INTERVIEW: Tony Jebara, chief scientist for New York start-up Sense Networks and a professor at Columbia University, tells how location-based data is being used to predict consumer behavior and preferences.

Jebara, who is delivering a keynote presentation at this year’s MetaPlaces09 conference, said the results can be used to highlight hot spots where different urban "tribes" gather – but can also give advertisers a better idea of where and when to advertise to certain groups of people.

Someone who goes to Starbucks at 4PM a few times a week probably has some similarities with others who also visit the coffee chain at around the same time – regardless if they are in San Francisco or New York.

Equally, knowing where someone in San Francisco has dinner on a Friday night could help a visitor to the city make a better restaurant choice, according to Tony Jebara, chief scientist for New York start-up Sense Networks.

His company has developed a phone application that highlights hot spots where people are gathering around a city.

App Like "Sixth Sense"

Called Citysense, the app uses frequently updated cell-phone and taxi GPS data to produce a heat map of where users are in the city.

Jebara said people who have used the software love it because it is like a "sixth sense about what’s going on in the city".

The application is currently up and running in San Francisco and is expected to be launched in New York in August before being rolled-out to other cities.

Tony Jebara, chief scientist Sense Networks

But providing basic activity information is only the start.

Sense Networks’ platform, Macrosense, is able to receive streaming location data in real-time, analyze and process the data in the context of billions of historical data points.

This can then be stored in a way that, the company says, can be easily queried "to better understand aggregate human activity".

Users Categorized in Urban Tribes

So in a new version of CitySense, expected shortly, this data will be used to reveal the movement of people with certain behavior patterns – urban "tribes" such as students, tourists, or business people, for example.

What this means in practice is that users could arrive in a new city and with the help of CitySense find bars, restaurants or other activities that chime with their tastes and socio-economic profile.

Jebara, who is also a professor at Columbia University, said that while people loved the fact they can see a street map of city-wide activity, they wanted something that is customised for them to show "people like me" or "tribal clustering".

Location Data Potential

He said the Sense Networks’ software was initially developed to allow stores to use location data in order to monitor consumer activity.

But they quickly realised that the information had much more powerful applications.

"What the platform does is it looks at different places and figures out what happens there," he said.

"At different times, what kind of common activity is taking place? It looks at individuals and how they are exposed to different types of commercial activity and how they spend their leisure time."

This can be whether someone choses to go, say, to a high-end restaurant or a nature park at weekends.

The data on an individual’s movements then allows them to be categorised and their probability of doing different activities calculated.

Sense Networks has defined 24 "types" or "tribes": student, business, young and edgy, stay-at-home parent etc.

These tribes are determined using three types of data:

  • a person’s "flow" or movements around a city
  • publicly available data concerning the company addresses in a city
  • demographic data collected by the US Census Bureau

Jebara said someone can be a mix of tribes, such as student and stay-at-home parent.

"What’s interesting is that these tribes carry across different tribes and cities," he said.

"If two women like to shop at high-end stores, they will have a similar profile even though one is in New York and the other in Dallas.

"They are more similar than two women in Dallas, if one does not shop in high-end stores."

A Next-Generation Facebook

Jebara said this is a good way of modelling for marketing, indicating if someone is likely to be interested in a particular advert, or would download certain mobile applications or upgrade a phone.

"There are a variety of business decisions that we can derive by using location data to look at what people are doing," he said.

"It’s a way of building the next-generation Facebook. Instead of having someone’s profile typed in, we figure out where they hang out and the activities they do.

"That determines their profile and they can be linked to similar people."

Jebara’s keynote presentation at MetaPlaces09 is titled A Snapshot of the Location Industry.

The two-day conference in San Jose, California is attended by the leading location platform and service providers, as well as wireless carriers and device manufacturers.

Privacy Issue

To increase the accuracy and effectiveness of its software, Sense Networks stores historical location data.

So the number of times a person goes to a particular store or restaurant is saved to build up a profile.

This idea of being tracked and logged understandably makes people uncomfortable, but Jebara stressed this is not exact data.

The raw data is used to analyse commercial activity and demographics and then disposed off.

"We do not have latitude-longitude information about any individual, so if the FBI asked us for information they would never be able to figure out where someone was in the past," he said.

"The data just tells us someone likes high-end restaurants with a family crowd, for example.

"It tells us the probability of different commercial, demographic and tribal exposure.

"There is a lot of anonymity in that prediction."

Sense Networks, headquartered in New York City’s SoHo neighborhood, was founded in 2003 and incorporated in early 2006. The founding team is composed of top computer scientists from MIT and Columbia University.

Among them is Alex Pentland of MIT, who pioneered reality mining, a research trend that is trying to tap into the potential of location-based data.

Accuracy Improved

Jebara said one of the key things holding location-based applications back is concerns over their accuracy due to poor signals or infrequent pinging.

However, he said Sense Networks’ software looks at the long-term history of activity and summarises that to say what people are doing or could be doing.

"We do not just look at the current latitude and longitude or time, but we augment that with the history of what someone was doing for the last three weeks," he said.

Again, Jebara said this is not the exact data – just that someone went to a certain type of restaurant or certain type of nightclub, and so on.

"Combining this history with more recent things overcomes the problem of just using single location pings," he said.

Jebara said Sense Networks’ intention isn’t to keep the technology in-house but to make the analytics engine available to other people to use on their apps.

This would even apply to a company building a rival app to Citysense.

He said the beauty of location data is that it has the same format everywhere (latitude, longitude, time and an error measurement).

This Lingua Franca doesn’t need to be translated and it can be used anywhere in the world.

Sense Networks plans to provide its location data on city activity to advertisers.

Tailored Advertising

This would comprise details on where certain types of people congregate and when.

So, for example, Sense Networks’ data-analysis algorithms may show that a particular demographic heads to bars downtown between 6 and 9 PM on weekdays.

Advertisers could then tailor ads on a billboard screen to that specific crowd.

While operators and advertisers stand to gain from the use of this location data, Jebara believes the consumer will actually benefit more.

"People don’t want to fill in forms and answer questions. Consumers want customised recommendations rather than generic advertising," he said.

"If this data is properly leveraged we will trust it a lot more. It will empower the user."

Once consumer trust in Macrosense is there, Jebara said it can be combined with Citysense to offer something of value to users.

He compared it to the early days of Google when the search engine had to first earn users’ respect by proving its worth in finding things accurately.

Once that was achieved it was possible to include some relevant adverts as well.

"It becomes much more palatable to the user if it is combined with something useful," he said.

"So first we have to win over the hearts and minds of customers. Then the business opportunities will be great."

For more information on the MetaPlaces09 conference (22-23 September 2009) in San Jose, California, please click HERE

Subscribe to our Newsletter