Better together than apart 

Neil Sandle is Head of Product Management Asset Control Data is increasingly talked about as ‘the new gold’ and the fuel to power business performance. In few sectors of the economy is data more highly prized than in finance where it is widely used for everything from risk management to investment decision-making, and from customer analysis to fraud prevention. But collecting data for its own sake is of little use in finance. Bringing intelligence and facilitating access to that data is key. Keeping the cost of change in mind, the emphasis needs to be as much on the adaptability and extensibility of data models and the onboarding of new data sources as it is on data aggregation capabilities. We are witnessing a shift in the way financial institutions focus on data management and analytics. Historically, the two disciplines have been largely separate. The data management process typically involves activities such as data sourcing, cross-referencing and ironing out any discrepancies via reconciliations and data cleansing processes. Data analytics processes are typically carried out afterwards in a variety of desk-level tools and libraries, close to the users and typically operating on separately stored subsets of data. This divide has created problems for many financial institutions, with the separation impacting time to insight and acting as a brake on the decision-making processes that drive business success. Data was typically held, and often still is siloed in data stores and in legacy systems where accessing it was difficult. The metadata surrounding the data was often not updated frequently, making data lineage and understanding of the relevant permissions around the data difficult. Judging whether data was fit for purpose was complex and frequently models came to suboptimal results because they were based on stale, incomplete or otherwise inappropriate data. Scoping the challenge Certainly many quants and data scientists today still face logistical issues in accessing the data they need for their decision-making. Often, these analysts still find they have to contact the IT department to write a query, set up a report, or they might confront licensing restrictions or run into other permission issues. Even when quants access the data they need, there are often additional issues to address. Invariably they find that the metadata that should provide insights into the quality, the origins of the data, what the license permissions are and who has approved its use, has not been tracked. As a result, they may conclude that they do not have enough context to decide whether the data is fit for purpose. Furthermore, because data and analytics typically remain decoupled within many organisations, quants will need to run two separate processes to get hold of usable data. Apart from that, the scope, breadth and depth of data often changes, leading to repeat request to keep the data up to date and as comprehensive as possible. Different selections of data may need to be presented in different ways, depending on the use case. If quants want to run a financial model, they will typically look to gather the data that is most relevant for their use case, store it in their own database and then run analytics on it. They will not be able to push their own model to a central processing framework that runs as a shared store of market data. These are limiting factors on user enablement within financial organisations and stimulate redundant copies of the data – with all the overhead and operational risks that stem from that. In the new world order, where data and analytics are increasingly integrated, none of this, in theory, should need to happen any longer. Building the right structure Today, many firms understand they need a better way to provision their data scientists and other key users with clean price and market data. So how do they go about making that happen? The technological capability is certainly increasingly in place. A shift to the cloud and the adoption of cloud native technologies is helping firms transition to a more integrated approach to data management and analytics. Apache Cassandra for example, has emerged as a highly-scalable, open-source, distributed database, that makes it easier to securely store and managing large volumes of financial time series data. Apache Spark is a unified data engine for big data processing. Taken together, the two, and other associated tools, are helping to facilitate the integration of data and analytics. At least in part as a result of this, we are seeing data management and analytics increasingly joined at the hip. This integration is also blurring the line between source or primary data (from the stock exchange, for instance); and what is derived and calculated (for example, interest rate curves, correlations or volatilities). Data management and analytics are today two sides of the same coin in user workflows. Increasingly also, the focus is on bringing analytics to the data rather than vice versa. In other words, it is about moving the analytics capability to where the data resides rather than moving large stores of often siloed data over to the analytics function which typically led to inconsistent copies over time and lots of analyst time spent verifying and gathering data before the actual analysis could start. Data quality matters Despite all the above, though, for analytics to work effectively and efficiently, the data that fuels it has to be of the highest quality. Good quality input data makes analytics more reliable. Conversely, even the best model will produce useless results when fed with poor quality data. This drive towards efficiency and accuracy as businesses look to turbo-charge their analytics function is one reason why data quality matters to the finance sector today. The other is that regulators are increasingly scrutinising data quality, and in particular the quality of data that feeds into models. Financial businesses will often need to explain the results, not only the mathematics of the model itself but also the data that went into it, what the quality issues were, what the sources were and who touched it on the way. That can be difficult if they treat data management and analytics as separate disciplines. Without having the ability to analyse the data and the oversight of where it has come from and is going to, businesses struggle to gain transparency over how they are provisioning their models with data. They also need the data quality capability in place to ensure that data is consistent and validated and that the burden of reconciliation is reduced. Making it happen for users With the above technological capability delivered, financial services firms should be in a position to better provision their business users. As the cycle of managing and processing data extends to take in analytics, users within financial services organisations increasingly want to be empowered by the process and use these new capabilities to drive better informed decision-making. This move to data-as-a-service (‘DaaS’), when combined with the latest analytics capabilities, is starting to make this happen for financial organisations today. By adopting this approach, they can gain access to multiple data sources and also multiple data types, from pricing and reference data to curves and benchmark data, ESG and alternative data. With the help of quant languages like Python and R, firms can create a robust and scalable data meeting place enabling users to share these analytics across their entire data supply chain and develop a common approach to risk management, performance management and compliance. Quants and data scientists benefit from this through increased productivity. We are seeing many data analysts today that are looking to dig into the data to find signals that help them discover investment signals and returns in the market. Data scientists are looking at historical data across asset classes looking to distil information down into factors including ESG criteria to operationalise it into their investment decision-making process, and increasingly too, they are incorporating innovative data science solutions, including AI and machine learning, into market analysis and investment processes. The new methodology enables the construction of proprietary analytics, with any combination of data types, to support activities across the supply chain (including investment decisions, valuations, stress-tests, return series, performance calculations and risk figures,) and also to conjoin views upon the newly combined data. But this approach to user enablement is also helping to democratise analytics, bringing it into the orbit of those who are not data experts. Today, thanks to the contextualisation provided alongside analytics, it is not just the preserve of the quant or the data scientist, but a key tool that those less expert in data, can use to drive business decisions. The key once again is ensuring that data is easy to use. With data and analytics seamlessly integrated in a new methodology, it is straightforward to be able to seamlessly onboard and maintain data (vendor) sources and data types, adding them to the data supply chain for easy integration. It is also easier to perform data quality checks on incoming data and troubleshoot data quality queries through data lineage track and trace capabilities. This in itself drives business agility but the combination of data and analytics can also help businesses optimise costs. It does this by supporting greater agility with the data, selecting only the best sources and data (external or in-house) to fuel investment management processes and to help drive the business forward. This is, however, also about avoiding data duplication, controlling data access and centralising all data requests. In more granular terms, the coming together of data management and analytics enables organisations to reduce cost on data spend by implementing optimal data sets and drive further efficiencies by readily deploying standardised data models. It also allows them to lower the cost of data integration, switching, changes and ongoing maintenance and further control over costs by achieving oversight of how and where data is being used across the organisation. The reverse side of the cost control coin is the ability that the new approach provides organisations to maximise the achievable return on their market data investment. Part of this comes from improved data access – in particular, the ability to access and switch between data from multiple (vendor) sources across multiple data types, including pricing, reference, curves, benchmark, ESG, fund, alternative and in-house. Another key element is the greater return on market data investment that comes from enhanced data quality. By validating, checking and normalising sources, financial institutions can reduce the reconciliation and cleansing burden. Furthermore, by ensuring they keep data quality levels high. They help to eliminate model risk and unify data across securities, positions, portfolios and benchmarks, further accelerating the process of maximising market data RoI. More broadly, the approach brings a wide range of benefits, helping financial institutions get much more from their investment in market data. At the top level, the integration between data and analytics helps here simply by allowing firms to make more and better data-driven decisions. But the ability to manage data more efficiently brings further advantages, enabling financial businesses to more easily switch source(s) without interrupting the established data flow and processes and more readily embed new data-driven concepts to their investment decision process. Closer integration between data and analytics can also help to support regulatory and audit processes by demonstrating data quality checks and full data lineage and allow firms to produce key performance indicators (KPIs) for continuous data improvement. Future focus Looking ahead, we are on the cusp of a new age in financial data management. Today, technology, process, macro-economic factors and business awareness are all joining forces to bring analytics and data together. The result for financial institutions is a new world of opportunity where they optimise costs, drive user enablement and maximise the value they get from 
... we are on the cusp of a new age in financial data management. The result for financial institutions is a new world of opportunity where they optimise costs, drive user enablement and maximise the value they get from data