The Buy Side Delves into Mobile Data

August 30, 2017 | By: Ivy Schmerken

mobile data

By Ivy Schmerken, Editorial Director

As hedge funds and quantitative asset managers hunt for unique sources of alpha, Wall Street’s attention is turning to analysis of location data generated by mobile phones.

Location data from mobile phones is now at the forefront of the push to combine data science techniques with machine learning techniques to produce actionable information on company fundamentals.

Third-party aggregators are seeking to quench the thirst of hedge funds for non-traditional datasets.

On Aug. 9, Thasos Group, an alternative data intelligence startup, launched Streams, a real-time location information feed that transforms mobile phone data into actionable insights for traders, investors and other business managers.

With GPS enabled phones spewing out torrents of data every minute, Thasos has built up the largest repository of location data after Apple and Google, the company said.

Thasos obtains data from hundreds of millions of mobile phones, which is then mapped to every store or venue owned by over 400 public companies. It then filters the location data from the cell phones, with the longitude/latitude coordinates from every venue owned by the 400 public companies.

This way, it can provide a more timely and accurate picture of foot traffic to retail stores, restaurants, malls and airports, deliveries to loading docks, hours worked on assembly lines and patient counts in hospitals.

Formed in 2011 and spun out of MIT Media Lab, Thasos has been selling its location information feed to about 25 hedge funds over the past two years.

Demand for location data is growing, “not because of its historical availability, but due to the ability to extract value from it,” said John Collins, founder and chief product officer at Thasos in New York.

The most comparable data set in terms of forecasting earnings has been credit card transaction data, which has been a staple of hedge funds for the past two years, explained Collins. “But location data is a lot messier and a lot larger and is a lot bigger of a problem to solve since it comes from many different sources,” he said.

One hurdle is that most apps have not historically logged the locations of their users. Thasos said it knows the sources that have consistently logged their data for the past two or more years. Thasos obtains the cell phone data from app providers, whom they contract with directly.

“We’re in a position to know so we are well positioned to deliver data in the hedge fund space with respect to calculating the year-over-year change,” said Collins.

The trend has exploded in the retail sector where analysts at hedge funds and quantitative fund managers are vying for an edge.

Even before a company reports its quarterly earnings or same store sales, Thasos can calculate the foot traffic in venues owned by companies like Home Depot (NYSE:HD). “We cover 2,000 Home Depots, all 4,500 Walmarts and hundreds of thousands of locations for public companies,” said Collins.

“A lot of hedge funds and other kinds of buy-side firms are analyzing that data to see which retailers people are going to,” said Valerie Bogard, analyst with TABB Group, who follows the alternative data space for institutional buy-side and sell-side firms.

When consumers download apps on their phones, they are often asked if they want to allow location tracking. Popular social networking platforms like FourSquare and Yelp (NYSE:YELP) will make recommendations based on the user’s location. “These apps are aggregating the data, making data anonymous and they’re selling to hedge funds and whoever wants it,” said Bogard. Hedge funds presented with this data can get an early indication of foot traffic at Walmart (NYSE:WMT)  vs. Target (NYSE:TGT).

Predicting Economic Activity With Mobile Data

Because the mobile data is collected in real time, and is associated with ticker symbols of the public companies, it can be analyzed to predict economic activity ahead of quarterly earnings releases.

On July 26, Thasos released a report on real estate investment trusts  (REITS) that invest in shopping malls, predicting a decline in foot traffic by anchor stores, including including Macy’s (NYSE:M), Nordstrom (NYSE:JWN), Dillard’s (NYSE:DDS) and Sears (NASDAQ:SHLD). The report accurately predicted a sell-off in REITS that occurred after Q2 earnings were released. Prior to that release, many mall REITs reported or suggested that year-to-year growth in foot traffic across their mall properties was positive.

“What we’re talking about here is macro-trends based on mall traffic counts,” said Jamie Benincasa, senior vice president, FlexTrade Systems.  Mobile is the next step in terms of sourcing data sets in order to make information investment decisions, said Benincasa.

The Trend Accelerates

While hedge funds are consuming market data – stock quotes, historical data and company fundamentals — there is a push for less obvious data.

“Everyone has consumed the normal data sets. Therefore everybody gets the same ideas. In order to be more original, they are looking at these alternative sources of data,” said Benincasa.

While the largest quants are the early adopters of alt data, this most recent dataset is becoming more distributed and accessible to firms that don’t have the internal resources to analyze it on their own. And as mobile phones, satellites and drones become cheaper to produce, alternative data has gained more adoption from smaller hedge funds, long-only asset managers and sell-side firms, said Bogard.

Aggregators are out there pulling in location data, credit card transaction data on an anonymous basis. Then there’s satellite data and public web site scraping and then using analytic techniques to find comparisons, added Bogard.

In 2016, TABB conservatively estimated that the market for alternative data in the U.S. is currently around $200 million and could double within the next five years.

London-based Neudata Limited is one such firm that is helping hedge funds and the institutional investment managers find and evaluate alternative data sources.

Unlike firms that sell alternative data, Neudata provides an independent data research platform, which covers datasets that span commodities, equities, FX and fixed-income asset classes. Varieties of data types include: crowdsourced, economic, location, satellite and drone, sentiment, social media and weather. “We are an outsourced resource to the buy side,” said Rado Lipuš, founder and CEO of Neudata. “We find new data sources for them, and second we do due diligence on these data sources and provide metadata about the data sources,” said Lipuš.

“We try to make sense of hundreds of these data sources,” continues Lipuš. His firms answers questions, such as what’s the content; what’s the universe; how much history is there; is it an aggregated format; is it clean; is it unstructured, etc. “There are a lot of questions and the questions will be different for the fundamental or quant,” said Lipuš.

According to Lipuš, many of the quants have an in-house team vetting the alternative data sources. “We do the same thing on an outsourced capacity.”

Outsourcing the Data Science

The value in outsourcing is that aggregators are performing the heavy lifting so that hedge funds and asset managers can hire fewer data scientists.

“Normalizing the data is one of the hardest problems to solve,” said Collins, who explains that Thasos removes the noise, such as cars that drive by or people who not actually customers of casual dining restaurants.

“Obviously, the large quants are doing this internally, but in terms of processing and analyzing [data], there is a lot of data that people don’t care about,” said Bogard. “There is a lot of processing involved in terms of finding that diamond in the rough, in terms of [finding] a specific data point.”

Those working with large-scale data sets, whether from mobile phones or sensors, are applying machine learning to extrapolate the data they are interested in, said Bogard.

“It all ties into the big data construct. It would be impossible to weigh these factors, such as foot traffic and credit card transactions, without sophisticated hardware to tune the artificial intelligence. That is becoming less cost prohibitive and available to a wider audience,” continued Bogard.

“Now with the ability to handle larger and larger volumes of data in a shorter amount of time, and really discern these patterns with the application of technology, we should expect to see more and more of this,” said Benincasa.

However, firms that are examining alternative data sets “don’t necessarily want to broadcast how they are interpreting the information,” said Benincasa.  They want to do the analysis in-house and keep it proprietary, he said.

How funds are using the location data varies from fund to fund, and depends on their level of sophistication, said Collins.

Traditional hedge funds that are not automated might plug the information into Excel and sum up the foot traffic data for the current quarter of the year, and the same quarter last year.  They might compare the location data to credit card transaction data, said Collins.

On the opposite end, there are quantitative firms with thousands of positions that ingest the data directly into their multi-factor models that cover entire industries. “And so our data helps them understand whether they should go long or short on every company in a particular industry relative to another industry,” he said.

“Ideally, what the managers want is finding sources they can trade on, or finding sources that add to their factor model,” said Lipuš referring to the rules-based algorithms that quantitative asset managers build in-house. “They may be looking to address a small problem, just a factor that is weak or is missing in the model. It’s very subjective. That is where the art and science come together,” said Lipuš.  Systematic funds will backtest the data to see how it behaves. Fundamental firms are looking for a supporting data source, which will prove an idea, he said.

Ultimately, most hedge funds are looking for a signal they can trade on.

While Thasos has been delivering its products as weekly deliverables, the overwhelming demand from clients is to move to a daily delivery system, said Collins

As location-tracking data is becoming more distributed beyond the top-tier quantitative firms, smaller hedge funds are also delving into it.

Though location tracking is not yet considered to be mainstream, it’s moving in that direction, said Bogard.

Past blog posts about Data issues

MiFID II Transparency Puts Stress on Data Architecture

Discretionary Managers Seek Alpha in Alternative Data

Quant Funds Get Sentimental About Big Data

Managing Big Data After Brexit