As part of my year-end reflection and 2010 road map, I have been thinking deeply about how exactly I invest. How do I pick the companies and sectors on which to focus? I quickly realized that the principles I applied in the math of derivatives are the same ones I use as a venture investor. In short, I look at everything as a limit: "What does this (company/sector) look like at infinity?" I see most progress as being asymptotic, e.g., there is a period of accelerated development and rapid growth, beyond which gains are much harder and more costly to achieve. The goal is to find businesses at the early part of the rapid growth phase, but where they are close enough to rapid growth that they don't require decades of funding to get there. For instance, if you had invested in machine learning in the 1960s (or natural language processing in the 1980s) back when it held so much promise, you'd have gone bust many times over waiting for commercial success. Does it mean that machine learning and NLP are unattractive fields? No. But if you were 20-40 years early then I'm sure it seemed that way. My goal as an investor is to avoid such delayed gratification.
While other investors invariably have their own language to describe this exercise, I find calculus to be a helpful tool for testing one's assumptions around a particular investment and for providing valuable discipline to avoid "style drift." Though I make the occasional investment to learn about an interesting business or to work with a particular group of investors, my investing is largely characterized by the discipline noted above.
From a process perspective, this generally means that I have a vision of the future "at infinity" and work backwards to identify potential investment candidates. The key for me is once a candidate is identified, are they close enough to the rapid growth phase and is the market sufficiently ready for this growth to take place? How many times have you heard an entrepreneur say "We were just too early." The road is littered with great ideas whose time had not yet come from a commercial perspective. It is this timing issue where I spend a tremendous amount of time on due diligence, reaching out to industry contacts and testing their receptivity (and willingness to buy) the product/technology in question. Does this approach mean I'll miss some huge ideas that were simply on nobody's radar screen? Sure. But is it a more risk controlled way of reaching for big gains? I think so.
So given this approach, what are some of my "visions for the future?"
1. Machine-driven trading will continue to proliferate, and represent a sustained source of alpha.
If my thesis is right, the most attractive opportunities will exist in the following areas:
- Alternative data. The key question is whether it is more valuable as a distributed vendor product or as a closely-held proprietary product that is traded.
- Modeling platforms. Most platforms today have the ability to ingest structured, quantitative data. Future platforms will be able to consume and model both structured and unstructured data, and to mash up disparate data sources with a limitless number of systematic trading models. They will also provide for straight-through processing, generating trading instructions and executing trades directly from the modeling platform.
- Database architecture. The relational database of today will be inadequate to process the massive amounts of unstructured data - in real time - that tomorrow's (and, in fact, some of today's) trading models require. This also encompasses distributed and high-performance computing.
- Predictive analytics. Extracting insight from large bodies of textual data will challenge current analytical frameworks. New methodologies will arise to meet this challenge. This also includes event notification and anomaly detection, which has relevance for anti-terrorism and anti-fraud applications as well.
2. Tomorrow's ad exchanges will resemble the stock and options markets for equities.
This has implications for liquidity, price discovery and hedging:
- Aggregated buy- and sell-side ad demand. Fragmented exchanges will be be stitched together, leading to a consolidated view of sell-side inventory and buy-side interest. Price efficiency will skyrocket. Buyers will be able to hedge and speculators will be able to take a view on the direction and volatility of context-specific impressions.
- Consolidated buy-side optimization platforms. Agencies and brands will have access to platforms that integrate disparate forms of data and metadata, exchange prices and enable ad campaigns to be optimized and ROIs to be calculated. They will control the structuring, buying and monitoring of online ad programs.
3. Social media will simply be called "media," and viewed as a fully-integrated part of the overall media buy.
Distribution, monetization and ROI measurement will be key drivers of success:
- Apps that aggregate fragmented audiences across Twitter, Facebook and other social media outlets.
- Platforms that facilitate monetization of these fragmented audiences, and deliver a powerful suite of tools for agencies, brands and content owners to use for advertising, promotion and ROI measurement.
- Tools that enable agencies, brands and content owners to carefully control which ads and promotions are displayed to which audiences, placing reputation protection and control in the hands of those with the brand and relationship equity.