Mobile Data Primer: App Install Data

The history of customer targeting has been one of moving from the broadcast methods available to the earliest marketers — television, radio, newspapers, billboards — to the modern, data-driven methods available in the digital age. But even in the Internet era, most marketers have rarely known much beyond general demographics about a person who visited a webpage, and couldn’t even stitch together a good picture of what other webpages the person might have visited.

But mobile has changed the story completely, delivering marketers a way to draw deep, rich pictures of the customers they’d like to target.

App install data, or app graph data, can offer detailed information about the psychographics of mobile handset users, making it a valuable resource for mobile marketers and publishers alike. As the name implies, app install data is simply information about which apps a user (as identified by their anonymized Mobile Ad ID) has installed on their mobile device.

Let’s take a look at what we know about those users, how we know it, and the best ways to leverage that data.

A Look Inside App Install Data

App install data creates what we refer to at Twine as a very clean “rectangular data set.” Let’s go back to algebra’s coordinate system to understand this better: with a large set of app install data, you can place on the Y-axis all your Mobile Ad IDs (MAIDs; for more information on what that is, see our previous explainer here) and on the X-axis the potential universe of millions of available mobile apps. Each data point on this graph is the intersection of one app install per MAID.

For data scientists, this large rectangle of data creates a stable base to do a lot of very interesting targeting and modeling work. Data models need both scale and consistency of data to do their best work — the more stable the underlying data set, the better the ability to make accurate predictions. One of the largest rectangular data sets for marketing is owned by Facebook. They have hundreds of millions of MAIDs on one axis and demographic, psychographic, and behavioral data on the other axis. The result is that they have a baseline data set to do data modeling better than almost anyone else.

The data itself typically comes from one of four sources:

SDKs with large install footprints by their nature can start to identify which mobile devices have which apps installed. This data is incomplete as no SDK has even 25% market penetration, so you can only get a sampling of apps. SDKs also generally only cover certain categories of apps, resulting in selection bias.

On Android, a full set of app install data is accessible if the consumer has consented to share this information and the app making the call has a valid business reason for collecting it.

With ownership over the customer relationship and deep access to mobile devices and network traffic, wireless carriers have ready access to this data.

When ad-powered apps run ads, ad buyers can make associations between apps and mobile handsets. They can also presumably intuit that a user has used an app at a specific day and time. This data source can be very problematic, as can any ad-based or bid stream-based data set is it is vulnerable to fraud. Many unscrupulous publishers send fake app information across the bid request to make their ad inventory worth a lot more.

The most sophisticated advertisers tend to be most wary of bid stream data due to fraud and privacy concerns.

A Goldmine of Information — The App Graph

Never before have mobile users had access to such specialized and personalized software, with apps for everything from gaming and fitness tracking to on-demand services like Uber and Postmates. On average, a typical device has more than 70 apps installed on it. The cross-section of apps someone has installed leaves a very unique fingerprint of their demographic and psychographic profile — a unique app graph.

With this data you can learn things about a user like:

  • What they like to read
  • What brands and products they have loyalty programs with
  • Their leisure or non-work activities
  • Companies they use for basic services like insurance, banking, transportation

With this information you can also start to paint a detailed profile for precise targeting of niche audiences. For example, you could target a bargain shopper with expendable income based on app usage from big box stores, a monthly wine subscription and deal-travel apps.

Recency of install can be a key factor for this data set, revealing even more detail about a user’s immediate interests. For example, users who have recently installed a real estate app may likely be in market for real estate services. And app install data can go even deeper than install, into the realm of behavioral data, including information on app usage or in-app purchases.

App Install and Privacy

As with any MAID-connected data set, users have the power to block the collection and sharing of their app install data. Both the Android and iOS systems allow users to limit ad tracking on their mobile devices, which in the case of iOS prevents the collection of the MAID entirely. And both Google and Apple are very clear in their terms of services that app install data cannot be collected without a specific business purpose and permission from the user.

Putting App Install Data to Use

Here are some ways that marketers today are using app install data:

Marketers can target the customers of key competitors who have installed that competitor’s mobile app. Wells Fargo can accurately target and advertise to users of the Bank of America mobile app, competing for market share directly.

Beyond specific competitors, marketers can accurately target broad interest groups by the mobile apps that audience typically installs. If a major men’s sportswear brand is trying to reach large numbers of sports fans, targeting users who have installed network apps like ESPN, Fox Sports and league apps like MLB and NBA is a good strategy. Broad interest categories like finance, entertainment and travel can be reached at scale with app install data.

App graph data is also effective at reaching very niche audience groups by targeting a strategic cross-section of related apps. For example, an SBA Lender could target small business owners via people who have installed Quickbooks Self Employed, eBay Seller, Amazon Fulfillment, or Etsy Seller. Marketers could reach fashionistas with expendable income with apps like Poshmark, Zara, and the Outnet.

While there are some shortcomings when it comes to app install data, we have a lot to look forward to. Over time the data community will have more access at scale to app usage behavior and in-app events. This means the signals we are using to target audiences and learn about our customers will become richer and more accurate. With a large app install data set, data companies have a competitive tool to build highly effective and differentiated data products to compete with the largest mobile ad solutions.

The Mobile Source

A round-up of information about mobile data, mobile marketing, and programmatic advertising

Elliott Easterling

Written by

Founder and CEO of TrueData. www.TrueData.co

The Mobile Source

A round-up of information about mobile data, mobile marketing, and programmatic advertising