8 min read
Top Misconceptions Money Managers have about Alternative Data: Part One
Perhaps the most common misconception we encounter among managers who are relatively new to the field, is that Alternative Data is only useful for predicting company earnings shortly before they are announced. The idea is that traders will use data to bet against the consensus view, anticipating a surprise in the market and taking quick profits from the subsequent move in the stock price. For instance, let’s say you have data that indicates that Wayfair is growing sales recently and your model implies it will surprise on Revenue causing the markets to react, so you buy W and sell after the market “catches on”. This over-simplification is almost never the reality.
While some traders do use Alternative Data to get a different perspective on an important company KPI ahead of an announcement, the reality is that this data is used for much more than turning a quick profit. To call it the tip of an iceberg would be a big understatement. Alternative Data can add value to almost any aspect of the research process for a long-term investment as opposed to just a trade.
What are some of the most important aspects of a research process? Managers running fundamental strategies perform rigorous analysis on companies: they scrutinize their industries, their products, their competitors, and the behaviors of customers regarding their products and services. Alternative Data can yield critical, long-term, forward-looking insights on all those concepts and more.
For example, managers can use years of historical transactions data to analyze the adoption of smartwatches compared to luxury watch brands. Then they can use a different cut of the data to see the changing income demographics of people purchasing those smartwatches. They can also analyze the brands that are most favored over time and how the market composition has evolved. Alternative Data can tell rich and nuanced stories about public or private companies that support a long-term thesis or view. In fact, in some cases, managers analyze Alternative Data spanning a decade of trends to better gauge the competitive landscape in areas such as fitness, consumer goods, healthcare, and transportation. Often, this data is used to build and substantiate a thesis for a multi-year investment.
Upfront data costs can be tough to swallow for smaller managers, but they would be wise to look at the overall value they are getting from Alternative Data.
There are two main costs to using Alternative Data for investment managers: Data acquisition costs and costs of Data Management & Analysis. It’s true that the upfront costs of buying data sets remain high, but they have stabilized and even come down on average over the years. At the same time, the value managers can get from data has increased as they find more ways to apply the data to research portfolio management.
Although Alternative Data sources are costly relative to traditional market and financial data, the real costs that concern most managers come after the initial outlay: the costs of running an effective Alternative Data operation.
Managers feel that they need to hire droves of data scientists or consultants to onboard, integrate, and cleanse the data, or develop their own data collection infrastructure, or even ask their analysts to sign up for the next available online Data Science course touting Python. Unlike structured market data, Alternative Data is thought to be noisy, biased, and not always relevant to certain investment strategies. The cost of overcoming these challenges may seem daunting to growing managers.
And even after the data has been onboarded, cleansed, and tagged, analyzing it can also be expensive. Fund managers feel they may need to hire additional analysts or build specialized internal tools to process and analyze the data so that they can finally get to the insights locked away in the data.
Again, reality is somewhat different from perception. The costs of running an effective data operation are not as high as you may think if you invest in the right technology, and the value you can get from fully utilizing Alternative Data in your process more than justifies the costs.
What are the managers missing?
In recent years, technology has progressed to the point where organizations can become very efficient with their Alternative Data spend. For instance, knowing which data sources are most relevant to your portfolio and watchlist can save millions by only focusing on the relevant sources.
Furthermore, automation can take the place of large engineering teams focused on resource-intensive staging, mapping, cleansing, integrating, and combining data sets.
Finally, the analysis and reporting layers can also be largely automated through the use of technology such as monitoring and alerting processes that managers can define themselves without writing code or spending millions on teams of developers. Often, a technology partner can surface insights on a portfolio more efficiently than internal staff, effectively replacing a team of data scientists and analysts. This brings us to the next misconception – the need to be a Data Scientist to extract data insights.
The rush of portfolio managers to educate themselves in data science has been nothing short of astounding. Dramatic statements such as “I will not have a job in two years unless I learn to code” were commonplace until very recently. This sentiment has largely been fueled by the advent of Quant shops that ingest an enormous amount of data and use computers to process and extract signal out of the jungle of Alternative Data sources.
While good data practices are extremely important, and you do need the right tools to navigate the data jungle, it is not necessary for all portfolio managers to drop what they are doing and subscribe to a data science bootcamp. A data science course from Coursera will not have the raw credit card transactions data for you to practice on and your lessons will be broad and not directly applicable to the problems you will face in the real and chaotic world of data. A week’s online course may not be enough and there are better ways to get where you need to be.
The technical skills needed to run an Alternative Data operation are two-fold. First, you need the programming chops to build data pipes for data ingestion and ETL process for each delivery mechanism your vendors use to serve up data. Second, you need the modeling skills to build ML models, statistical models, and pattern recognition software to effectively use the data in your internal process. You will need a lot of problem-solving skills as data always comes with issues – mapping issues, gaps, biases, and more. Writing code to detect and deal with issues before they cause damage is key.
An apparent problem arises. Different teams of developers are performing the same repetitive data work on the same exact datasets and are doing it continually, as new data becomes available.
Some managers wisely choose to side-step the challenge by partnering up with an organization that has already done the heavy lifting of ETL and modeling for each data set. Not only does the manager reduce time to value on their data spend, but the overall manhours saved writing essentially the same code is astronomical.
Finding the right technology partner can offer a great deal of scale to an organization with limited technology resources. After all, the core expertise of a hedge fund manager should be portfolio management and research, not necessarily data management. In contrast, the data company’s sole purpose is to help investors get insight out of data quickly and efficiently. Their expertise is in data and technology.
PMs and analysts are excellent at analyzing data, reports, and dashboards. And even though common belief holds that vast technical resources are required to get Alternative Data to the point where it can be consumed like traditional data, more and more managers are finding ways to sidestep writing code through the right partnerships.
What do you think about Alternative Data? Talk to us about your current process and see how Alternative Data can fit in, or how to improve the value of your current data investment.