Measuring Cross-Device Attribution

With the rise of multi-device usage, users now bounce from device to device and it is not always the case that the device they started the conversion path on (for example, saw an ad on their tablet) is the same device the final conversion occurred on (for example, downloaded an app on their smartphone). As an advertiser, you want to run campaigns across multiple devices and platforms to gain the widest reach. Of course, you also want to measure these cross-device campaigns, but how?

To attribute cross-device campaigns, Attribution Analytics implements deterministic device pairing, which links the device on which an ad is viewed with the resulting usage of the correlated mobile app on another device (the linking or attribution uses a specific user identifier that is common to both devices). Since the same identifier is used when measuring the click of the ad, events by that user on another device can be attributed back to the ad click.

Attribution Analytics supports the following unique user identifiers for deterministic cross-device attribution:

  • user email – The email of the user defined by the advertiser.
  • user id – The ID of the user defined by the advertiser.
  • username – The username of the user defined by the advertiser.
  • facebook user id – The ID of the user defined by Facebook.
  • google user id – The ID of the user defined by Google.
  • twitter user id – The ID of the user defined by Twitter.

Note: We recommend that you hash all data you consider to be “personal data”- including user names and email addresses. Regardless, TUNE will hash user names and user email addresses that are uploaded to the Attribution Analytics platform. For German clients, we will also hash and salt IP address data stored in our logs. For more information about handling personal data, read our article on the Importance of Securing Data in the TMC.

Normally, third-parties use a unique email address for deterministic cross-device attribution because third-parties can’t interact with or target the user through the advertiser’s user ID or username for the user (they don’t have access to that information). The user ID and username are used solely by the advertiser to measure internal campaigns, since they can interact or target the user with these two user identifiers.

Measuring_Cross_Device_Attribution_1_User_Identifiers_Process.png
Through deterministic device pairing, cross-device attribution can be implemented to measure new user acquisition campaigns and also re-engagement campaigns for existing users.

Cross-device Attribution for User Acquisition Campaigns

When promoting user acquisition campaigns, you want to attribute the install back to the cross-device ad that drove the install. However, in almost all cases, the app doesn’t have access to the user email address, the user ID you’ve defined for them, or the username defined by the user on first app open (install). Therefore, cross-device attribution for user acquisition campaigns must be conducted on the registration or first login event, when those identifiers are available to be collected by Attribution Analytics.

Consider a loyal user of a desktop web site of a company (such as Nordstrom). The user is always authenticated and logged in when browsing Nordstrom.com. Nordstrom wants to run various internal campaigns on their desktop web site, promoting users to install their mobile apps. Nordstrom displays an ad encouraging them to download the mobile app via their desktop web site. When a user clicks on the ad, the user is redirected via TUNE link which includes a user ID, user email, or username. The user is then redirected to the app store on their desktop computer to install the mobile app.

To measure and attribute this cross-device promotion, the Attribution Analytics SDK must be set up to collect the user ID you’ve defined for that user, the user’s email, and/or the username defined by the user themself on the registration or login event. After the user opens the app for the first time (install) on their device and registers or logs in, these values are collected and matched against all of the measured clicks of the internal advertising campaign. After the click that contains the same user identifier is found, the user’s actions (the registration or login event) are attributed back to the specific internal advertising campaign.

Measuring_Cross_Device_Attribution_2_Cross_Device_Process.png
While the above scenario describes a user originating on a desktop computer, the device of origin could be a laptop, tablet, or any other device where they have the opportunity to see an ad that encourages them to open the app. The above user could just as easily be browsing the Nordstrom.com web site on their tablet and then decide to enage with the mobile app on their smartphone.

Some apps, however, are developed so that users aren’t required to actually register with the app itself. Users can use Facebook connect or another authentication system to login for the first time. The use of authentication systems create a unique situation where we do not have records of that user ever having registered with the app, but there is a request to measure a login event for the first time. To make sure we don’t miss these types of scenarios, the same attribution process is completed in an effort to measure any relevant user acquisition campaigns; in this case the fact that a user did download the app for the first time, but logged in using credentials over a third-party authentication system.

Register_vs_First_Login_-_New_Page.png
Alternatively, a user who installs an app for the first time on their device may also have a valid account created from another device. If the user has created an account on another device, but is logging in for the first time on their mobile app, we still want to provide attribution insight on where the user came from before installing the app, if doing so is possible.

By default, Attribution Analytics only includes clicks that occur directly before the event you want to attribute (in this case, the install). So technically, impressions and clicks from another device wouldn’t be included when attributing the Registration or first Login event, because those clicks occurred before the preceding install event.

However, Attribution Analytics has extended the attribution window for the Registration and first Login event to include impressions and clicks that occur before the install event. After the user ID, user email, or username is collected on the registration or login event, the user identifier can be matched to the ad click (with the same identifier) that generated the install, completing the cross-device attribution.

clicks_before_registration___login_events_-_New_Page.png
Thus, instead of simply using the registration and first login events for Post Install Insight into New vs. Existing Users, these events can be used for deterministic cross-device attribution of user acquisition campaigns.

Cross-device Attribution for Re-engagement Campaigns

Cross-device attribution can also be conducted on Re-engagement campaigns; for example when a user already has the mobile app installed on one device, sees an ad on another device, and then opens the app on the first device.

Consider a user who is browsing and logged in to Nordstrom.com via their desktop computer at work. Since the user is logged into Nordstrom.com with their user profile, Nordstrom can encourage the user to click on ads relevant to offers available in their mobile app. On the user’s way home, the user opens the Nordstrom mobile app on their smartphone.

Measuring_Cross_Device_Attribution_5_Cross_Device_Process2.png
To measure and attribute this cross-device re-engagement promotion,on app open, the user ID, user email, and/or username should be collected with the open event via the Attribution Analytics SDK to measure the event. When the open event is processed via “measureSession“, our attribution process searches for valid clicks by user ID, user email, and username. If a matching click is found, then Attribution Analytics attributes the event accordingly with re-engagement—linking the open and subsequent events with the partner information from the ad click.

Note that the “open” event is measured automatically by measuring the session. Measuring sessions automatically determines when installs, update and/or opens are logged.

As mentioned previously, the attribution window for Re-engagement campaigns, by default, only includes impressions and clicks that occur directly before the event you want to attribute (and after any previous events). As you can see in the diagram below, only impressions and clicks after the Registration or first Login are included to attribute the 2nd Open event (the install event is the first app open).

clicks_before_registration___login_events_-_New_Page_1_.png
Following this logic, if a third app open event occurred, then only impressions and clicks after the last event prior to the 3rd Open are included to attribute it. And so on. This process of only including impressions and clicks after the last prior event ensures that partners are getting credit for events appropriately.

This same process of deterministic cross-device attribution can also be used to attribute advertising campaigns via email and SMS, even on the same device. For example, when an advertiser sends an email to their users, they can include the user email address in the TUNE link which can then be used to measure the click. When the user opens the app, the click can be attributed 1:1 to the open event (along with subsequent events). Similarly, if the user email address is known when sending an SMS message to the user, then the email address can be included in the TUNE link as well. When the user opens the app from the SMS message, the open and subsequent events can be attributed 1:1.

Measuring_Cross_Device_Attribution_7_Cross_Device_Re_Engagement_Process.png
To learn more about measuring app-to-app cross promotion campaigns (not desktop-to-mobile app or Email/SMS-to-mobile app), read our article on Measuring App-to-App Cross Promotions.

Cross-Device Attribution Requirements

Before you can begin using deterministic cross-device attribution, make sure that you fulfill the following requirements:

  • TUNE links include at least one of the user identifiers for cross-device attribution.
  • The SDK is implemented to collect at least one of the user identifiers for cross-device attribution with the Registration or Login event for User Acquisition campaigns and/or with measuring sessions for Re-engagement campaigns. Note that the user identifiers included in the TUNE links on click are the same parameters that must be set when measuring the event for cross-device attribution (else Attribution Analytics cannot link the event with the click and/or impression).

Make Sure TUNE Links Include Unique Identifiers for Cross-Device

To attribute an install from a User Acquisition campaign cross-device, the TUNE link measuring the click needs to include the unique identifier the advertiser collects via SDK on conversion.

User identifiers for cross-device attribution:

  • user email – The email of the user defined by the advertiser.
  • user id – The ID of the user defined by the advertiser.
  • username – The username of the user defined by the advertiser.
  • facebook user id – The ID of the user defined by Facebook.
  • google user id – The ID of the user defined by Google.
  • twitter user id – The ID of the user defined by Twitter.

Note: We recommend that you hash all data you consider to be “personal data”- including user names and email addresses. Regardless, TUNE will hash user names and user email addresses that are uploaded to the Attribution Analytics platform. For German clients, we will also hash and salt IP address data stored in our logs.

The values of the parameters above can take the following formats:

  • MD5 – MD5 hash of string
  • SHA1 – SHA1 hash of string
  • SHA256 – SHA256 hash of string.

As noted above, if you are a third-party partner, then you’ll likely only be able to include the user email address using our “user_email” parameter.

If you are the advertiser running an internal campaign, then you might be able to specify the user ID that you already defined for them or their username.

Ensure Your Attribution Analytics SDK Collects the Corresponding Values

The unique identifier specified in the TUNE link should be one of the same identifiers already being collected by the SDK. This way the click on the TUNE link is associated 1:1 with the Registration/first Login event for User Acquisition campaigns and 1:1 with the Open event for Re-engagement campaigns (by matching the unique identifier from the event to a matching click with the same unique identifier). If the SDK is only implemented to collect user ID, then attribution cannot be performed with user email address nor username. Whichever user parameters are collected on conversion, you should set these same parameters in the TUNE link on click so Attribution Analytics can use it to perform attribution.

For User Acquisition Campaigns

To enable cross-device attribution for User Acquisition campaigns, the user ID, user email address, and/or username must be set prior to measuring the Registration or Login event.

To do so, implement the appropriate code in the Attribution Analytics SDK.

  • If you want to measure the Registration event, implement the SDK code found here.
  • If you want to measure the Login event, implement the SDK code found here.

For Re-engagement Campaigns

To enable cross-device attribution for Re-engagement campaigns, the user ID, user email address, and/or username must be set prior to measuring the Session (Android: but after the onResume (open) function). Note that these values cannot be set when measuring the first app open for the install.

To do so, implement the appropriate code in the Attribution Analytics SDK.

  • Android SDK
  • super.onResume();
    mobileAppTracker.setUserId(userId);
    mobileAppTracker.setUserEmail(userEmail);
    mobileAppTracker.setUserName(userName);
    mobileAppTracker.measureSession();
    
  • iOS SDK
  • [MobileAppTracker setUserId:userId];
    [MobileAppTracker setUserEmail:userEmail];
    [MobileAppTracker setUserName:userName];
    [MobileAppTracker measureSession];
    

For more information about attributing in-app events, read our developer documentation on Measuring In-App Events with the SDK.

1 Comment

Leave a reply