Location is the first step to understanding an end user’s real life context. So what’s a CTO to do when he receives mobile IPs which are both inaccurate and unreliable, and he has to turn them into real life places and experiences? Never one to shy away from a challenge, Oded walks through the Ubimo method for turning a mobile IP into a real life location.
Ad dollars are rapidly shifting to mobile, as advertisers communicate with their audiences throughout their day, at any location. At Ubimo we have been advocating for marketers and advertisers to use location and local data in order to define an audience’s context, and serve them with the most relevant ad.
The cornerstone for solving the mystery of an end-user’s context, is location. We’ve been working with the MMA to encourage all parties to work towards improving location accuracy for mobile – for more details, check out our collaborative white paper.
Location is the pivot point of our data offering and so we have built the tools and methodologies to improve the location accuracy of each bid, and to subsequently extend the reach of our location data. In this post we will tackle the right way to translate an IP address on mobile, to a physical location in the real world.
IP2Location for web
For traditional web advertising, IP addresses are the key to understanding the end user’s location. The IP address, assigned to a network by the ISP, directly correlates to the router’s physical location. Whilst most routers will only use the same IP address for an average of a week, the ISP will always assign a new IP from a pre-defined block of addresses for that specific region or city, so that traditional IP addresses are considered good indicators of location at city level.
There are a few providers of IP2Location tables, and they enable technology partners to derive a real world location from an IP. Typically however, these providers can only provide an accuracy level of a few miles radius, which could be acceptable for city level granularity, but is not specific enough for venue level targeting.
IP addresses in a mobile world
An obvious challenge for anyone dealing with location on mobile devices is the inability to use IP as a reference for the device’s location. With the advent of mobile advertising, platform providers attempted to use mobile IP addresses to determine location with the same methodology as IP addresses assigned to desktop. It became immediately apparent that these methodologies are not transferable to mobile. IP addresses are usually assigned to a mobile device with no reference to its real world location. Even when there is consideration for the end user’s location, the IP assignment becomes irrelevant in terms of location as soon as the mobile device moves to another location. In practise that means that an IP address assigned to a user whilst they’re in NY, might still be used by their device when they travel to Florida in the same week.
Consequently, an IP address is useless for determining real world location on mobile. For devices connected with both GPS and wifi, the GPS will always identify the device’s location in real time, but it is not uncommon for the IP address to show that the end user is in a different city or even in an entirely different state. So a tourist in Florida relaxing at a beach bar and browsing on their cell phone, might be using an IP which would tell us that they are in Detroit.
Mobile is in our DNA – our platform was built entirely with mobile in mind, and we address the challenges of mobile marketing without the constraints of our desktop predecessor platforms. We enrich location data with local data, to enable our clients to understand what experiences are taking place at any given location. If someone is with their children at a holiday market in the snow, we can serve them a dynamic ad which will adapt to suit them, based on their real time experience. A bid which comes in with false location information is unusable for a contextual targeting platform like ours, so we’re compelled to address this critical challenge and make sure that we maximize available inventory.
As part of our commitment to providing our clients with as much valuable inventory as possible, we’ve come up with a simple solution to solve the challenge of inaccurate location data derived from IP.
IP to Real-Life Location
Traditional IP2Location services tend not to work on mobile IPs. On the rare occasions when IP2Location methods provide a result, the process will still yield suboptimal accuracy. With this in mind, we decided to create our own mapping between an IP to a location using GPS data. By leveraging impressions with GPS location data, we designed an adaptive model to map IP addresses to locations. Our system continuously collects fresh traffic containing a valid GPS location and an IP address, and updates the model accordingly. The model reflects the data seen in the last few days, and is used to estimate real world locations for the next day’s traffic. Whenever we receive a request which includes an IP address but no location information, we are able to accurately assign the bid with a real world location. Whenever we receive a request which includes an IP address that includes location information, we use it to evaluate and improve the accuracy of our online model.