Vendors might expect a system to processed in the same manner, but once you see how things really work they may be in for a shock.
Investment teams develop a certain type of pride when they pull out a few nice cards and chips, usually in a stealthy fashion. Models are running, draw downs are on green and dashboards are green. Until one day, something happens: maybe half a point in a central bank’s rate signal catches everyone’s attention, or maybe the port and its attendant shipping channels on the other side of the world has been closed and the impact has spilled out across three supply chains, or maybe a geopolitical flashpoint has been reached, and the sovereign risk is repriced in a whole region in 48 hours, leaving the team with less to show for than they had hoped. It was a "reporting system that was a "reporting system".
The difference between platforms that show you what happened and platforms that can give you insight into what might — is the single biggest question institutional investment operations should be asking about their analytical infrastructure, today. The answer can mean the difference between maintaining capital, adhering to regulations, and even company survival, should the situation be considered a stress event.
The DNA of Change that Makes the Old Risk Paradigms Unfit for Today.
The past ten years post-crisis was very favorable for institutions that engage in risk management. Structurally, volatility of the former was suppressed. By recent history, the central bank's policy was fairly easy to forecast. In general, the correlations among asset classes were as one would expect.Since the period during which the risk frameworks were developed and built around Value at Risk (VaR) calculations, with historical covariance matrices and a set of exposure index sectors specific to beta exposures, these are fairly decent. None of the unexpected cross-asset contagion shocks, such as those of the past 10 years since 2020, tested them.
What the pandemic, the inflation blaze, the rate cycle that followed and the successive geopolitical crisises have uncovered is the basic problem with traditional measures of risk based on assets: They find what markets did, not what markets can do when markets have a very rarely seen and unprecedented set of conditions. Now that a central bank has made a 425 basis-point hike in a 14-month period of a fixed-income portfolio with a long duration of 3 to 4 years, a VaR model designed with five years of daily returns will not provide that answer. Finally, past pre-2020 evidence will fail to take into account the speed at which the link between the quality and momentum factors can break down during an episode of reduced liquidity and credit tightening.
It is not a criticism of the use of the quantitative approach. It's a note on models as instruments and models as decision aids. What institutional investors now require, are the platforms that see historical data as one of many inputs, instead of the key ingredient to comprehending risk.
This show evaluates the strategies and technologies apt for building a modern Financial Analysis platform.
Despite the fact that the interfaces and business models may be completely different, the most powerful platforms in use today can be seen to have the same general architectural logic for institutional use. There are three layers of logic that are applied and are Coercively related.
The first is data integration – not just market data but the alternative and unstructured data feeds that are increasingly becoming the critical data to understand risk before it manifests itself in price. Taking all the above inputs, normalizing and ingesting data from them and linking it with the portfolio team allows the portfolio team to gain a signal before the corresponding move has occurred, as shown by traditional price feeds.
Models and scenario frameworks as well as stress test models are part of the second layer, known as the analytical engine, which turns raw data into actionable risk intelligence. This is where the difference between the sophisticated and adequate platforms can be seen the most. Such a platform can execute pre-defined stress scenarios that are based on some of the past crises situations (2008 playbook, 2020 playbook), and is useful but limited. A platform capable of simulating forward looking, bespoke stress scenarios (such as a simultaneous dollar appreciation, commodity inflation and EM capital-inflow reversal) and overlay them in real time over actual portfolios today is a different kind of risk management.
The third layer, decision-support output, is where the analysis is presented in a way that can make a difference in portfolio decision making and doesn't add to the clutter. The failure mode here is well known: creating platforms that provide detailed reporting information that no one reads beyond the Executive Summary page, or platforms that produce so many reports and alerts, people don't pay attention to them. A sound analytical design approach has output design as a problem unto itself.
- There are some 'Non-Negotiable Features' for any serious Deployment!
Evaluating financial analysis platforms for institutional use can be like sitting in the middle of a conversation as they talk about everything that it sounds cool in the demo, but doesn't matter in real world situations. The key is a less bloated and more efficient list when it comes to preventing the capital loss and operational integrity:
- Multi asset data integration in real-time: You need real-time data integration and normalization of data from various asset classes, spanning equities, fixed income, FX, commodities and private asset proxies and derivatives, without any manual process. Risk data latency is not an inconvenient speed bump, it is a capital risk in volatile times.
A baseline requirement for any platform trying to make a play for the institutional level is now API connectivity to alternative data, much like S&P data, but from other providers as well such as satellite data, trade flow data, NLP processed news and filings or macro now casting models. This alpha and early risk signal is no longer drowned by aren't-EWs and broker-EWs, but resides in these datasets well before it is reflected in what has been considered traditional price feeds.
Pre-built historical scenarios are a starting point, not a capability: algorithmic / custom stress-testing are also required. The capability of teams to create customized, forward-looking, macro environment, policy and liquidity shock stress tests that directly impact their “live” portfolio.
- Portfolio level downside exposure decomposition: Risk attribution that will always allow you to look at the risks on a factor level, geographic level and on a liquidity tier in a way that a risk committee can act on — rather than only seeing an aggregate VaR or volatility.
An automated compliance and regulatory monitoring capability is crucial for institutions with operations in multiple jurisdictions that need to monitor position level exposures and ensure that they are in parallel with regulatory constraints (concentration limits, leverage limits, eligible collateral rules, etc.) without manual checks.
Against the backdrop of a risk of refinancing rates and increasing stress on balance sheets, the exposure to creditworthiness of counterparties who may be deteriorating requires more than just standard credit ratings when this data is structurally a lagging indicator.
- Audit grade reporting infrastructure: each and every risk decision, model input and output should be reported in a form which could meet the standard of an audit from financial regulators and/or internal governance processes. Platforms which do not have this feature puts compliance risks on top of compliance risks when it is under the spotlight of the regulators.
A bulk integration with portfolio management and execution systems across the entire platform: An “island” risk platform (where neither data export from the risk platform nor data match-up with the OMS/PMS is not automated) adds the risk of data errors and slowdown that undermines the purpose of real-time analysis.
This paper examines the increase in uncertainty about prices as companies have transitioned to event-driven risk models.
Among the more important changes in the risk thinking of institutions that have occurred in the last four years has been the change in viewpoint on how to think of uncertainty from as a statistical parameter of returns to as a discrete, event-driven variable to be actively priced.
The difference is significant when it comes to the way things are done. From a statistical point of, however, there is uncertainty represented by volatility estimates or confidence intervals around VaR, or fat tail adjustments to return distributions. By their design they are a backward looking. They say there were “fat tails” in the past and to be mindful that the future may have them as well.
In the event-driven framework, the uncertainties are priced by the evaluation of a set of specific scenarios that have a probability that are geopolitical, macro economical and/or operational and have an impact on the portfolio with enough precision for pricing, hedging and positioning decisions. This clearly needs a new type of analytic platform. Digital Risk and Reward scoring as an input is required, as is geopolitical risk and policy uncertainty indices, as well as central bank communication analysis and supply chain disruption signals, which then are translated into meaningful time granularity, portfolio-level impact estimates.
Already, the more advanced institutional teams work like this. They are not waiting for “volatility” to soar in order to begin to make adjustments in duration or risk exposure in correlated securities. They're trying to identify which pieces of information will be useful when to price certain risks in, such as sovereign credit risk in EM economies, sensitivity to energy prices in European manufacturing, refinancing cliff risk in the leveraged credit markets. Such future-oriented risk management would necessitate an infrastructure in the platform to support it, which is more demanding than the commercially available platforms can today at full functionality.
It's the Data Quality issue no one is talking about enough.
Underlying each of the higher-level risk models is a data pipeline and every data pipeline is underpinning a data governance issue, most of which are less carefully managed than the models themselves are.
This is by no means a theoretical issue. In reality most failures in managing risk in an institutional portfolio are not the result of a model failure, but rather data failure. Stale prices on markets where there is not sufficient liquidity. Wrong mappings of position due to restructuring. Lagged corporate actions that affect the factor exposures. Inconsistency in FX conversion in the different portfolio accounting systems. All of these, when put in a risk calculation, can provide apparently reasonable results which are nevertheless materially inaccurate.
As critical as financial analysis platforms' modeling tools, is their data governance mechanism: how to validate inputs, how to fill in missing data, how to look for anomalies and how to reconcile data from multiple sources. Without reviewing the data layer, it would be no different than checking some structural engineering report with no survey data.
Generally speaking, institutions that have taken a hit on the analytical infrastructure arm are the ones that have taken a big hit on the arm investing in the infrastructure. But those being built now would be wise to consider the Data Quality issue as not an IT implementation challenge, but rather a life-long operational discipline on which every implementer needs to invest and on which every data conversion event needs to be regularly audited.
How Regulatory Pressure Can be an Acceleration to Risk Infrastructure Improvement.
So it's worth noting that the drive for the more powerful investment performance analysis platforms isn't solely on the back of investment performance. The regulator pressure in key jurisdictions is equally important to push the institutional change.
The timeline for implementing Basel IV, the enhanced requirements from the SEC to report on alternative investment funds, the continuance of refinement from ESMA on the risk disclosure rules of AIFMD and MiFID and the greater focus on bank reporting of counter party exposure risk by CFTC have all created a new baseline for the amount that risk infrastructure needs to produce. By 2019 standards, well-managed institutions are discovering that their current risk management processes are unable to provide the granularity of regulations that is called for by regulators.
The ability to comply with regulations and the overall integrity of analytical infrastructure go hand-in-hand. Institutions with a focus on platforms equipped with sophisticated and automated reporting systems also seem to have a stronger internal view of risk—a reporting discipline that helps with the investment decision making process as well. In general, those firms seeing regulatory reporting as an investment in infrastructure (as opposed to a compliance cost to be reduced/eliminated) are more analytically equipped.
The difference between institutional and non-institutional e-services platforms
A large commercial market exists for financial-analysis software. There are platforms designed for retail investing with the naming of its brand over top of that branding. Mid-market solutions exist that have actual analytical functionality, but an architecture that won't scale to the amount of data, users or level of reporting required by a large investment business. But, there are indeed institutional class systems, which - though not as many as the market would have you believe - have been designed “from the bottom-up” with the particular needs of complex multi-strategy multi-jurisdiction portfolios in mind.
The differentiating factor of that level is usually not the most glitzy. That's not because of the quality of user interface or the wide array of pre-made chart library. Those factors are the dependability and speed of the information infrastructure at peak load, the ability of the scenario modeling engine to adapt to non-standard questions for analysis, the robustness of the technical setup to integrate with existing systems and information, and the robustness of any support for when that existing information system does go down during a fluttery period – it will!
As institutions look at platforms that can be sorted in this category, detailed technical questions should be asked on data structure, stability of API, disaster recovery plan, support team etc. They should be doing their own calculations on their portfolios in parallel to the model results to verify the risk results obtained by model systems prior to migration work. Also they should consider this assessment as more of a multi-months rather than an exercise based on procurement in weeks.
The platforms you can build that will be a little less flashy in a sales presentation, are the ones you should deploy for accuracy and resilience. That's the trade-off that you want in risk management.
Creating an entrepreneurial expertise to be successful on platforms.
Lastly, one aspect that often does not get sufficient consideration, when assessing platforms, is that the technology alone is not the only thing that matters – it's the team behind it. Just like the team of Piki Templates create such a professional themes like Adsense friendly blogger templates.
It can be particularly important as it relates to scenario modeling and alternative data integration, situations where the practitioners must draw insight from the economy underlying the variables modeled, with many more coming to fruition in 2018. A team that is familiar with building its own sovereign stress scenario, understanding how a sovereign stress scenario is going to be transmitted across asset classes, as well as how to calibrate its own estimate of probabilities based on a reference scenario of the state of the world today will make a lot more out of a sophisticated platform than a team that sees the scenario tools as a box to tick.
It's important to note that investing in investment analytics platforms and investing in investment talent go hand-in-hand, not at odds. Typically, the Universities and institutions that have undergone an institution-wide platform upgrade and have not necessarily invested in the required skills—quantitative and risk—are “technologically advanced, but operationally underused,” with the capability delivering a sophisticated report without creating a meaningful link between the report and operational decisions on how positions are sized and managed.
The companies that have got this right are those that developed the platform capability and human capability side-by-side and engaged risk practitioners in the platform design up front instead of delivering them a "finished" product. Finally, that's what set apart risk management that's functional from risk management that appears functional.
