative information marts that represent a problem with too much complexity, risk and cost to address in its entirety. However, the problem can be addressed incrementally with the right top-down support. The balance should be to move forward using a semantic approach capable of addressing the new as well as integrating with the current ecosystem, providing the capability to chip away at the old.
Graph-based analytics under the right circumstances are extremely powerful. For decades the dominant financial information structure has been relational, so I think it is naive to expect that to change overnight. However, I think it is equally naive, as well as dangerous, not to innovate and evolve. I see graph and graph analytics as an important evolutionary step forward
that also happens to lend itself to a semantic approach to data. Using a consistent conceptual representation to generate binary building blocks or atoms of “things” and “relationships between things” is a powerful foundation for any information structure. Graph is unique in that it represents its “relationships between things” (or edges) as formally as it does its “things” (nodes). So for example, if I model the transitive concept of “component of” between relevant things in my data center. Then, apply that to my inventory data and use the resulting atomic relationships to build a binary graph, something has broken. What do I need to triage? I need to change something - what change risks do I need to proactively manage? A graph is THE optimal information structure for this type of relationship mining or link analysis. Given the highly connected and continually connecting nature of the modern world I don’t think its hard to see the value proposition of graph.
Semantic technology is designed to drive consistency and connectivity. An ontology can formally represent these knowledge concepts. Cambridge Semantics Anzo Smart Data Lake® platform can take those concepts as a specification and generate an implementation that can be executed against raw data at massive scale using Big Data technologies such as Hadoop and SPARK.
To hear Carl discuss some of the most pervasive Financial Industry data challenges and their solutions, watch our on-demand webinar "Applying Data Engineering and Semantic Standards to Tame the 'Perfect Storm' of Data Management".
