-
-
Marketplace
-
Channel Resources
Articles from this Site
-
Bringing Sensitivity Analysis to BI
It's a concept so simple that you probably learned about it in college, but so complex that corporations and governments employ teams of economists to grapple with it. The concept is sensitivity analysis, or the process of analyzing future results under differing theoretical conditions.
If you slept through those college lectures, imagine that your company is considering introducing a new breakfast cereal. You conduct a conventional cost/benefit analysis considering the price of flour, sugar, and other ingredients; production costs; and marketing and distribution costs. The results show a small but respectable profit on each box.
Should you go ahead with this new cereal? Perhaps there are concerns that sugar prices might skyrocket. With sensitivity analysis capabilities, you'd plug in one or many alternate scenarios for future sugar prices, and determine under which conditions your new cereal would still be profitable. Of course, sensitivity analysis can't predict what the actual sugar prices will be; that's your judgment call. But it can tell you whether your new cereal will be profitable no matter what, or whether even a modest movement in sugar price will ruin everything.
Sensitivity analysis is tricky stuff in any scenario, and even more so when applied to BI. Yet when we looked at the decision support capabilities that sensitivity analysis could provide, we saw a real business opportunity. So too did our internal business partners in marketing, risk, sales, and finance. They've collaborated closely with us on BI from our first tentative data marts in 2003 to the complex, multi-variant data warehouse that helps drive our global business today. For sensitivity analysis, our internal partners helped us to pinpoint the real business opportunities, and deliver a valuable new component of BI.
Background: Our Business and BI
We're GE Real Estate, one of the world's leading resources for commercial real estate capital, with over $34 billion in assets and more than 35 offices throughout North America, Europe, and Asia/Pacific. A large part of that $34 billion is loans made on office buildings, shopping centers, warehouses, and multi-family housing, or sometimes even on hotels, golf courses, and parking garages.
One of the biggest users of BI has been our Risk Management organization, a global team that analyzes the credit worthiness of every loan request. The risk calculations include looking at the borrower's finances and those of the property itself; the latter is particularly important throughout the commercial real estate industry. As a simple example, consider a loan proposal for hotel in Hawaii. The borrower's credit is good, and our risk team's detailed on-site property evaluation (yes, it's a tough job, but someone's got to do it) determines that the hotel is worth $70mm. The borrower is asking for $55mm in financing; should we go ahead?
Since its inception, our BI system has played a major role in those decisions, providing detailed performance trend data on similar loans in our portfolio. Through BI, our risk managers can zero in on the performance of our loans in a single market, sub-market, or group of markets, targeting year of loan origination, fixed vs floating rate, loan-to-value or debt-service-coverage, property quality, property type, and over a dozen other variables. And it's not just our risk management team who benefits from this data. Our sales executives, who receive more and more loan requests every year, use BI to help determine which requests are most likely to be approved; our finance managers use it to predict future loan earnings; and our asset managers use it as an early-warning signal for possible changes in loan performance. Even our global markets development team soon be using BI to determine which new countries would benefit most from our products and services. Combining global market data such as market size, liquidity, and volatility with our own internal information, they'll be able to make better decisions on what areas of the world to focus our expansion.
Bringing Sensitivity Analysis to our BI
We've all seen how business users come to appreciate the power of BI, and then begin to visualize opportunities to expand its scope. That's what happened with us when sensitivity analysis was first proposed. Business support was strong, and we formed a dedicated project team that investigated and debated many different concepts priorities for it. Should our sensitivity analysis focus on the impact of employment changes on office building profitability? Exchange rate fluctuations on demand for manufacturing facilities? Gasoline price increases on parking garage revenue?
What settled the discussion was the steady rise in the benchmark Fed Funds interest rates by the Federal Reserve. These increases, which began at the start of '05 and continued uninterrupted into '06, prompted a lot of questions about the potential impact on our floating rate loan portfolio. A borrower with a loan pegged at 1.5 percent above the Fed Funds rate may have had no repayment concerns when the Fed Funds rate was 1 percent... but what about 4 percent or 5 percent? Our BI platform was already providing retrospective insight on impact of the ongoing rate increases; now, we would look ahead and estimate future performance based on varying scenarios of future interest rate increases or decreases.
Using a complex weighted mix of historical performance data from our own databases and from industry-wide data aggregators, loan-specific variables and future probability models developed by outside econometric consulting firms, we then identified several key indicators that would be particularly appropriate to our interest-rate based Sensitivity Analysis. In our case, they are short-term interest rates, long term interest rates, discount rates (a net present value industry term), Cap rates (another industry term capturing the net operating income as a percent of a building's market value). For some indicators, we developed the ability to provide analytics under every scenario in quarter-percent increments; for others, we use real time consensus ranges updated regularly by our econometric consulting firms.
Our business now uses sensitivity analysis capabilities to project ahead to any quarter this year or through the end of '09. They can examine our entire loan portfolio, or look at a single continent, country, state or metropolitan area; for example, they can compare the impact of an increase in the short term interest rate forward yield curve on our loans in Miami vs those in Orlando. They can home in on one, or several, of over 70 specific business segments, such as Affordable Housing, Manufactured Housing or Vacation Ownership. They can look at all deals with a single customer, or at a specific property. And they can even aggregate by markets considered low risk, medium risk, high risk or recovery.
The impact of sensitivity analysis on our business has been revolutionary. Our risk managers now analyze loan proposals not just by expected outcome under current scenarios, but under a broad range of future scenarios. Our finance, marketing and sales teams use inter-market, multi-scenario comparisons to determine the biggest market growth opportunities as well as those where caution is advised. And our asset managers, whose role is to work with the property managers at the buildings secured by our major loans, have an early warning system that lets them focus on the areas of potential greatest risk.
Best Practices
We've learned a lot through our successful deployment of sensitivity analysis within our BI platform. If you're considering a similar initiative, consider the following:
Partner with business leaders: As obvious as it sounds, there are still IT professionals who fall in love with a concept - and sensitivity analysis is a very exciting one. However, if there aren't business leaders clamoring for this, you risk ending up with expensive shelfware. At GE Real Estate, we take the business partnership seriously: it's not enough for business leaders to endorse our technology initiatives, they also need to fund them. Our BI platform, like all our IT initiatives, competes for funding that could otherwise go for marketing campaigns, additional hires, you name it. This ensures that only work that is truly meaningful for the business, which in our case includes sensitivity analysis, goes forward.
Don't build sensitivity analysis in an unproven BI system: Building even simple BI capabilities can be extremely difficult, as articles in this publication regularly attest. Sensitivity analysis brings with it significant additional complexity. Only after our BI system had been live for several years, and successfully integrated into a broad range of business processes, did we proceed with sensitivity analysis. By then, we had a strong understanding of what our users value in BI, and were able to apply that to our sensitivity analysis development efforts. Doing otherwise means risking the whole initiative.
Focus sensitivity analysis on truly meaningful data: As noted above, we looked at leveraging many types of underlying data, from employment growth to property tax trends to commercial construction permit issuance. Some of these would have been quicker to develop and easier to integrate, but did not carry the business value of the interest rate data that we chose. Utilizing this data allowed us to integrate a custom forward yield curve model, significantly enhancing our forecasting accuracy.
Do not boil the ocean: Break your projects into small, manageable components. While our choices lengthened the development time, having a robust, well-regarded BI application gave us the credibility to take longer and invest more. And there's always time for additional capabilities in the future.
The Future
Over the upcoming years, we'll continue to increase our scope of sensitivity analysis. Some of our planned features, like incorporating Monte Carlo simulations, will bring even greater sophistication and capability. Others are more straightforward, such as developing industry-specific and company-specific risk metrics sensitivities for major tenants in our office buildings.
At the moment, we're expanding our sensitivity analysis to include future exchange rate fluctuations for many of the currencies in which our loans are denominated, including the US dollar, British pound, Swiss franc, Euro, Japanese yen, and Canadian dollar. And while our risk and finance partners haven't seen the value of including the Tahitian franc in these currency analytics, we believe that it offers an untapped business opportunity - but to be sure, we'll be traveling there to conduct an extensive, onsite requirements analysis.
Hank Zupnick is CIO, and Jeff Rademaekers is VP & CTO, at GE Real Estate (GERE). GERE's Business Intelligence system has recently received an Edison Award, one of a very small number awarded annually within GE for outstanding technology contributions to business growth.
Hank Zupnick is CIO at GE Real Estate (GERE). GERE's Business Intelligence System has recently received an Edison Award, one of a very small number awarded annually within GE for outstanding technology contributions to business growth.
Jeff Rademaekers is VP and CTO at GE Real Estate (GERE). GERE's Business Intelligence System has recently received an Edison Award, one of a very small number awarded annually within GE for outstanding technology contributions to business growth.
For more information on related topics, visit the following channels:
