Digital Technology
Jul 11, 2018

The time is now to scale fund-of-funds business using artificial intelligence

A fund-of-funds firm operates at the intersection of investors, fund managers, and wealth advisors to provide strategic value, such as asset allocation, manager selection, portfolio oversight, and consolidated reporting.

While the front end of the fund-of-funds business is big dollar allocations, famed managers, and well-heeled investors, the back end is often not as glamorous. In most cases, middle- and back-office functions remain manual, slow, error-riddled, and risk-prone. The dependence on manual processes impedes the ability to scale efficiently. (Astute observers will agree that similar problems are prevalent among institutional investors, sovereign wealth funds, and private equity firms.)

Throughout the alternative investments value chain, there is a vast amount of unstructured data and documents floating around. Many of these documents and the data contained therein are extremely valuable pieces of information. So, you can imagine the magnitude of the problem involved in updating a diverse set of documents in distinct formats at unique frequencies. In such cases, dealing manually with unstructured data clearly hampers the ability to optimize operations and scale efficiently.

Let's look at a few examples.

Manager selection
There are over 15,000 hedge funds, as well as about 3,500 private equity firms globally. This makes the manager due diligence and selection process both onerous and critically important.

While some data is structured, such as filings to regulatory bodies like the Securities and Exchange Commission (SEC), much of the other data is often private, privileged, unstructured, and unavailable.

Imagine the challenge of digitizing and gleaning insights from thousands of documents and normalizing that data to facilitate manager research, selection, due diligence, and onboarding.

Even answering typical questions like the following takes time and much browsing through various documents:

  • What's the domicile? The Cayman or the Channel Islands?
  • What's the gate? 10% or 25%?
  • Does it have a side pocket?
  • What's the hurdle rate?
  • What's the high watermark?

Ongoing monitoring
Unlike mutual funds or exchange-traded fund (ETFs), ongoing monitoring of hedge funds is also not easy – one can't look up the latest net asset value (NAV) in a business journal or a financial website.

The investor letter, often a window into the mind of the manager, comes in all flavors. Some are crisp and quantitative, others are verbose and poetic – all, in their way, are informative and insightful.

Some bits of information, such as a change in a manager or sizeable redemptions or a large, risky bet, may have a significant impact on the viability of the strategy and the allocation to the manager.

But keeping tabs on factors, such as style drift, performance volatility, manager attribution, and market correlation, is difficult. Plus, in the absence of easily obtainable data, the problem compounds exponentially.

Consolidated reporting
Today's investors demand transparency, depth, and detail, in addition to a coherent and consumable synopsis of the investment performance. Many large family offices and institutional investors require specific formats and particular performance details based on an investment policy statement (IPS).

For firms, the need to meet the custom reporting requirements of investors is further complicated by the need to combine performance statements from different managers spanning diverging asset classes and varied valuation methods.

The foundation for a good custom reporting solution includes integration with regulatory databases for downloading periodic filings, APIs with custodians for transaction, price and position information, and ingestion of unstructured data, such as PDF statements from managers.

In addition to the aggregation, synthesis, and normalization of the data, downstream reporting also requires views that comprise portfolio decomposition, various performance measures, and sophisticated risk analytics.

Artificial intelligence and machine learning to the rescue

The information necessary for various fund-of-funds operations is spread across and often hidden in a high volume of documents. Each of these documents has a distinct format, structure, and style. The range of formation documents includes: Private Placement Memorandums (PPMs), Investment Management Agreements (IMAs), Form 13F, performance statements, investor letters, and so on.

Previously, in addition to buying a database of fund manager information, a team of junior analysts (or a team overseas) entered much of this unstructured information manually for downstream uses, and sometimes that meant entering the same information in more than one system, on account of application data silos. These teams also had to update the information monthly and quarterly to include the latest valuation and position or holdings information.

Today, traditional machine learning techniques can be applied to extract information from some documents, provided all fund managers are following a standard format. But this approach requires a large body of training data to achieve expected accuracy levels, and, while this approach may allow the fund-of-funds to take steps in the right direction, it is not scalable or optimal.

Genpact Cora LiveWealth approaches the same problem using a patented method that involves the disambiguation of information using computational linguistics and a semantic network. These techniques allow our neural network to grasp nuances of the text and extract meaning as well as actionable insights. These advanced methods reduce the need for a voluminous set of training data and eliminate the need to modify algorithms based on minor changes in form, format, structure, or content of the documents, opening up the opportunity to optimize operations and scale the fund-of-funds business.

About the author

Satya Iluri

Satya Iluri

Vice President, Wealth Management Consulting

Follow Satya Iluri on LinkedIn