Monetizing bank’s data: An example

Background

The Federal Reserve Board (Feds) and other regulators conduct supervisory hypothetical scenarios based simulation exercise (called stress tests) to help ensure that large bank holding companies operating in the United States will be able to lend to households and businesses even in a severe recession. The Feds tests are known as the Dodd-Frank Act stress test (DFAST) and the Comprehensive Capital Analysis and Review (CCAR).

The Fed hypothetical scenarios are categorized as baseline (most likely), adverse (moderate recession), and severely adverse (great recession). Banks in addition to these scenarios create multiple other hypothetical scenarios of their own. This help banks, estimate projected revenues, losses, reserves, and pro forma capital levels.

Each hypothetical scenario comprise of hypothetical affect on numerous macroeconomic variables (MEV), forecasted for several quarters in future, using sophisticated modeling techniques and taking in account correlation among them. Some big banks do projection for over 500 MEVs. Some examples of MEVs are:

• Economic activity and prices : Real and Nominal GDP, Unemployment rate, real and nominal disposable personal income and consumer price index (CPI)
• Aggregate measures of asset prices: House prices, commercial real estate prices, equity prices and U.S. stock market volatility.
• Measure of interest rates: 3-month T-Bills, 5 and 10 year Treasury Notes, 10 year BBB corporate, 30 year mortgage rates and prime rate.
• Major economies variables: Real GDP, CPI and USD exchange rate.

Why forecasting MEVs is important and who might use it ?

MEVs drive the business from a systematic perspective and help one to understand their impact on individual businesses. Some of the use cases for the use of MEVs are:

Asset Managers:

• Devising asset allocation strategies
• Managing long-term risk
• Modeling expected cash flows and discount rates

Banking and Insurance :

• Conducting regulatory exercises like CCAR/DFAST, ICAAP, CECL etc.
• Projecting revenue and expenses
• Asset liability management

Government

• Formulating appropriate polices (Monetary, Fiscal, Industrial etc). It also provides insights into the intricate relationship among factors like CPI, Income output and employment.

Financial Planning & Analysis (FP&A)

• Developing multi-year Financial Plan that directly links into the annual Corporate Plan

Corporate Strategy

• Identify opportunities to expand business or reduce costs

Note:

• MEVs in itself are not sufficient to make financial or risk forecast. They need to be modeled with company specific business drivers and assumptions. For example, Auto sales are not only affected by MEVs but also by any campaign or promotions one plan to undertake.

How do we monetize MEV forecasts?

Forecasting MEVs is a complex process. Banks have been investing heavily in this area, mainly to meet regulatory requirements. They have built army of Economist, Statisticians and Programmers, who collectively build sophisticated forecasting models. In addition, massive amount of investment have been put into to building modern and sophisticated infrastructure to support modeling process and conduct simulations like Monte Carlo simulation. Regulators have now high degree of confidence in the big banks’ end to end process. The processes are very well controlled , with proper oversight, audit and governance. Finally, the MEV forecasting models are constantly being back-tested and refined, while also leveraging innovation in this area.

This has allowed banks to not only produce high quality forecast but also to quickly turnaround on new scenarios. For example, during initial stages of Covid-19, banks within matter of couple of weeks were not only able to generate multiple new scenarios but also able to assess impact on their financial health. This allowed them to build right level of reserves to withstand losses.

Good MEV forecasts can be a source for competitive advantage for any company. I believe, there is an opportunity for the banks to monetize this. They can do so by selling forecast data and related services to other companies who cannot produce on their own. Our initial prime target could be the companies, who already buy this data from a market data providers. The data they get is non-customized, not as comprehensive (i.e. number of MEVs covered) and available only for limited standard scenarios.

Product Offerings

Banks can provide four level of product offerings

1. Raw and/or derived data for standard scenarios

a. Raw MEV forecast for predefined scenarios.

b. Additional curation services to re-format data , apply business rules and/or resolve data quality issues.

c. The data can be delivered via secure APIs or as feeds.

2. Scenario design

a. Ability for clients to design their own custom scenarios.

b. Since banks’ proprietary model is aware of the intricate relationship among various MEVs, the existing models should be able to spit out the forecast for other MEVs as well.

c. This will also allow clients to conduct What-If simulation exercises.

3. Modeling software as a service.

a. Some big banks are in position to offer their modelling platform like a Saas (software as service) provider on cloud. The software they have created caters to multiple lines of diverse businesses, which now can be offered to external clients as well.

b. The external clients will have forecasted MEV data available and they can also supply their own data in a secure environment. This will allow them to build sophisticated projection models using programming languages like Python and R.

c. It also provides opportunity for the banks to provide professional services to build these models , if clients do not have resources to do so.

My this post describes how we can build a such a SaaS platform

4. Analytics

a. The banks can also provide additional services to visualize and analyze the data. For example, to provide visual insights into the major drivers for the change in revenue between two chosen scenarios.

How clients can be charged?

• Product offerings
• Number of MEV subscription
• Number of models coded
• MB/Day usage of data
• Processing power (i.e. CPU and Memory usage)
• Concurrent users
• Online monitoring
• Support time response