Tapping user location data from mobile phone application vendors, start-up Thasos has launched a time-series location data offering, dubbed Streams, company officials announced today, August 9.
The vendor, founded in 2011, spent three years developing and training its internally developed machine-learning algorithms to process the vast amount of data into its final form.
“We have signed data license agreements with hundreds of sources, take their data, and evaluated it,” Greg Skibiski, founder and CEO of Thasos, told Markets Media.
Thasos looks for data sets with samples sizes in the hundreds of millions with at least two years of historical data.
“We see hundreds of millions of phones pinging the server every hour,” said Skibiski. “It is quite a high-frequency global data set from everywhere in the world.”
The vendor began its industry coverage with a focus on 150 listed retailers but has expanded to cover 400 of public companies across multiple industry verticals.
“We started there because they did us a great crowdsourcing service of creating all of this training data by releasing their numbers every quarter,” he explained.
Institutional investors typically have used Thasos data initially to confirm signals the investors detected within credit card data.
“When clients see that our data and the credit card data are moving along together, they have a lot of confidence that they know there’s something weird going on," he said. " Whereas, they know sometimes that our data and credit card data diverge, which simply gives them a risk signal that says maybe they shouldn't trust this right now and to pull out.”
However, Thasos provides coverage for industry verticals and subsets of the retail sales where credit card data cannot provide insight, such as the hotel industry, where people book via third-party websites, and dollar stores where credit cards aren't widely used.
Thasos also draws traffic data from locations that they geofence like shopping malls, airport terminals, manufacturing plants, and hospitals. By tracking movements in and out of these areas, Thasos can which provides additional metrics, such as factory hours worked by full-time equivalents and occupancy rates
Users can access the previous week's data starting Sundays nights from any Streams to which they are subscribed.
"It takes time to collect and process the location data in the cloud, do the machine learning, validate the results, check them, and deal with every little error and glitch that comes in,” said Skibiski. “It is a pretty formidable technical challenge even with the technology that is available today.”
Thasos processes its data in the Amazon cloud and Microsoft's Azure platform. The dual-cloud environment is due to many of the company's data providers delivering data to one cloud or the other, he added.