Audience Analysis While Respecting Consumer Privacy

By Simon Khalaf, CEO

A Letter from Simon Khalaf, Flurry President and CEO

Two years ago, we were a scrappy start-up with a new idea.  Today, as our reach exceeds 100 million unique monthly devices, we have become a bona fide leader in the new era of mobile computing, proving to be as significant as the PC revolution of the 80s and Internet boom of the 90s.

At the same time, Flurry is more deeply connected to this new mobile era than the average company.  As a provider of application analytics to over 35,000 companies whose combined end user sessions total more than 5 billion monthly sessions across more than a total of 180 million unique devices, Flurry manages a lot of data.  This is a large responsibility.

Personally, as someone who has worked in software security for fifteen years and online advertising for another five, I personally appreciate both the sensitivity and the opportunity of managing this data set.  And this is what I want to write about today.

Across our industry, there has been increasing discussion about protecting consumer privacy.  Flurry is proud to hold high standards in this regard.

Flurry and Consumer Data

To start, it’s important to understand that Flurry does not have direct relationships with consumers.  Instead, we have relationships with our customers, mobile application developers, who agree to our Terms of Service and Privacy Policy before integrating our code into theirs.  What we collect, on their behalf, as consumers use their applications is anonymous, aggregated data.  We do not know anything about the individuals behind the device upon which our customers’ apps are running – not their name, physical address, phone number, email address, etc.  Collectively, this kind of data is called “personally identifiable information” or PII for short.  Flurry does not collect, store, buy, trade or sell consumer PII.

Flurry Customer Access to Data

Flurry collects and aggregates anonymous usage data for its customers, one app at a time.  After collecting and processing this data, we present that back to our customers in a data dashboard for them to better understand how consumers interact with their live applications.  Most often, customers use Flurry Analytics to improve their application experience to increase consumer satisfaction, retention and revenue.  On average, good apps earn consumer trust, and companies using Flurry Analytics care about quality.  Toward that end, Flurry has played a vital role in helping increase app quality across the mobile app ecosystem.

Flurry’s Own Use of Data

Frequently, we are asked about our ability and intention to build consumer profiles from the data we collect for ad placement, application recommendations and other services.  Let’s face it: all companies want efficient marketing, which includes the ability to match the right consumer to the right product at the right time, for the least amount of money.  This is where audience analysis and reach can help, and Flurry is committed to helping its customers improve end-user acquisition, retention and monetization.  However, to protect consumer privacy, we are very careful about our approach in doing so.  So how can Flurry reach end-users while respecting privacy?  In short, Flurry does not profile consumers.  Rather, it profiles applications.

Application Profiling

Think of applications like magazines.  If an advertiser wants to reach teen females, she will run an ad in Teen Vogue.  Males 25 – 49?  Try Wired magazine.   Like in the magazine industry, Flurry is able to extrapolate the composition of an application audience without knowing individual consumer information.  We start with aggregate behavioral information such as category interest.  For example, we know what proportion of an application’s active users also uses different app categories such as news or sports.  We can also add geographic information, which we obfuscate to the city level.

Then, if available, we start adding demographic information.  We do this using a small sample.  This simply works like exit polls on election nights.  Pollsters ask a few people how they voted and then project the election results based on this sample.  Similarly, Flurry looks at a usage sample among our own data set and extrapolates.  As a simple example, if our sample for an app is made up of 40 males and 60 females, then we say that 60% of that application’s user base is female and 40% are male.

Audience Analysis and Recommendations

Audience analysis is where our algorithms come into play.  When a request comes in to reach a specific audience we look at the profiles of the applications typically used by that kind of audience.  Then, the intersection of these profiles is used for app recommendations.  For example, let’s consider two apps that appeal to two kinds of audiences.  The first one is profiled as “80% female, 60% U.S.-based and 80% social gaming” and the second one is profiled as “95% female, 78% U.S.-based and 30% social gaming.”  In total, we would then reach apps that are “95% female, 78% U.S.-based" and appeal to “those who like social gaming 80% of the time.”  As we run this over thousands of applications our confidence in accuracy increases.

Accuracy vs. Privacy

Some may argue that this kind of extrapolation is not accurate enough.  However, Flurry is confident that it has the richest data set and the best recommendation engines in the industry. At Flurry, we believe it is unnecessary to force a trade-off between effective advertising and consumer privacy. On the contrary, with good technology and a large enough data set, we can deliver great recommendations without compromising privacy.  Not only is this exactly what we believe, but also exactly what we provide.