BUZZ Indexes Enables the First ETF Built Using AI

The application of artificial intelligence (AI) and big data to the investing process is an exciting one, and up until now this application has been limited to only the most sophisticated institutional investors. One of our readers recently informed us about a company called BUZZ Indexes that takes all the big data being generated by social media, interprets that data using AI, and then uses it as a momentum factor to determine which stocks are most likely to outperform their peers. While we generally look at social media as a time wasting black hole as opposed to a legitimate technological advancement, we’re genuinely interested in learning more about this new investment product that allows any retail investor to get in the game.

The way an ETF works is that you have an index provider who tracks a basket of stocks and then an ETF provider who then tracks that index and turns it into a tradable investment vehicle. In this case, a Canadian company called BUZZ Indexes has built an index called the “BUZZ Social Media Insights Index” which identified the 75 most bullish stocks based on social media “big data”. This index powers the Sprott BUZZ Social Media Insights ETF (BUZ).

Here’s how the index is produced. Firstly, BUZZ Indexes needs to identify their big data sets. These are not just traditional social media outlets like Twitter and Facebook but also websites like Seeking Alpha which has a treasure trove of useful comments, in fact, the comments are often better than the articles themselves. Here’s a look at just some of the sites in the “social universe” that BUZZ gets their data from (there are actually 50 different blogs, websites, and news services they query):


Next, BUZZ needs to “scrape” all the data, in this case 50 million unique-specific data points, and then interpret their sentiment. Extracting meaning form all this big data is where the artificial intelligence piece comes in. Do you recall a company that we wrote about before called Aspectiva that was scraping product review comments and then measuring positive or negative sentiment? That’s kind of what BUZZ is doing. Each month, they use “natural language processing” algorithms to determine which stocks are being talked about and whether or not people are bullish or bearish across the 4 previous quarters as seen below:

BUZZ AI Big Data Process

The universe of securities is all stocks that trade on major U.S. exchanges and that have a market cap greater than $5 billion and $1 million in trading volume or more. Each stock in this universe is assigned an “insight score” and the top-75 stocks are included in the index. Holdings for August can be seen below with the stocks removed during their last rebalancing marked in yellow:


Since rebalancing occurs monthly, this seems to be a fairly high level of turnover which results in higher transaction costs. However in reading the methodology, BUZZ states that excessive turnover is undesirable which is great to hear. It’s unclear whether or not transaction costs are included in the total expense ratio, but at .75% (which is rather high for an ETF) we would assume they are. Here’s the list of positon changes that took place in the last rebalance:

buzz-september-index-rebalanceAfter additions and removals, the weightings for each of these 75 stocks are then rebalanced according to a proprietary model which caps the maximum weight in of any stock in the index at 3%. It’s really quite slick though they may not yet have worked out all the kinks. While they originally debuted the index with 25 stocks, just last month they changed their methodology and changed this to 75 stocks. That’s a pretty significant change to make just 4 months after creating the ETF but it makes a whole lot more sense.

So how has the Sprott Buzz Social Media Insights ETF performed so far? Here’s a look at the fund’s performance so far:

buz-etf-returnsSo let’s compare this to the S&P500 over the same time frames;

  • One Month: -.56%%
  • Three Month 4.71%
  • Since Inception 4.21%

Since investing is a much longer game than one counted in months, the jury is still out on how this investment vehicle will perform against the broader market. When you have an investing strategy, you can test its historical performance by doing something that’s called “backtesting”. In this case, they could only “backtest” back to January of 2013 since the historical social media data just wasn’t there. During these backtests, the BUZ ETF outperformed the S&P by about 30% and Nasdaq by 10 percent. It would be interesting to hear how these results change following their major methodology changes.


The idea of mining social media “big data” wasn’t just invented by BUZZ Indexes but was covered in at least 14 different academic research papers cited by the Company. It’s also not that different from traditional “factor investing” where you overweight momentum stocks and therefore get exposure to a particular factor. You could also look for negative sentiment and short stocks based on these same ideas. It’s certainly a very interesting application of AI and NLP and something that allows retail investors to take advantage of some of the same advanced technologies used by the world’s highest performing hedge funds. We’re going to mark our calendars for 1 year from now so we can check back and see if the “BUZZ Social Media Insights Index” is actually generating the alpha we expect it to. The BUZZ Social Media Insights ETF trades under the ticker BUZ.

While NLP is now mainstream, there's plenty of growth left to be had. Four of the five AI stocks we're holding are SaaS businesses selling AI-powered enterprise software. Become a Nanalyze Premium annual member and get immediate access to our entire tech stock portfolio.

Leave a Reply

Your email address will not be published.