Spanning the analytics spectrum: How financial services firms continue to evolve from reports to self-serve analytics

By: Venkat Mullur

The transformation of data into information should be seen as a continuum, from static reporting (which still has its place in most businesses), dashboards (mainstay in managing repeatable operations), to more sophisticated self-service analytics tools that answer forward-looking questions.

Timely and accurate data has always been critical to financial operations such as trading, lending, hedging, and disclosure. Over time, the types of questions business users have come to ask of data have changed, and business software has evolved accordingly. Timely insights are now critical to a sustainable competitive advantage.

The first generation of data queries sought to gain an accurate picture of the business, mostly from a historical perspective - "...which region performed best in Q4?", or "...what was the revenue from credit cards, for each month in the preceding 12 months". Accurate source systems and robust databases capable of returning results fast was the need of the hour.  The software industry responded and we saw a boom, from about 1980 to 1998, in enterprise databases and reporting systems.

The next generation of data queries involved more sophisticated questions that sought to glean second-order insights from data. Financial institutions, faced with increasing competition caused by deregulation, were seeking new ways to be profitable independent of business cycles.  Business users were forced to look at performance metrics both temporally and across other classifications. So questions often took the form of "...what are my top three products, both across a line of business and across geographies..." , "...what was the daily 200-day moving average for treasury yields over the past two years, and on what days did yields fall 5% below and rise 5% above that average...". As a result, the boom in reporting software soon gave way to business intelligence software, or BI, characterized by dashboards, pivot tables, and multidimensional data cubes.

But despite the massive investment in business intelligence software since 1997, financial services firms have continued to rely on after-mart approaches to extracting more insights out of data. The most famous of these approaches involves the pervasive use of spreadsheets to conduct ad-hoc analysis, perform statistical transformations, and study the effect of core assumptions - involving such metrics as interest rates, inflation rate, or correlations - on entire portfolios of assets. While this is not the article to discuss the pros and cons of using spreadsheets, it is fair to say that it takes significant skill and persistence to extract critical third-order insights buried in spreadsheet data.

Optimal decision-making hinges on a thorough understanding of underlying causes behind observed metrics, understanding relationships between performance measures and cost drivers, visualizing risk drivers and their effect on measures, and gaining sufficient confidence to assign causality. Such questions have applications in risk management, profitability analysis, and most importantly, in shaping future investments and new growth strategies. Examples are "...what would demand be for our products in the coming quarters or years?"; "...will our insurance premiums be sufficient to pay out claims over a period of five years?"; "...who is most likely to respond to credit card offers?"

 While high-end statistical and optimization software do exist to answer such questions, the specialized skills needed are often beyond the average business user.

This situation is untenable because the chasm between those that provide the answers (modelers) and those that act on the answers often results in sub-optimal, or even wrong, decisions. As an example, absent knowledge of stochastic distributions, it is hard to intuit that VaR at 99.99% confidence level can be five times the VaR at a 95% confidence level!  Understanding this is crucial to an accurate estimation of economic capital at a bank!
Analytics tools that interpret complex quantitative measures and present the ramifications to end-users in an easy to interpret format represent the next generation of business intelligence tools.  The separation between reports, dashboards, and analysis is fast blurring, and we are now witnessing the first generation of tools that span the analytics spectrum and still appeal to the broadest range of business users. Let us consider a specific example from the banking world.

In banking, tracking loan performance - be it mortgages or revolving lines of credit - is a core business function. Loan performance is both a key performance metric and a risk metric, one that has immediate effects on loan origination, internal rating systems, P&L analysis, AML, risk management, and capital deployment. However, at most banks, each of the abovementioned functions is handled by a different system and to get an answer to questions like the ones in the picture below the business-line manager must request answers from the risk management system, the AML system, the econometric modeling system, and finally the reporting system.
 
By the time, the user gets the answer, and gets it in a way that is actionable, market conditions might have changed materially so as to render the answer obsolete for decision-making. As seen in the picture above, the lack of an analytic layer forces the business user to rely on a multitude of systems and system owners.

Upon further reflection, one can easily surmise that the only interface a business user really cares about is a portal that helps pose relevant questions and receive the answers. This analytical layer ought to be tailored to the type of user. It must, without major architectural reconfigurations, be able to provision reports, dashboards, or interactive screens capable of answering ad-hoc questions. The figure below illustrates exactly such an interface.
 
To conclude, financial services firms are facing an increasingly competitive landscape, and the pressure on management to grow, while tightly managing risks, is tremendous. Analytics ought to enable decision-making, and must be positioned to appeal to a wider range of business users, and not just those with "analyst" or "quant" in their job titles. The analytics spectrum discussed here is a culmination of many generations of software evolution, and is now poised to re-set user expectations and change the way financial firms use business intelligence software.