Alternative data: Lessons from the world’s top investors…are you asking the right questions?
Over the past few years, alternative data has become ubiquitous in the investment world. As head of Business Development for DataMonster, I’ve been fortunate to meet with hundreds of institutional fund managers (and increasingly corporates)and to learn how they are integrating data into their fundamental process. While we’re still very early in the adoption phase, I’ve seen commonalities in the approach of those firms that are finding the most success. Conversely, many who are struggling often run into the same pitfalls.
Successful funds in this process have made a long-term commitment to primary research. They’ve set realistic expectations around the time and cost required to scale their efforts. They have a centralized process that fits the needs of different types of users. Finally, they use data to answer questions around directionality of investment themes, not solely trying to “call the quarter”.
Here are some helpful starting points for those who are curious about what “the other guys” are doing right. (For starters, they’ve dropped “alternative” and just call it “data”).
Build a Central Process
I’ve seen a wide spectrum of “commitment to data” across the street. In some cases, there is a top-down push from the C-suite without much end-user buy-in. In other cases, a single portfolio manger or analyst struggles to convince management of data’s merits. Successful funds have a firm wide commitment to imbed data in their primary research process. Data isn’t viewed as a short-term experiment that is validated by an insignificant sample of successful outcomes (or often vice versa).
On the back of that commitment, they’ve built a central data team (or sometimes a “data champion”) that serves as a resource for the broader investment team. This team is highly valued and not just seen as a glorified IT resource. They are a part of the investment committee and crucial to the overall investment process. Much like “have we talked to management?” or “have we spoken to experts?”, “what is the data telling us?” is now an essential question for any thesis.
The central data team must be independent. The last thing you want is a data scientist that is motivated to find conclusions that agree with what a portfolio manager wants to hear. The risk of cognitive bias is increased when you have a data scientist embedded within a portfolio team. These risks can be mitigated by ensuring the data scientist focuses on the thesis question itself, not the desired direction of the underlying investment.
Ask the right questions.
Data is commonly (and mistakenly) seen as a short-term tool focused on “calling the quarter”. Funds that are further down the journey have moved past this initial use case toward more nuanced questions around their theses. Savvy fund managers realize that there is greater edge when they use data to understand the direction of specific narratives. They are not focusing their efforts on trying to nail the quarter with 99.9% accuracy. Investors that have experience working with data have realized that predicting a point estimate with alternative data is extremely challenging, bordering on unrealistic for a variety of statistical realities. Getting a feel for the direction of an investment narrative/question is MUCH more achievable.
Successful investors have differentiated, long-term views with a few core tenants which they feel are misunderstood by the market. If those tenants hold over the investment horizon, the investment has a good chance of working. These investment tenants can easily be turned into a few big questions, which can then be broken down into a series of smaller questions. These smaller questions tend to be more quantifiable and are good targets for primary research (data and/or surveys). For those interested, I’d highly recommend the book “Super Forecasters” by Phil Tetlock. While not focused on data, the techniques and theories around good forecasting can be well adapted to the world of data and fundamental investing.
See below for an example of what I mean on what is by now a highly consensual narrative:
As you can see, by taking a big picture question and breaking it into smaller and smaller questions, you arrive at questions that are much more quantifiable. I can think of ways to quantify every one of the smaller questions on the right side of this illustration by using alternative data.
Build a Mosaic
Data is best used as a part of a mosaic. This is relevant for data sourcing, as well as the investment process. There are no silver bullets or magical data sets. Successful funds look to layer multiple data sets to find situations when all or most of the data is pointing in the same direction around a thesis. From there, you can look for confirmation across other research methods — i.e. expert network calls, surveys or calls with management — and be armed with data that allows you to ask the right questions. When all aspects of this research process align, you can make the big call.
Expectations need to be realistic in terms of cost and time commitment. Data can certainly be expensive. However, even at the largest and most advanced funds, a data budget is often a fraction of the overall research budget. Large funds have been paying traditional research firms tens of millions of dollars a year with very little ability to quantify the associated alpha. The same holds true for time spent across traditional channel checks, expert networks, etc.
Data, on the other hand, is more heavily scrutinized since it’s much easier to quantify its efficacy. Quantification is very important and useful when done over the appropriate time horizon. In other words, funds must look to evaluate ALL of their research inputs, but the time horizon must be sufficiently long to get an accurate view (not, “I tried the data a few times, it didn’t work”). In evaluating the results, a sense of intellectual honesty must also be applied in comparing these data led research results to traditional research methods. Data isn’t no silver bullet, but the approach is likely to get you much more useful insights than the old scatter gun approach of channel checks you have performed historically.
Despite all of the hype, it’s clear that we are still very early in the adoption phase of alternative data. As we are now in a much more volatile phase of the investment cycle, the pendulum of advantage should swing back towards active managers from passive. The most successful of this new breed of active managers will inevitably incorporate alternative data into their process. The operative word being “incorporate”, not supplant. This will be an incredibly exciting time for the funds that get it right.