Harnessing the Power of Mobile Data

The numbers prove it: Mobile is dominating. Mobile devices are now the main source of Google searches, and mobile app marketplaces are on track to do over $76 billion worth of business worldwide next year. And, unsurprisingly, with all that usage comes lots and lots of data.

Hailing from either the hardware of the mobile device itself or from mobile app usage (which grew by 58% in 2015), these increasingly large and varied datasets are a fertile ground for data scientists. So is there an inherent difference between data science for mobile and data science for desktop? No, there isn’t. But the mobile platform presents new opportunities for data scientists to add value, including the five areas listed below.

Connectivity

Ideally, the 7.6 billion mobile connections in the world (yes, mobile devices now outnumber human beings, according to GSMA Intelligence’s live count) would be on a 4G network at all times. But in reality, those devices are often operating on 3G or something slower.

Data science can help ease any connectivity-based struggles your users might be experiencing by showing you how to be smarter about managing your network. For instance, data from failure logs will tell you how often users don’t complete a download due to a dropped connection. That information will help you pinpoint the exact size your downloads should be to minimize the problem.

Input

For the average user, typing on a smartphone screen is more difficult than typing on a full-sized keyboard — hence the endless (and hilarious) lists of “autocorrect fails” on the Internet. This presents a challenge for mobile sites or apps that utilize forms for functions like job application processing or lead scoring.

If you’re using lead scoring to rank sales prospects, you don’t want to devalue a lead due to a typo. A machine learning approach to lead scoring would take the method of entry into account, and help you identify possible errors. The result? Your lead scoring is more accurate, regardless of the device it came from.

Device Attributes

You can’t change the size of the screen or processor speed your users have, but you can use data science to improve the mobile experience. A/B testing allows you to compare versions of your mobile application or site across similar groups of users to see which ones perform the way you want them too. Do users with larger screens perform better or worse? How do different processor types affect the experience?

In fact, you can answer a lot of mobile product questions using techniques like correlation, clustering or logistic regression. For instance, you can see which version of your marketing copy makes the biggest impact, which color works best on your call-to-action buttons, or how fast the gameplay on your app should be to satisfy your users. All of these questions can be segmented by device attributes to ensure that customer behavior is optimized across platforms.

Environmental and Location Data  

Since mobile devices often travel with their users, a whole new type of data is potentially available for analysis. Accelerometer and GPS data will give you the opportunity to predict things like motion, orientation, and location.

Built-in and accessory-based sensor data means the possibilities here are nearly endless: Most Android devices have built-in sensors that measure things like motion and environmental conditions (for example, acceleration forces or humidity), meaning a data scientist could hypothetically track an occurrence using sensor data (like when a user drives over a pothole) and then plot those occurrences on a map using GPS data.  

Data Visualization

Most mobile applications use data visualization in some capacity, no matter what kind of app they are. For example, fitness and finance applications are two places where data visualizations inevitably crop up to tell the user how many steps he or she took or where his or her money was spent.

But displaying data on a mobile platform is challenging. Many of the data visualization features that a user may encounter on a desktop (like the ability to roll over images) might not translate to a mobile device. Looking at user interactions, a data scientist can tell you which data visualizations are having the best effect on users (which did they spend the most time on, etc.) and if your app is maintaining the integrity of the data it is meant to portray.

It doesn't stop there. The possibilities mobile data holds are nearly endless. And a data scientist's take on that data might very well change your users' mobile experience for the better.