You May Beat an Algorithm Today. What about Tomorrow?
Technology will upend the way investment decisions are made.
How should investment decision makers respond?
This conversation has a tendency to assume an emotional tone even before it gets existential. For instance, you hear genuine angst over Microsoft Office’s new layout, and that’s just a redesigned application. Innovation has cumulative effects though, and after more than 50 years of Moore’s law and numerous other advances, algorithmic decision making fits too tightly into investment processes to be ignored.
That doesn’t mean it’s perfect. One purpose of this essay is to invite you into friendly competition with an algorithm designed by Ashby Monk and his colleagues at Stanford University. Their process can select managers with limited information and little time. You may be able to predict a team’s success more effectively.
Whether or not you feel like you beat the machine today, it’s worthwhile to have a plan for what happens next so that your organization can fully capitalize on its native strengths.
In our imagination, the word “algorithm” tends to evoke fast-paced and high-stakes processes.
The mundane truth is that the word refers only to a process or set of rules that describe how to accomplish a task. There is no naturally associated time horizon or risk appetite.
And that’s good, because most public pension funds, endowments, and sovereign wealth funds wouldn’t count speed or agility among their strengths.
Our goal here is to underscore that they don’t need to, and explore what a technological transformation looks like when its primary purpose is to reinforce risk-aware patience at institutions that may outlast many of their assets.
Know Who You Are
Many organizations have no formalized models or systems for characterizing data or judging data quality.
No matter. Ongoing competition among market participants to exploit new forms of information and data squarely qualifies as an arms race, definable as a situation in which parties are locked in perpetual efforts to outcompete one another without a defined endpoint. And it gets better: You can’t escape. It’s difficult enough to achieve most current risk and performance targets as things are, but alternative datasets and associated analytical techniques are already augmenting the pricing process in many markets.
Transacting in markets you don’t understand is not a recommended practice. The fear is that it may become business as usual for many investment organizations unless they develop an internal capability to assess and integrate alternative data. As we all know: Fear leads to anger and is the path to the dark side.
One manifestation of this is when entities “leap before looking” and obtain datasets without considering the actionability of what they’ve bought. This can lead to efforts that are poorly aligned with organizational capabilities or priorities and offer little long-term value. The wrong purchase can also drive pursuit of shorter payback periods to offset their costs, and so compress the time horizons of decisions made with them.
Seek Defensive, Defensible Value
Accessing novel data should not be a goal in its own right. The idea is to manifest the best possible version of your organization.
Many large investment organizations struggle with innovation for the simple fact that they lack internal agreement about what direction they should go in. The allure of working with alternative data could make it a common point of agreement in the process of building support for new initiatives internally, and the learning from efforts to integrate it can drive significant future innovation.
The work isn’t necessarily all that different from what you do already. For instance, consider:
- Assessing the strength of a potential general partner’s professional and social network with LinkedIn data and news reports of prior deals before committing to their fund. An allocator is likely to have little ex ante clarity about the specific start-up companies in which a venture capitalist will invest, and no control over how it does so once capital is pledged. The quality of the venture capitalist’s likely co-investors, however, may be easier to discern and serve as an indicator of the ultimate riskiness of its portfolio.
- Conducting due diligence on candidate direct investments in leisure-related properties — say, hotels or casinos — by assembling online price and ratings histories of possible competitors — think Airbnb, TripAdvisor, or Yelp — or price-series of airfares to that locale.
- Controlling reputational risk from investee companies by monitoring controversies about them that arise in social media posts or other localized/unconventional news outlets.
The precise tools and techniques of executing those research operations may not yet be clear, but hopefully the practice seems less like a dark art and more like defense.
There’s still a place for present value, probability, and the rest of the body of knowledge that characterizes the investment profession. Alternative data just adds new colors and textures to the mosaic of information investors have long been evaluating.
It also argues for a degree of operational openness.
Some of the most compelling implementations of alternative data strategies involve the investment organization itself. Here are a few examples:
- Ensure your cost structure is not a simple function of market capitalization by continuously monitoring contracts with external asset managers. Fee structures based on assets under management (AUM) can creep into costs because growth in an account balance does not necessarily reflect manager skill.
- Inventive collation and synthesis of documents like e-mails, investment memos, and contracts can uncover precious metadata with insight into communication, culture, negotiation, time allocation, benchmarking, and diligence.
- Organizations can map their internal knowledge flows by tracking how internal users query and access documents in organizational databases. This also allows for examinations of typical approaches analysts use in problem solving. More granular visibility into these activities can expose not only areas for improvement, but also help better identify best practices.
Everyone faces technological competition.
Strategically, this means that investment organizations can invest in homegrown sources of technological leverage, and they ought to. If they don’t, the decision should come after a thorough analysis of long-term tradeoffs to the organization.
From these seats, it’s hard to imagine a winning case against it. Data is perhaps the most important input into every investment decision-making process already, and new applications for it are proliferating every day. Unforeseen (and sometimes unidentifiable) risk still hits portfolios frequently. Unrealized efficiencies abound.
You really should try to outperform the algorithm, but here is some further reading to help you seize the opportunity:
- “The Investment Firm of the Future: Alternative Business Models and Strategies for a More Forward-Thinking Industry” (CFA Institute)
- “Rethinking Alternative Data in Institutional Investment” (SSRN)
- “Ambiguity Tolerance Beats Artificial Intelligence” (Enterprising Investor)
- “The Technological Investor: Deeper Innovation through Reorientation” (SSRN)
- “Artificial Intelligence, Machine Learning, and Deep Learning: A Primer” (Enterprising Investor)
If you liked this post, don’t forget to subscribe to the Enterprising Investor.
All posts are the opinion of the author. As such, they should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute or the author’s employer.
Image credit: ©Getty Images/monsitj
5 thoughts on “You May Beat an Algorithm Today. What about Tomorrow?”
Really great question and survey Sloane and Ashby – can we republish with attribution and link in our Global Corporate Venturing trade paper?
Thanks for your interest! Sending you an email.
AI as a substitute for personal responsibility is a slippery slope. As a student of karma in financial matters, I can only assume that turning it over to an algorithm will have different results for different client/customers even if the computer(s) have similar inputs. The idea that this or, for example, indexing, will turn out the same for all involved will be tested in the near future, and the explanations of managers will be great reading.