Main Methodologies for Attribution

To accurately measure and attribute the interactions (for example, installs and in-app events) that a user takes with your mobile app, Attribution Analytics uses four different types of attribution methods:

  • Google Install Referrer
  • Identifier Matching
  • Fingerprint Matching
  • Open URL with Click ID

While you can use each method to attribute app installs and in-app events, the methods you implement depend heavily on a combination of three factors:

  • Platform/App Store: iOS/Apple® iTunes® App Store, Android™/Google Play™ Store, Amazon Appstore for Android™, etc.
  • Engagement Type: Click-through vs. View-through measurement
  • Conversion Type: Install or In-App Event

The table below shows the four attribution methods and the circumstances under which you apply them.

Attributions_Method_1_-_New_Page.png
Additionally, Attribution Analytics completes the attribution process using the last-click/last-view model, where advertising partners who interacted with a user (but were not responsible for the last-click/last-view before the conversion) receive credit for an assist and/or non-windowed contribution instead.

For information about how Attribution Analytics prevents duplicate attributions, please visit Preventing Duplicate Attributions for Mobile App Installs. If you’re interested in how Attribution Analytics attributes contributions before the last click, please visit Assists and Non-windowed Contributions with Multi-Touch Attribution.

For more information on our attribution method waterfall – the sequence in which we implement the following attribution methods – please visit TUNE Attribution Method Waterfall.

This page provides details about our attribution methodology categorized by engagement type, conversion type, and platform/app store type.

 

Install Attribution

The following chart summarizes the various attribution methods that you can use to log install events. For iOS apps in the Apple iTunes App Store, you can use Identifier Matching and Fingerprint Matching for attribution.

With Android apps in Google Play, you can use the Google Install referrer to attribute installs on a click-basis; this method does not work with Android apps in other stores, however. For Android apps in other app stores (such as the Amazon Appstore for Android), you can use Identifier Matching and Fingerprint Matching for attribution.

Attributions_Method_2_-_New_Page.png
Since there can be up to 100 times more impressions than clicks, Attribution Analytics does not perform Fingerprint Matching on impression for view-through attribution (to minimize false positives and ensure statistical accuracy for attribution of clicks via Fingerprint Matching).

 

Event Attribution

When performing attribution for in-app events, Attribution Analytics uses a proprietary method that involves the Open URL with a Click ID (which works both with Android apps in Google Play and other Android stores, as well as with iOS apps in the Apple iTunes App Store). Using the Open URL with a Click ID provides 1:1 accuracy for click-through attribution, but because it requires a click to set the Open URL, you cannot use it for view-through attribution.

Attributions Method 3 - New Page (2)
Since both Fingerprint Matching and Open URL with a Click ID are applicable only for click-through attribution, and the Open URL with a Click ID provides 1:1 accuracy, Attribution Analytics uses Open URL with a Click ID for event attribution where Fingerprint Matching is not.

Similar to install attribution, you can use Identifier Matching with Android and iOS apps for both click-through and view-through attribution with 1:1 accuracy as well.

 

App-to-App vs. Web-to-App

The attribution methods may vary from the standards listed in the previous sections, depending on where users are coming from. App-to-app is the most common user experience where:

  1. User is in one app (in-app) such as the “Zynga Words with Friends” app and sees an ad in the app for another app (such as the “Supercell Clash of Clans” app).
  2. User clicks on the ad, which opens the respective App Store (or opens the “Clash of Clans” app if it’s an existing user who already installed it).
  3. User installs or opens the “Clash of Clans” app (assuming it’s not an existing user).

Web-to-app is the scenario where:

  1. User is in a web browser on their mobile device (such as Safari or Chrome on an iOS device).
  2. User sees an ad on a web page for the “Supercell Clash of Clans” app.
  3. User clicks on the ad, which opens the respective App Store (or opens the “Clash of Clans” app if it’s an existing user who already installed it).
  4. User installs the “Clash of Clans” app (assuming it’s not an existing user).

The main difference between app-to-app and web-to-app install attribution is that Attribution Analytics can’t use Identifier Matching for web-to-app scenarios (only Fingerprint Matching with iOS). With Google Play, Attribution Analytics leverages the Google Install Referrer string for both app-to-app and web-to-app install attribution.

Install Attribution with App-to-App vs. Web-to-App

Attributions Method 4 - New Page (1)

While methods vary between platforms and stores when performing install attribution , the methods are very similar between Android and iOS apps (in both app-to-app and web-to-app scenarios). This similarity leads to a very big advantage when it’s time to attribute events via the Open URL with a Click ID method (to attribute app-to-app and web-to-app install events).

Event Attribution with App-to-App vs. Web-to-App

Attributions Method 5 - New Page (3)

Google Install Referrer

With Google Play, you can include an install referrer string in your TUNE links to uniquely identify an advertiser or advertising partner (publisher). Attribution Analytics leverages this capability to pass Google a Unique User Session Id (Click ID) generated when a user clicks on the TUNE links. Since this value is set on ad click and redirects users to the Google Play Store, Attribution Analytics can only perform click-based attribution (and not impression-based/view-through attribution).

Google_Install_Referrer_-_New_Page.png
Attribution using the Google Install Referrer method matches at 1:1 accuracy with a loss of approximately 10%. This attribution method also supports a customizable attribution window.

The following code shows an example of how Attribution Analytics appends its Click ID into the Google Install Referrer parameter:

https://play.google.com/store/apps/details?id=com.hellochatty&referrer=mat_click_id%3DCLICK-ID

On install (first app open) by the user, the TUNE SDK collects the Google Install Referrer value (which includes the Click ID in the query string). The Attribution Analytics platform takes the Click ID, finds the corresponding click record, and attributes the install to the click accordingly. For information about the Google Install Referrer, please visit How Android Install Referrer Works.

For information on how Google Play Android Apps are attributed using the Google Install Referrer and Identifier Matching (but not device fingerprinting), please visit Attribution for Google Play Android Apps.

Attribution based on the Google Install Referrer is compatible with Google Play only. The Apple iTunes App Store, Amazon Appstore, or any other app stores do not provide or support an install referrer string. There are some intricacies when using the Google Install Referrer method with other attribution methods. For information about using the Google Install Referrer with other attribution methods, please visit Dynamics of Using the Google Play Install Referrer with Multi-Touch Attribution.

There is also an issue with the Google Install Referrer not working properly when the web version of Google Play is used. Therefore, using Direct Tracking links for Android is not a fool-proof method of attributing Android app installs on Google Play. You should use Redirect Links instead to allow the Attribution Analytics platform to perform attribution using the other two methods.

 

Identifier Matching

Attribution using unique identifiers involves matching them from the conversion (install or post-install event) to a click or impression. On conversion, the TUNE SDK collects multiple unique identifiers for attribution and uses all of them to search for clicks and impressions with matching identifiers.

Main_Methodologies_for_Attribution_7_Identifier_Matching_Process_GEF.png
 

Identifier Matching attributes matches at 1:1 accuracy with absolutely no loss. This attribution method also supports a customizable attribution window. You can modify the attribution window in the Attribution Analytics platform at the account level, and at the partner level. For more information, read our article on Setting Your Attribution Windows.

For example, an advertising partner includes “ios_ifa=AAAAAAAAA-BBBB-CCCC-1111-222222220000” in their TUNE link. On install, the TUNE SDK collects the IDFA of “AAAAAAAAA-BBBB-CCCC-1111-222222220000” and sends it to the Attribution Analytics platform, which receives the install request and searches for all clicks and impressions that match the IDFA. The results include the click by the partner who included the IDFA (of “AAAAAAAAA-BBBB-CCCC-1111-222222220000” in the TUNE link. Then Attribution Analytics attributes the install to this partner based on the unique identifier that matched from install to click. The same logic applies for impression-based/view-through attribution.

Unique Identifier Matching enables 1:1 accuracy of clicks or impressions to conversions where identifiers are passed app-to-app. Our integrated advertising partners are passing unique identifiers into TUNE links where possible, so they can use this method of attribution.

When an advertising partner specifies unique identifiers in a TUNE link, they can improve performance by loading the TUNE link on the server-side instead of redirecting on the client-side. For information about loading TUNE link on the server-side, read our developer documentation on server-side click notifications.

The most common device identifiers include the Apple Identifier for Advertisers (IFA) and the Google Advertising Identifier (AID). For more information about these identifiers, read our Unique Identifiers for Attribution article.

 

Fingerprint Matching

Fingerprint Matching pulls basic (and not always 100% unique) information (such as IP address) from mobile device headers to connect a user from ad-click to conversion. Fingerprint Matching occurs asynchronously in the background of the mobile app, which does not interfere with the user experience by forcing a browser to open.

device_fingerprinting_methodology_-_New_Page.png
Fingerprint Matching works by redirecting a user through a TUNE link and collecting the publicly available HTTP headers about the device. Attribution Analytics uses this information to create a unique fingerprint about the device (whose user clicked a TUNE link). When a user installs the mobile app, the TUNE SDK collects the same data points from within the mobile app and sends it to the Attribution Analytics platform, which searches for clicks with the matching fingerprint of the install.

Since attribution via the other three methods provide 1:1 accuracy and Fingerprint Matching relies on statistical probability, the three other methods of attribution always trump Fingerprint Matching. So if Attribution Analytics can perform attribution using a method with 1:1 accuracy, it may attribute an advertising partner (who has an older impression or click) with an install or event over an advertising partner with a more recent click with matching fingerprint. In addition, Attribution Analytics does not use Fingerprint Matching to credit Assists or Non-windowed Contributions with multi-touch attribution.

The attribution window for Fingerprint Matching is set to 24 hours by default. You can also modify this setting to reflect a 1/3/6/12 hour window or disable it all together.

For more information about Fingerprint Matching, read our Device Fingerprinting Methodology article.

 

Open URL with Click ID

As described in the previous tables, the Open URL with Click ID method provides 1:1 accuracy for attribution of events. It works for both app-to-app and web-to-app scenarios to attribute on a click-through basis by using the URL (deep link) that initiated the “app open”. This method does not work for install attribution because on the first “app open” after the install, the deep link is inaccessible.

Open URL with Click ID - New Page

Attribution using an Open URL with Click ID also matches 1:1 accuracy with absolutely no loss. It also supports a customizable attribution window.

When users click on an ad and they already have the app installed, they are deep linked into the app (through the Deep Link URL) causing the app to open. In the deep link URL, Attribution Analytics includes the ID of the click (Click ID) as shown in the following example.

app://details?action=newchat?mat_click_id=CLICK-ID

When users open the app, the TUNE SDK collects the Open URL (which includes the Click ID in the query string). Then the Attribution Analytics platform takes the Click ID, finds the corresponding click record, and attributes the event to the click accordingly.

When an advertising partner uses this method of attribution for events, they can use server-side click notifications instead. For more information, read our developer documentation on server-side click notifications.

4 Comments

Leave a reply