Eight Ways to Use Alternative Data for Trading
In the past year, we’ve looked at various ways alternative data can be used for investing, from satellite imagery that forecasts metal prices to predicting geopolitical risk between the world’s countries. If you’re a large company of any kind, it’s likely that you have your own valuable alternative data laying around. A company called Eagle Alpha can help you monetize that data by selling it to firms that use it to make better investment decisions. Today, we’re going to look at some ways that alternative data can be used for making better trading decisions. Most of these examples are courtesy of a report published by Eagle Alpha titled Alternative Data Uses Cases – Edition 6. Our first example comes from a startup called SpaceKnow.
An Eye in the Chinese Sky
Founded in 2013, San Francisco startup SpaceKnow has taken in $5.2 million in funding to enable “ultra large-scale planetary analysis” through the use of satellite imagery. The company’s platform allows you to scan satellite imagery and choose to have machine learning algorithms automatically identify objects like cars, boats, trees, and even swimming pools. The usefulness of this tool is evident in examples like SpaceKnow’s Satellite Manufacturing Index (SMI) which is said to be an upgrade to China’s Purchasing Managers Index (PMI). (The PMI is an index compiled by the National Bureau of Statistics of China which compiles the results of a monthly survey of enterprises’ purchasing managers). SpaceKnow’s SMI uses an algorithm that compares satellite images of more than 6,000 industrial facilities, and produces a result that’s remarkably correlated to the index China produces.
Since the days of Mau’s Great Leap Forward, investors have been suspicious of official Chinese economic data. Now, they can use the SMI which is an independent measure that’s published much more frequently – every Monday, Wednesday, and Friday – making it the fastest, and most reliable, gauge of Chinese industrial activity.
Following Jets and CEOs
Founded in 2017, New Yawk startup JetTrack has taken in $1.8 million in funding to provide ten years’ worth of flight tracking information for 1,700 different aircraft across more than 25,000 airports. These aren’t just any aircraft, but rather planes operated by all the companies in the Russell 3000, the largest activist hedge funds, and the largest private equity funds. Sounds like there will be a “no more using the corporate jet for sensitive meetings” policy and the C-types will all be relegated to traveling First Class alongside the commoners. That’s because this data has been used in numerous cases to predict corporate events, like when Constellation Brands kept flying up to Ottawa at an increasing frequency before making a $3.8 billion investment in Canopy Growth Corp.
The parent company of JetTrack, Paragon Intel, has also developed a product called CEO Change that will provide subscribers with inside information on each external CEO Change in the Russell 1000. This includes an Investment Memo, a Background Check, and Full Transcripts with Interviews of a CEO’s former colleagues. The company claims that superior alpha generation can happen as a result.
Maybe the CEO Change product can help us better understand what’s happening over at MedMen.
Founded in 2013, Connecticut startup RS Metrics has taken in $14 million in funding to develop a geospatial analytics platform for businesses and investors. In a previous article, we talked about how they’re Using Satellites to Forecast Metals and Commodity Prices and looking for “ground truth” in the world of ESG investing. We recently visited RS Metrics’ headquarters in Connecticut to learn about how they’re using satellite images of Tesla’s production lots to gauge just how many cars are being produced and shipped. The machine learning algorithms are not only able to discern between various Tesla vehicle models, but they’re also able to tell which cars are still parked in the same space and haven’t been moved between images.
In the above images, we can see how the North Production Lot at Tesla’s factory in Fremont, California has changed in just four days’ time. It now seems much fuller and contains almost exclusively Tesla models S and X. If you’re an analyst covering Tesla, this kind of information is pretty much mandatory for fine-tuning those Excel valuation models.
10 Years of Google Trends
Eagle Alpha is a vendor of alternative data that we discussed in our previous article on Monetizing Your Alternative Data With Eagle Alpha. Founded in 2012, the Irish startup has taken in $8 million to disseminate more than 1,000 alternative datasets to the masses. Among those are Eagle Alpha’s own proprietary datasets, one being based on Google Trends, a public web tool based on Google Search that shows how often a particular search-term is entered relative to the total search-volume over time across various regions of the world. This search data is a byproduct of people’s normal day-to-day activities, and consequently is an excellent determinant of economic activity with over 10 years of history.
Eagle Alpha specializes in analyzing Google search data which isn’t as easy as it sounds. That’s because Google “supplies trend data at varying frequencies depending on the time span requested.” This means you may get different results when running the same query. Eagle Alpha has developed a methodology for handling these sampling problems by using “multi-sample averaging of identical time spans for the same term.” The end result is an accurate dataset with remarkable predictive powers. For example, the data powers Eagle Alpha’s US Unemployment index which has a 5-year correlation of 0.9 with the US Unemployment Rate.
For the next four examples we’re going to cover, we’ll look at alternative datasets that are available on Eagle Alpha and provided by undisclosed vendors.
We’ve talked before about how geolocation data can provide us with interesting insights such as where America’s fractured society likes to dine. There’s enough data now that we can observe foot traffic changes for various retail outlets. In one example, the geolocation data provider was able to monitor the foot traffic at Whole Foods locations in response to a price reduction event while also controlling for the impact of seasonality and Labor Day weekend. In another example, the data was used to anticipate disappointing sales at RadioShack stores.
One unnamed vendor is using data derived “from more than 900 million mobile phone users, mainly in China,” to “track specific Chinese industries, such as steel production and consumption, as well as consumer trends.” Applying the data to an investment methodology that times the market results in some meaningful performance.
Hiring Freezes and Hiring Sprees
In many corporations, company executives are allocated a budget for things like training and headcount. When things tighten up, the CEO may then commence a “hiring freeze” which means all hiring stops immediately. This might be followed by a round of layoffs during which the talent pool consolidates a bit before things get back to normal. These hiring freezes are signals of the purse strings tightening in much the same way a “hiring spree” would indicate some new budget has been allocated to grow a particular area of the business. Even the types of hires being advertised can telegraph hidden messages.
Completely unique in the industry, a job listing alternative dataset from one of Eagle Alpha’s vendors indexes jobs directly from employer websites. Updated daily with over 4 million jobs from more than 30,000 employers, the platform eliminates duplicate and expired job listings. Testing shows that there is alpha in the data, with the “Jobs Active” variable (number of job postings that are currently active and open) producing the highest and most consistent returns (6-8% yearly).
Then, there’s a vendor that provides “public profile information for CRM system enrichment” which tracks job changes. Sounds like it might be sourcing from all those public LinkedIn profiles out there as the dataset includes “344 million profiles with basic details and 262 million profiles with employment records,” from over 10,000 publicly traded companies across the globe. This data was then used to produce 1.4 billion “job change events,” that then showed which companies had the highest and lowest turnover. Using a long/short portfolio from the top and bottom quintiles, the vendor was able to demonstrate a viable short-term trading strategy.
Email Receipt Data
One vendor collects anonymized purchase data from around 2 million active shoppers, scanned from email purchase receipts. It covers over 600 merchants from more than 25 industries. The unstructured email receipt data is transformed into a normalized consumer transaction dataset which includes item and SKU-level transaction data which is filtered into 53 product categories. JP Morgan performed a cross-sectional analysis of the data and used it to create a trading strategy which “generated annualized returns of 16.2% and a Sharpe ratio of 1.13.” In another example, Eagle Alpha used the data to produce predictive reports of earnings expectations of which 80% correctly predicted Beat/In-line/Miss.
If you’re wondering where all that receipt data is coming from, the answer may lie in our article on Smart Receipts and Why We Should Use Them.
Only One Room Left
If you’ve spent much time booking hotels online, you’re probably familiar with the “4 other people are looking at this room” message which some might dismiss as a clever marketing tactic. The same holds true for the “only one room left” message. One data vendor visits travel sites owned by the Expedia Group (EXPE) such as Travelocity, Orbitz, and Hotels.com to find all active properties and then record key facts for each property including location, property type, and booking messages like “10 people booked this property in the last 48 hours.” They’ve been collecting this data since mid-2015 and have been able to accurately predict strengths and weaknesses for Expedia’s various brands. While this use case may be limited to one stock, there’s no reason this method couldn’t be extended to other hotel chains – or even Airbnb for when they decide to go public.
You can’t prove that an alternative dataset generates alpha by cherry-picking your time frames like some ESG companies like to do. The way to demonstrate the usefulness of alternative data is to have some ground truth that you can verify against, like SpaceKnow’s China manufacturing index which replicates an existing data index but is far more timely. We also see some examples of very company-specific alternative datasets, like counting cars in Tesla’s production lot or figuring out how many hotel rooms are being booked by perusing online travel sites. While hedge funds have been doing this stuff for decades, the explosion of big data availability means we can all use it in different ways to gain insights that lead to better investment decisions.