Mobile Marketing: Discrepancies in Installation Statistics

Original author: Roman Garbar
  • Transfer
Show me two dashboards with the same data, and I will say that one of them lies or at least blindly copies numbers. Like it or not, discrepancies are inevitable in the work of a modern marketer.

Is it possible to get rid of them? Unfortunately not. Is it worth it to figure out the reason? Yes, because it will allow you to:

  • better understand user behavior;
  • Find the most suitable data sources for specific reports;
  • respond more effectively to data distortions;
  • notice technical errors in the application without looking into the code;
  • determine the acceptable level of discrepancy.

In this article we will look at the main metric of attracting users - installation.



App Store and attribution platform analytics: when installation doesn't count


What is install in terms of mobile attribution platforms (Adjust, Tune, Appsflyer, Kochava, etc.)? Logically, the installation occurs immediately after downloading the application from the App Store, right? But no.

Most analytics systems count the installation when the user first launches the application (some systems even call it download). The reason is that most analytics systems receive data exclusively from the SDK installed in the application. Therefore, the earliest stage when they can determine a new installation is only after starting the application.

There are only two systems that can count actual installations - Google Play Console and iTunes Connect. Comparing their statistics with the data of attribution platforms, you will probably see discrepancies.



Below are a few scenarios that may lead to data discrepancies for this reason.

  • The user installed the application, but never launched it.
  • The user launched the application a few days after installation.
  • The user installed the application, uninstalled it, and then installed it again.
  • The user installed the application, but before the SDK started, there was a failure.
  • The SDK did not start immediately, but later, for example, after downloading additional content.

Of course, the last two cases can be avoided by properly debugging the SDK.

Accounts or devices?


Another interesting fact: Google Play Console and Itunes Connect consider installations based on account IDs, and attribution platforms based on device IDs (GAID / IDFA).

For instance:

  • The user installs the application on two Android devices with the same account identifier. Google Play will count this for one installation, and the attribution platform for two.

In such cases, discrepancies cannot be avoided. You can only determine their acceptable level, which, as a rule, will depend on your application.

Affiliates and ad networks: have you set up an attribution window?


Let's start with the definition. The attribution window is the time period from the first contact with the application’s advertisement to its installation. The problem is that attribution windows are not always configured the same.

Suppose a user clicked on an ad on August 1, and installed the application on August 3.

  • If the attribution window is set to 3 days or more, the installation will be counted.
  • If the attribution window is set to 1 day, the installation will not be counted.


For the installation to be counted, the time between the click and the installation should not exceed the attribution window.

It is worth noting that attribution windows for clicks and impressions can be configured separately. A good example is Facebook. Here we can set up attribution windows and watch on the fly how the data changes.

Facebook allows you to choose one of three attribution windows for views and clicks: 1, 7 or 28 days. If you do not configure the windows manually, the data will be tracked by default (1 day for views and 28 days for clicks). You rarely see this kind of flexibility when working with attribution platforms.


Here you can set up attribution windows on Facebook.

Remember that before comparing settings data, make sure that the attribution windows of the ad network and the attribution platforms match.

My click or last click?


The following type of discrepancies is often found in major advertising networks: Google Adwords, Facebook, Yahoo, Twitter.

Ad networks pretend to be installed, not knowing about user interactions with other networks. In other words, different networks can attribute to themselves the same installation.

But here attribution platforms come into play. They collect data about all clicks in the attribution window and count only the last one.

As an example, consider Google and Facebook.



For example, before installing the application, a user clicked on an advertisement on Facebook, and then on Google. Naturally, Google Adwords will attribute the installation to itself, and Facebook - to itself. But, as a last-click rule, the attribution platform will account for the installation of Adwords.

Did you get my postback?


We talked about industry giants, but what about everyone else. How does everyone else work? After the attribution platform credited the installation with the last click, it passes the data to the advertising partner via postback. Without proper configuration of the postback sending mechanism, statistics on settings in the ad network may not be displayed at all. Sometimes data sent via postback lacks required parameters (for example, clickID). In such cases, the installation will not be registered.

To solve the problem, it is best to contact the account managers of both parties. But this can take some time, so if a partner can provide you access to the logs, this will be a big plus.

Click, view or both?


Some analytics systems show views and clicks separately, while others show together. Therefore, before comparing the data, specify how the settings are displayed in analytics.

What time zone are you in?


Everything is extremely simple here. When comparing the number of installations on two different platforms, you need to consider in which time zone they work. Most modern systems display data for different time zones, so there are hardly any problems here.


Facebook shows the time zone of your account.

Are we looking at the same thing? Report Types


Sometimes it seems to us that we look at the data from one angle, but in reality this is not so. Different partners use different approaches to displaying data.

Consider this with the Google AdWords example.

AdWords displays all conversions, including settings, based on the date of the click, not the installation. And most attribution platforms are just the opposite. If a user clicks on an ad on August 3 and installs the 4th app, AdWords analytics will display the 3rd and partner’s analytics the 4th.

So, we examined most of the reasons for the discrepancies in the number of installations in the mobile market. If you have encountered similar problems, join the discussion in the comments.

Also popular now: