Tracking and optimizing mobile app installs: The newcomer’s guide to mobile app attribution
How do you measure mobile app marketing?
Mobile is now the default way we consume information. We’re spending an average of 4.3 hours a day on our smartphones, and that means we’re installing apps to access the information, services, messaging, products, and tools we need. That also means even now, a full 12 years after the launch of the first modern mobile app store, thousands of new apps are launched every single day on Google Play and the iOS App Store, joining the more than five million apps already published.
That means new games. New stores. New travel services (yes, someday we’ll be traveling again!). New on-demand services. And much, much more.
It also means a lot of competition. And that means publishers of mobile apps need what we used to call mobile app tracking, and now call mobile attribution, to understand and optimize the results of their mobile marketing campaigns.
The first thing you’ll want to know is that mobile attribution, or mobile app tracking, is very different than web-based marketing attribution. Because web marketers are adding mobile marketing skills to their toolsets, there are always smart, strategic marketers who need an introduction to mobile measurement. One of the key differences: while web attribution relies on cookies and pixel tags, mobile app attribution for Android and iOS apps works very differently.
This post is designed to help you if you’re new to mobile attribution and mobile app marketing to understand how mobile app attribution works, and how it differs from other forms of digital attribution.
See Singular’s mobile app attribution solution
Quick summary: You’re not in Kansas anymore!
Sophisticated data-centric app marketers use third-party mobile device app attribution systems for mobile app tracking, or mobile attribution. They are looking to track and attribute installs and remarketing events like re-engagements, just like web-based marketers might use an ad server like DoubleClick to track and attribute display ad campaigns on the desktop web.
Because Singular focuses on Android and iPhone app tracking or attribution for enterprise marketers, I’ll confine this explanation for how large app publishers get the breadth and depth of information necessary for attribution and campaign optimization at scale.
Web-Based Tracking Versus Mobile App Tracking
If you’ve been active in digital marketing for at least a few years, you’re familiar with both third-party cookies and pixel tags, and how they are used for marketing attribution. Third-party cookies are used by ad servers to anonymously identify users across a variety of touchpoints.
When a user’s browser requests a web page that will contain an advertiser’s promotional creative, the server places a cookie on their device and assigns it a unique user ID. When the user’s browser requests other pages containing ads from the advertiser, the cookie on the device delivers the unique ID to the server, so that the various ad views can be associated to the same user. And, when a purchase or sign-up or other conversion event happens, you can attribute the result to the ad campaign.
In this way, a third-party cookie is designed to associate all of a user’s exposure and actions in that browser to one ID. When a user converts, the completion action is tied back to specific marketing activity through the cookie ID.
Cookies are the go-to technology for most web tracking, but they have serious shortcomings because:
- Many consumers clear their cookies or set their browsers to block them
- Most cookies expire — often 28 or 30 days after they are initially placed
- In addition, many people use multiple browser instances, like one on a home and another on a work PC (plus their mobile phone too)
- A cookie ID by itself can only track a user’s activity within one browser instance
- Modern web browsers already deny third-party cookies (Safari, Firefox) or will soon do so (Chrome)
Mobile app attribution (or mobile app tracking)
Many web marketers who transition to the app industry assume that cookies and pixel tags will also play a central role in these mobile analytics use cases. But cookies are problematic in mobile: they don’t work in apps.
Both Google and Apple have deliberately designed their respective iOS and Android device app ecosystems to limit tracking with cookies. So if you want accurate information for attribution and optimization, you need an entirely different system.
That includes brand and performance mobile marketing in apps, and it specifically includes mobile marketing designed to increase app installs or downloads.
App downloads, app installs, and attribution
When a user installs a mobile application, they must first download the software. Once downloaded, it is installed on the device automatically. But an install can only be counted — and can only be confirmed and attributed — at the moment of first launch.
Think of all this as a multistep process:
- A user clicks an ad
- They visit the app’s page in the appropriate app store
- They click to download the app in the app store
- The app automatically installs on the device
- The user then opens the app for the first time
An iOS or Android install can be counted only after step five is completed. Thus, when advertisers refer to installs, what they really mean are apps that are installed AND THEN used once. This might sound like a pedantic nuance, but it’s important to understand this to get the full value from the rest of this post.
This additional step adds further complexity to an already complex and opaque system.
Get a personal demo of Singular’s mobile attribution tools
Let’s dig a little deeper into the steps and the way that desirable consumer actions like installs can be attributed to specific marketing activity. To do that, we need to draw some distinctions between the processes for Android and iOS, the two dominant operating systems. For both operating systems, the key task is to connect activity that occurs post-app-store with marketing activity that drove the visit to the app store. But beyond that basic similarity, there are some significant differences.
Google Play versus the iOS App Store
There are some process differences between iOS and Android app tracking that require explanation.
First, there is no measurement standard common across both systems. Google Play has been designed to enable a limited amount of tracking through Google’s sister product, Google Analytics. Google offers an SDK, or software development kit, which enables a referrer url and parameter to be passed to Google Analytics when a user clicks on an ad. If the user installs, the marketing activity can be easily credited. Using the referral parameter, you can track back all post-install activity to the referral source.
WIth Apple, things are a little more complicated. The App Store does not permit activity tracking within its walled garden. To connect specific installs to the marketing activity that drove them, you need a separate measurement solution that can trace the device ID to specific marketing activity.
Enter third-party mobile attribution, or mobile app tracking
Brands, of course, want to trace desirable marketing outcomes to the marketing activity that drove them. Third-party mobile attribution, or app tracking, like that offered as part of the Singular Unified Analytics solution, makes it possible to connect paid installs across Android and iOS back to the marketing programs that drove them.
To overcome the measurement issues discussed above — the multi-state app install process — most third-party mobile app tracking is delivered using a small piece of software called an SDK that is installed within the publisher’s app.
The SDK (Software Development Kit) connects the marketing activity to the desirable action like an install by matching the device advertising ID, which is a semi-permanent device identifier specifically developed to enable advertisers to measure marketing effectiveness.
There are two leading device IDs you will hear about a lot.
For iOS (Apple) devices, the ID is called the device identifier for advertisers, or the IDFA. IDFAs help an advertiser identify the specific phone where the ad action takes place. The Apple IDFA doesn’t change unless a user decides to change it in their phone settings. Few consumers feel a need to take this action, so IDFAs can offer a great foundation for a persistent and anonymized consumer profile.
Apple introduced IDFA to offer consumers a choice when it comes to interest-based advertising and tracking. IDFA is the only ID that Apple allows advertisers to use to understand the advertising actions on its phones, replacing the Universal Device ID, or UDID.
Note that as of iOS 14, the IDFA is going opt-in on a per-app basis, which means that it will be less available for mobile app attribution. Apple is providing the SKAdNetwork framework for privacy-safe attribution. Singular was the first mobile measurement company to declare support for SKAdNetwork, and has the leading solution for powerful mobile marketing measurement with SKAdNetwork.
Google Advertising ID (GAID)
The Android or Google equivalent to the IDFA is called a Google Advertising ID or GAID. It was previously referred to as the Android Advertising ID. These IDs share critical characteristics with IDFAs, in that they are persistent unless or until the user decides to change them. Before the Google Advertising ID, advertisers could track actions on Android phones using a device identifier called Android ID (or ANDI.) Users have the option to opt out of Android Advertising ID tracking, or to change their ID periodically.
That said, few consumers feel the need to take such action.
With most mobile app tracking, these advertising IDs are the means through which pre-app-store activity like ad clicks are connected to post-app-store activity like first launches.
Sometimes, though, it is not possible to link ad activity to an install with a device advertising ID.
This can be because the consumer has changed their advertising settings by choosing to Limit Ad Tracking on their device (iOS), or turn off ads personalization (Android). In those situations, a second attribution method called fingerprinting was historically deployed. Fingerprinting uses other signals to match marketing activity to an install. Such signals are collected by using non-personally identifiable information like IP addresses to infer a connection between an ad click and an install. The accuracy of matching through fingerprinting varies based on the signals available and utilized by the attribution provider, but such matching can often approach the level of accuracy available by matching device IDs.
Note that as of iOS 14.5, Apple does not allow the use of fingerprinting.
Attributing credit for an install
Advertisers want to know if their paid marketing programs are driving desirable customer actions like installs. Not every install, after all, is driven by advertising. Some installs will occur naturally, without advertising. Others will occur after a user clicks (or views) an ad. By connecting installs and ad clicks to the same device ID, you can identify which of your installs were driven by marketing activity.
But advertisers want to know more than just whether an install was driven by paid marketing.
They also want to know which programs, partners, and creatives were responsible for the install. When a user clicks only one ad prior to an install, it is very straightforward to connect that install to a specific campaign, media vendor, and execution.
But many installs occur after a user has taken multiple ad actions. For those instances, attribution providers play a critical role in pinpointing which marketing program and partner was ultimately responsible.
To attribute credit, attribution providers usually leverage last-click attribution, which attributes 100% of the credit for an install to the last partner that drove a click. Think of it as an assumption that the last click is the one that got the user to install.
Last-click attribution is the generally accepted model for attribution even though most people in the industry recognize that it has some shortcomings. It’s almost certain, for example, that clicks other than the last click prior to an install deserve some of the credit for that install. That said, last-click is viewed as the best available option given the complex measurement dynamics. And, in iOS 14 with SKAdNetwork, last-click attribution will literally be the only option for most devices.
One key issue is that many of the largest vendors are “self-attributing” ad networks. In fact, self-attributing networks make up 80% or more of ad spend across the industry.
Here’s how a third-party attribution provider can capture necessary information from a self-attributing network.
A self-attributing network must be queried after an install has occurred to determine whether they drove an ad click. By contrast, most partners automatically report all ad clicks immediately after they occur. Here is a simplified version of the steps for querying a self-attributing partner:
- The attribution provider captures the device ID when a user first launches a new app
- The attribution platform queries all self-attributing partners, asking whether the device ID was influenced by marketing activity on their platforms
- Any self-attributing partner responds by telling the attribution provider about the last click that occurred in their networks for a given app’s advertising, and when it occurred
- The attribution provider compares the timestamps from those self-attributing networks to the timing of installs recorded by other partners
- The most recent click prior to an install is credited with the install
- Since self-attributing networks do not report every click — only the most recent one, there would be no way to attribute installs using a method other than last-click
All partners accept last-click attribution as the accepted attribution method.
Third-party attribution tools for mobile app tracking
Without a third-party attribution platform, it would be impossible for marketers to understand which program and partner drove an install. Further, attribution partners ensure that an install is credited only once. This is so that an app publisher doesn’t double-pay or triple-pay multiple vendors for the same install.
To properly credit every paid install to a marketing program, your attribution provider must have visibility into activity across all your media partners. While many media partners will integrate with any attribution provider, some strictly limit such integrations to a small number of partners.
Attribution on iOS is changing. Find out more here …
Facebook, for example, limits access to attribution information, including timestamps, to a handful of companies that they have identified as high quality partners that adhere to stringent data and privacy regulations. Since Facebook often represents 20% to 40% of an advertiser’s spend, visibility into this leading platform is a critical component of a robust attribution offering.
These partners are called Facebook Mobile Measurement Partners, or MMPs.
Similarly, other self-attributing networks choose a select set of partners for attribution. Singular is very proud to offer its clients universal visibility into all of their app media providers, including Facebook, Google, Apple, Snap, Twitter, Pinterest, and more.
In addition to tracking installs, attribution providers also track post-install events like registrations, purchases, and the like. These can also be tracked back to the last-click media partner.
Marketers are increasingly investing time and ad spend to drive desirable consumer actions other than installs. For example, many brands now use advertising to drive incremental purchases or to get cart abandoners to return and purchase items that they added to a shopping cart but did not immediately buy.
Mobile app attribution or tracking providers can also be used to track in-app actions like purchases to the media partner that drove them.
Mobile app attribution or tracking without an SDK
Some brands do not want or permit the addition of SDKs into their apps. For these companies, server-to-server-based mobile app tracking offers a viable measurement alternative. In such cases, mobile app attribution tasks are still executed by the attribution provider, but the transmission of information becomes the responsibility of the advertiser.
Server to server integrations are fairly uncommon. For Singular, SDK-based attribution accounts for more than 70% of our client business.
An increasing number of advertisers are now measuring view-through attribution when no clicks precede an install. View-through identifies the media partner or campaign that displayed an ad to the user, even if there were no clicks. View-through measurement is generally only used when no click precedes the install.
Get in touch with Singular to learn more about our attribution services and our unified mobile analytics platform.