The catchphrase "artificial intelligence" may be ubiquitous, but it's hard to find people who actually know how to use it. That discrepancy represents an opportunity for products like Watson Analytics, IBM's point-and-click AI tool that lets non-experts analyze business data by asking questions in colloquial speech. In recent years, IBM has been building out the data sources that Watson Analytics can draw from—and is now adding granular mapping info through a deal with geodata provider Mapbox.
Watson Analytics is not the sexiest form of AI—pretty far from Ex Machina. But it's attractive to people in fields like marketing, who can start with as little as a spreadsheet of sales figures and get artificial insights into how and where they might sell more product. (Single user accounts range from free to $80 per month.) Since announcing Watson Analytics in September 2014, IBM has been adding built-in data sources that customers can pair with their uploads. In March, 2015, it announced a deal with Twitter to let Watson Analytics users study online chatter. That same month, IBM also made a deal for detailed meteorological data from The Weather Company (owner of The Weather Channel and Weather Underground)—including moving its data to IBM's cloud. A few months later, IBM purchased The Weather Company, a deal that closed in January 2016.
"For all these companies, the signal may not be in your own data," says Marc Altshuller, IBM's general manager of business analytics. "Most likely, the signal is in a combination of market data like what Mapbox has, plus your own internal data."
Currently, Watson Analytics' geodata stops at the county level; when Mapbox is integrated, it will go down to street addresses. (IBM hasn't provided a time frame for the integration.) IBM expects this to open up opportunities for brick-and-mortar businesses to figure out how sales vary by store location, for instance, or the best site for a new store based on factors like traffic. Without street-level detail, "you may not see that it's urban or rural that matters, whether it's in a shopping mall or outside a shopping mall that matters," says Altshuller. "Maybe it's something related to traffic and time of that traffic."
Other potential ways to use Mapbox data: analyzing national sales by geographic region, evaluating the success of junk mail campaigns, or projecting home values based on factors like crime and traffic. Maybe not exciting, but potentially lucrative. None of these are new applications in the business world, but IBM is offering the ability for non-techies to do this kind of analysis for a few bucks a month—and without needing much technical know-how. Natural language processing can return data on straightforward typed questions like: "What is the average income on the Upper East Side of New York City?"
The deal will also integrate Mapbox into another IBM product called Cognos Analytics, which is expected to happen by the end of the year. It's a drag-and-drop interface to create reports or interactive "dashboards," like a single screen that represents data in bar graphs, line graphs, and tables. IBM has a lot of competition here, with rivals such as Microsoft Power BI, Qlik, and Tableau.
IBM and Mapbox created for us a few examples of zoomable, interactive maps for Cognos. One breaks down a week of 311 call complaints in San Francisco, with the ability to drill down to each address. Another uses census data to explore how income and population vary by county around the U.S. While Watson and Cognos already had county-level geographic resolution, Mapbox adds spiffier ways to visualize it, such as the zooming capability, says IBM.
The deal is as much an enhancement of Watson and Cognos as an endorsement of the young mapping company that's taking on behemoths like Google and Apple. Founded in 2010, Mapbox has raised about $62.5 million in the past three years. It provides the mapping component for over 5,000 platforms, including, Etsy, Foursquare, Instacart, Lonely Planet, MapQuest, and Pinterest. In June, the company launched a service called Mapbox Drive for powering semi-autonomous vehicles.
CEO Eric Gundersen has a background in international development, and he was originally targeting clients like the UN and the World Bank through a consultancy company called Development Seed. "These are organizations working on the ground, figuring out where to put clinics, mapping malaria across Africa, [doing] election monitoring," says Gundersen.
"Bit by bit we had to build up not just the tools to visualize the data, but the underlying data set, the map of the world where all the cities are, the terrain and the satellite imagery," he says. Mapbox, which spun out from Development Seed, did this on the cheap, harvesting data from OpenStreetMap, a nonprofit, crowdsourced mapping project. Because OpenStreetMap is open-source, Mapbox's refinements are available to everyone, and vice versa. (The work was funded by a $575,000 grant from the Knight Foundation.) It also pulls from other public sources, such as USGS, Landsat, Natural Earth, and OpenAddresses. In 2015, though, Mapbox purchased the rights to 3 million square kilometers of imagery from DigitalGlobe, which captures some of the highest-resolution (nonmilitary) satellite views available.
The company also makes Mapbox Studio, user-friendly mapmaking software, which got a thorough rework in 2015. It's not the only such point-and-click tool. Carto (formally CartoDB), for instance, produces gorgeous geographic visualizations of things like Twitter traffic during different moments of the Academy Awards Carto's new geographic analysis tool, called Builder, is a competitor to the new mapping abilities that Cognos will get. (Mapbox supplies data to Cognos competitors, as well, such as Tableau.)
Mapbox was already working out a deal to provide data to The Weather Company, says Gundersen, when IBM bought The Weather Company. A new update to the Weather Channel iPhone app uses Mapbox for customizable, zoomable time-lapse visuals like showing how storms move across the Caribbean or how temperatures fluctuate among city neighborhoods. The feature is coming to the Android version of the Weather Channel app, as well as to the Weather Underground app, says Gundersen.
IBM so valued weather data that it ultimately bought the company that was providing it. Might it feel the same about The Weather Company's erstwhile business partner and snap up Mapbox, as well? "We're not for sale," says Gundersen. "It's the early days of the entire space."