Today we talk to AirSage’s Matthew Martimo, VP of R&D. Matthew is an accomplished R&D executive and respected thought leader in the rapidly evolving big data and predictive analytics sectors with nine years of experience driving high-tech product development and revenue goals. We talk to Matthew about how AirSage’s innovative wireless signaling analytics has changed the transportation industry, the current state of population analytics, and where the transportation planning industry is headed. You can find out more about Matthew and AirSage at airsage.com
M: We work with cellphone carriers to place lat/long coordinates onto live events in their wireless network. We can then understand the population as a whole and its movements. We feed that population movement data back to the carriers, agencies, and enterprise to provide data products and services on “where and when” things happen.
A: What services and applications do you provide and to whom?
M: We provide a couple different slices of population analytics. Fundamentally we provide trip matrix, select zone analysis, home-work matrix, and arrival/departure data, but there are lots of different movement data points we diffuse. We’ve been working with the transportation industry for some time to provide data that informs transportation demand models as well as analysis. We’ll produce trip and origin-destination (O-D) data. We can focus on specific areas and start to look at who is going where, associate that with demographic information, and then we can understand the “who” behind the usage of transportation facilities. We sell these kinds of data products and are used a lot by transportation engineers and planners. We make sure that transportation infrastructure investments are analyzed as efficiently as possible.
A: Who is your typical user?
M: My own background has been in transportation modeling, population modeling, simulation, forecasting, etc. I myself used to look for real live data that directly answers some of these fundamental population movement questions that before would have to be completely synthetic. When I joined AirSage over a year ago it seemed like this data was a unicorn, people didn’t believe it really existed. Over the last year and a half part of my job has been to educate people to let them know our data is a real thing and they can have direct answers to some fundamental transportation questions. For new developments, whether that’s state transportation infrastructure or building a new casino, we can let agencies or developers know how many people might use the new highway or casino. We can look at other locations that are comparable across the nation and see what’s going on at those locations. For us it seems like when an engineering firm or agency starts using the data it becomes much more regular for them. After a few studies agencies or engineering firms do with us it becomes a regular thing in their toolbox. They are our typical user.
A: If I never have to do a synthetic traffic model itself it would be a beautiful thing. Have you seen any customers say “AirSage are now the platinum standard and we don’t want the VISSIM style modeling, we want only AirSage data?”
M: We see that a lot more from commercial applications. It’s always fascinating to work with transportation engineers, and speaking as one myself, we are very skeptical and quite deliberately weigh pros/cons. AirSage’s data is absolutely massive to the point where we can sample 20% to 30% of the population at anytime. AirSage can actually observe what’s happening as opposed to the old-school survey of an O-D study and dealing with observation or reporting errors. For the engineers there can be downsides though. For example our data is completely anonymous and so the information I have on demographics and income and household size is only what I can observe from the data itself. I don’t anything about any individual phone number, address, who that person is, or who is carrying that device. And so in only dealing with this data in aggregate we can certainly see and understand the population movement but we don’t have the specific path trace you’d get from a transportation survey and all the specific demographic data that goes with it. Like anything else, with AirSage there’s a learning curve, there’s a new understanding, but there’s also a blending of data. So we are adding something new. Our integrated data, the data quality is really unmatched by anything out there. Now, there are questions that we can’t answer. Yet there is value in blending real-live cell and static survey data together. Once people come in and use our data they certainly don’t want to start going back to what they were doing before necessarily, but it certainly allows them to reduce the need of traditional surveys and helps them get a much higher quality solution. I don’t necessarily think the traditional survey methods will go away. It’s just a matter of changing how we’ve been using them. AirSage improves the final quality of the results.
A: Let’s talk some shop here. I work for an urban mobility startup that produces 24/7/365 data that enables transportation asset owners to do things they couldn’t do before like dynamic pricing. Convincing the owners that there are benefits in changing their current strategies, however, is our main difficulty. How do you convince potential new users to change their approach and that there is benefit in using live and continuous large data sets?
M: Engineers are naturally more skeptical since “they know how to do it” and have been doing it their way for a long time, so influencing change is difficult. We work with younger engineers, groups like YPT that are more open to looking at alternative solutions, and experienced but experimental engineers that are open to finding the best way to do this now. So it’s a little bit of selecting the right audience to begin with.
A: So besides the traditional transportation firms and agencies, what other types of users is AirSage accommodating?
M: Some of the tangential stuff is really interesting. When you look at things like the National Park Service or the National Forest Service, they cover millions and millions of acres of land and tens of thousands of miles or roadway. But they don’t have the same kind of budget that FHWA or a State DOT has. They own tertiary roads at best and they’re making decisions whether to maintain them at all or let them deteriorate to trail access. Our data is able to provide a much broader solution to some of these agencies who may not have the budget or who’s size make it unfeasible to do surveys and studies of their assets. We can also give them a national picture of what’s going on. We’re able to help some agencies who would never have the funding to do the job right. Our data set allows them to look holistically at any potential access or transport problems. Also on the commercial side it’s been fascinating. For example, a company wants to put up a high-end car dealership. They are going to be marketing to the wealthy population. The land where the wealthy are, though, is expensive. We’re able to provide, using our mix of cellular and population data, an approximate mix of locations where the wealthy live and plot over the course of a city where they may travel to. We can then advise the developer where they can build the dealership to better attract their demographic markets. Another market is analyzing population activities over specified time frames. Where do people meet? These are the kinds of questions from companies who are deploying Wi-Fi hotspots. Where are people spending time when not at work or at home? We can map that in aggregate over an entire city. For example, “Give us a location where people were stationary for more than 80 minutes.” We can map a city with very targeted market segments and activity patterns that companies are looking for. This decision support helps people develop more informed strategies and better use their resources in deploying new technologies.
A: You mentioned before that you worked with telecom companies. Do you have partnerships with them? What about other partnerships necessary to deliver your products and services?
M: AirSage has been doing this the last 13 years. We started off assessing locations where different types of cellular events occurred, allowing the telecoms to do network analytics. If a call drops, they want to know the location and closest cellular tower. Our technology defines event locations with very high precision. Telecom’s are sharing data with us so they can get usage insights back. It’s a mutually beneficial relationship.
We work with a lot of different technology vendors. We are a true big data company, receiving over 90 billion events a day, and we extract location data on more than 15 billion of those daily events. There is a lot of effort involved acquiring, maintaining, and looking in a geospatial way at this absolutely massive amount of data. The volume and velocity of the data really exceeds almost any other data source out there right now. We have a lot of partnerships with technology and hardware partners that allow us to push the envelope on storage, recalling, etc. Then down on the customer side and going into specific markets and new applications of the data there is certainly some very specialized companies that come to us asking tough questions, if it’s something we can answer we’ll form partnerships to look into different markets and different applications of the data.
A: At a high level how do you guys handle all that big data? What’s your architecture?
M: We certainly use a lot of different types of technologies. When most people think of big data, they think of Hadoop and Storm, and we definitely use tools like that. For some of our activity we handle raw processing with scripts on massive computing clusters using grid computing utilities like Condor, which is a job distribution framework. We use Storm to keep track of things. A lot of processing is still done with Python scripts and deployments on a massive scale. Our Operations people deserve a lot of credit at making the data accessible and managing these clusters who run and process as fast as they do.
A: Where do you see AirSage going in the next 5 years?
M: Coming from a background of transportation and population modeling, there’s still more research to be done in this field. What kinds of understandings and insights can we get from the population and its movement from these types of cellular datasets and others? We work a lot with universities, with a specific shout out to MIT, to get as much value as we can from the activity patterns that we’re seeing. Conventional methods include doing a one week travel survey of a few thousand households. It will take teams of planners and analysts 6 months to evaluate the data, then they come back with their findings and results.
In metro Atlanta alone there are more than 2 million devices that I get a full survey out of every day. We absolutely need to continue the research on this but really work with engineering groups and researchers to help answer some fundamental questions like how do we even characterize this data? For example we look at a traditional GPS survey and come back with these kinds of insights that this is the “average” day. “Is that the average day?” is a question I have every day of the year. How do we quantify these seasonal, daily, or year over year changes in the travel and movements of the population? How do we assess those types of impacts? The models will become exponentially more accurate when we can see the true behavior of the population and start looking at these predictive models and forecasting models that are used now. Our forecasts will become much better and more reliable and we’re going to be able to answer a lot more questions then we can today.
A: What AirSage is doing is beyond fascinating and I’m sure will really catch on more and more. This will fundamentally change the way that people analyze movement and mobility.
M: To me it’s as you say, absolutely exciting. Our product is a real game changer which is why I’m here working with AirSage now. We need fresh perspectives that folks like YPT bring. We’re absolutely proud of our participation with YPT and want to nurture that relationship. Every day we’re investing more and more in the future. That includes helping develop new engineers as much as it includes developing products on the technology side. We appreciate the opportunity to talk to YPT today.
Arthur Pazdan is a transportation geek, attempted polymath, and humbled surfer. When he’s not paddling out, you can find him at fine coffee purveyors across the San Francisco Bay Area, dreaming of ways to bring our crumbling infrastructure into the 21st century. Flying cars, here we come. @arthurpazdan Arthur Pazdan on LinkedIn