In a past life, I was the R&D manager responsible for the rewrite of our aging Risk Management and ALM platform. The rewrite was extremely successful except for one specific area: the end- user facing analytics support across the various points of use in the software.
What an ALM / Risk Management / Regulatory platform looks like
An Asset and Liability Management, Risk Management and Regulatory platform is a data transformation stack starting at the raw data input from the bank's back-office systems and ending at reports and results used to make strategic decisions, control the operations of the bank, satisfy the regulatory burden and help the bank navigate their present and future exposures and understand the impact of their potential growth targets. Along the path of data transformation there is a lot of modeling activities, from option pricing to future market prediction scenarios.
When looking at any particular contractual or transactional data and corresponding impacts on the end results, a huge amount of data points contribute to the calculations and both traceability and understandability are complex matters. If we add data or interface level quality issues sometimes finding information is like chasing a needle in a haystack.
Many of the results that allow us to understand how the raw data aggregates into final results are the result of additive linear processes, many others are not which makes the need for powerful and flexible analytics and data exploration features even more strong.
Example of activities inside such a FinTech software platform
Exposure Analysis
Exposure analysis is the assessment of how a certain movement in the outside world - from changes in the interest rates up to changes in the credit rating of a given client - impact the potential gains and losses of the institution using the platform.
The activity consists in tracking over time and inside future simulations the gains and losses and associated metrics and understand how the exposure to the market is spread along the huge amount of data that contributes to the exposure. This means that from the results, we would ideally like to be able to drill-back to the various groups and individual contractual relations that build up the end result. With multiple millions of such contracts, if we want to avoid the needle in the haystack exploration problem we need to be able to group, select, slice and dice the results or even perform more complicated cluster analytics that highlight the movers automatically.
Regulatory Requirements
Regulatory requirements come in multiple flavors, from pure disclosure of the financial institution's reality to prescribed limits and ratios it needs to hold as reserve in order be legally compliant. Small deviations in some elements of the commercial strategy or the market can massively change the amount of cash and other liquid assets that need to be held in liquidable positions and thus impacting the availability of liquidity allocated to more profitable activities. Again, finding what contributes how and where to the reserve and other capital allocation constraints is key to run efficiently the business.
Forecasting
As a going concern business, the current position is only the starting point of the work, simulating future investment strategies along with potential future market scenarios is the core of Asset and Liability Management in the modern age. This means multiplying everything from exposure analysis to regulatory requirement by the number of strategies and scenarios considered and the time granularity of future simulations that properly map the dynamics of the whole system. From an analytics standpoint, we move from analyzing the interaction of millions of contracts with 1000s of risk factors and internal and external constraints to the same multiplied by 1000s of scenario / strategy pairs along 10 or even 100s of future time horizons. The haystack quickly turns into a sea of data.
Where did we want to have Data Analytics from a Product Management point of view?
The short answer to that question was everywhere; from the place where custom inserted contracts were editable to the place where future scenario of market evolutions could be modeled up to the end reports or the place where strategies could be created. Data Analytics should have been a supporting function throughout each and every point of use of the software and not only at the reporting level.
Why not implement it ourselves or outside the platform?
The skills and competences needed for the development of an ALM / RIsk / Regulatory platform are the same like with usual operational systems: compute a lot of things on individual atoms of data and roll them up into higher and higher level end results. It means, from a technical perspective, to be able to distribute and process the data in a performant and scalable way and store the results back on the granular level for future use. Data Analytics is almost the opposite activity: be efficiently able to work with the whole data reality and allow to slice, dice, explore and explore the impact of a small change on the whole system in near real-time setting. It's a separate project comparable to the platform itself. We ended up doing it but spent a lot of time on a not completely optimal solution because of budget and time constraints.
The ideal Analytics Platform we never found
In order to achieve these requirements what we needed was:
- Ability to efficiently query a large volume of data (think billion data points)
- Ability to organize the results in a multi-dimensional way
- Ability to integrate existing computations into the data analytics platform
- Ability to integrate authorization in a transparent way
- Ability to have data flow control on a granular level between our host platform and the data analytics platform
- Ability to completely blend the visual elements of data presentation into our software
- Ability to integrate back from the visual part of the analytics into the platform's functional screens including filters and more contextual data
We tried working with existing BI and reporting tools but embeddability was and still is a second thought for those so we would never have had a user experience of one platform nor any reasonable way to integrate on the technical level like authorization, process flow automation and - most crucially - at a functional level allowing to reuse the features and functions of our existing calculation engine in the BI tool as part of their rollup and aggregation strategy, as well as the connection back from their reports or exploratory features into the more operational parts of our platform.
The ideal platform we never found is icCube, a product meant for integration at any level from the calculations to the security and branding level. A product able to sustain large data sets and allow hundreds or thousands of end-users to work with it in real time.
Does your FinTech Platform needs resonate with my story? Get in touch with us and let's discuss your needs!