Buzzwords, buzzwords, buzzwords. Every age has different buzzwords. Today, probably some of the most popular are machine data and alternative data, whether you work in finance or any other industry. Want to make more profits? Well, use machine learning, use alternative data etc. (ok, it isn’t that easy, but I’ll explain more later).
Rather than just hand waving, and chanting, we need to understand how these techniques can be helpful for us. In particular, how can machine learning be useful if we have alternative datasets? Machine learning can provide us with many useful techniques for making sense of alternative data. After all, in order to structure alt data we often use machine learning. Making sense of text, requires NLP, and many NLP models use machine learning these days. The same is true of computer vision. Whereas in the past it was dominated by rules.
Given that banding around buzzwords isn’t really sufficient, rather than thinking about the techniques or resources we’ll use (machine learning and alternative data), we need to think about the types of questions we want to solve, where potentially they might be useful. It’s just taking a plane People travel from A to B and a plane facilitates that. They don’t go from A to B, in order to spend hours queuing at an airport, and to spend time in a plane. Another important point with alternative data, is that it might help you solve questions, which you were never able to ask with “traditional data”.
Of course, these questions are going to differ from trader to trader, even if their objective might be the same, to generate returns. If we’re a high frequency trader, it might be, trying to find more data to seek out what can’t be explained by market data. For example, are there any important events in the calendar, or unexpected events which are explained by the news (and result in a trading halt).
If by contrast, we are a longer term macro investor, we might be trying to understand broader economic themes, and to assess which are driving the market (eg. Google Trends, looking at readership statistics of news articles etc.) or to try to forecast economic variables in a more timely and more accurate way, especially during these days of COVID-19, where the economic picture can change rapidly between traditional releases of economic statistics like GDP. Or you might be interested in understanding how consumers are spending at a consumer good firms, which constitutes a large equity holding in your portfolio (eg. using consumer transaction data).
With any sort of research or new exercise, we might not always be successful, but hopefully, along the way we’ll still learn something useful. Furthermore, the idea of using alternative data isn’t to solve a single question you have, but to help in the many questions you might ask. In some cases. Getting into alternative data is a gradual process, which takes time. It’s important to start thinking about how alternative data can help your business, whether you are involved in financial markets or outside. If you don’t explore this path, it’s still likely your competitors will do.