Say that you are planning a trip: in addition to emails from helpful friends and family members detailing all the places you ‘absolutely have to go to’, you must also sift through numerous websites, guidebooks and travel forums before making any final decisions.
How can you choose the places and attractions that you might like to visit based solely on the experiences and preferences of others? In other words, how can you make an objective decision based on the subjective experience of other people? Planning would be so much easier if you knew which neighbourhoods, shops and attractions provided the exact experience you were looking for. So how do you eliminate the guesswork?
Enter place graphs. These visual portraits are compiled using the data trail left by millions of users in the digital universe every day. On the surface, a place graph documents our habits, tastes and interests by colour coding, but it may also potentially help us identify the destinations that best suit our individual requirements.
The fodder for place graphs comes from everyday activities and interactions. This includes mobile calls and texts, requests and comments made using mobile apps, and postings on websites and social media sites. All of this information—known as “big data”—is captured and stored by developers with the intention of extracting significant correlations and trends that contribute to the development of new and improved products and services. Problem is, big data is too large and complex to manage, let alone to analyze and visualize. Big data requires exceptional technologies to aptly manage its awkward bulk and a big data technologist like Matt Biddulph to crunch the data with a strategic purpose in mind.
One such strategic purpose has been to leverage big data to create a ‘neighbourhood isomorphism’. Or more simply put, a tool enabling users to find the equivalent of certain neighbourhoods in other cities, like the Times Square of Tokyo, or the Shibuya of Los Angeles. This tool would be invaluable for travellers. For example, a single man living in a trendy neighbourhood, packed with coffee shops, galleries and niche boutiques would presumably like to stay in a neighbourhood with similar features. It would be an unpleasant surprise, once he got to his vacation spot, to find himself in a family hotel located far from the city centre. A place graph would help him locate the right neighbourhood before booking.
Matt Biddulph began chipping away at this challenge while working on location applications for Nokia, the Finland-based mobile technology company, in the UK. He took several weeks worth of tracking requests for driving routes, made by their maps application, and ran the data through an open-source software for graph visualization and subsequently clustered the information based on connections between towns. The more often people drove between Town A and Town B, the greater the concentration on the graph. The assumption being that people from Town A enjoy Town B, more so than Town C or D, because it has attractions that appeal to their particular interests. Biddulph's exercise is somewhat broad because the correlation is not supported by more precise data—what is it exactly that attracts people from Town A to Town B?—but it did get the ball rolling.1
A ball that was picked up by the engineers at FourSquare, a location-based social networking application for mobile devices, for a hackday project in January 2012. They too wanted to discover the equivalent of familiar neighbourhoods in other cities, but they had far more specific information to work with. These intrepid big data technologists mapped over 1.5 billion check-ins made in 140,000 neighbourhoods all over the world. They also calculated the number of check-ins per category, banking on the assumption that neighbourhood profiles could be created based on the types of check-ins made there.
They were right. Neighbourhoods with high levels of social and cultural activities were clearly differentiating from neighbourhoods with more check-ins at offices or retail stores. So for example, New York City’s East Village numbered thousands of check-ins at bars, pizza places, yoga studios and karaoke joints. The engineers dove back into the data and found neighbourhoods in other parts of the country that displayed similar profiles. So, for example, if someone likes the East Village they will most likely enjoy the Telegraph Hill neighbourhood of San Francisco, since it has similar check-in patterns to the East Village.
The challenge of creating meaningful place graphs also inspired the research team at Livehoods, a research project that redraws major cities—New York City, San Francisco and Pittsburgh—according to the activity patterns of urban dwellers. Simply put, rather than looking at what’s happening within a specific neighbourhood, this project maps check-ins across the city to create areas—Livehoods—with the most dynamic activity.
The Livehoods are colour-coded on the map and as one is selected, the five most popular places and unique things to do appear on the screen. This model, according to its creators, teaches us about the character of the people who live and play there, and can presumably be used by persons living in other parts of the world to decide if that Livehood would suit their interests.
This escalating interest in place graphs signals its replacement of the social graph. Social graphs—or “the global mapping of everybody and how they’re related” —is strictly two-dimensional and illustrates how people know one another online. It is used by social media platforms to efficiently spread information from one person to the next. Although they are easier to tap into than place graphs, which require greater effort and processing, social graphs cannot draw a portrait of individual communities and reveal what draws them together.
One last example is the Fixie Index, a project by the San Francisco-based Priceonomics, that mined a database of 1.3 million bicycle listings to determine the largest markets for used bikes, price variations by region and areas that fixed-gear bikes, also known as ‘fixies’, are most popular.
Priceonomics anticipated that New York boroughs, which have more fixies per capita, should also have more hipsters, who love the model’s aesthetic appeal and design. When their findings indicated that such a correlation was true for New York City, they decided to apply their theory to the rest of the United States. It turns out, however, that their predictions were wrong for the rest of the country.
Unlike social graphs, place graphs have a built-in versatility, meaning that the strategic focus can be shifted and the big data can be analyzed again. So when faced with such results, the Priceonomics team decided to apply the data to other questions, including the top cities for cycling in the United States and whether cities with larger bike markets also tended to have bustling growth and a more educated population.
As more progress is made with place graphs and their application, they will begin to have greater influence over how we shop, travel and connect with other people; drawing communities that stretch well beyond the limits of where we live.