The Background
Our client is Football Club Shakhtar. He is one of the leaders of the Ukrainian Premier League, a thirteen-time champion of Ukraine, an eight-time winner of the Super Cup of Ukraine, winner of the UEFA Cup in 2009, and a regular member of the UEFA Champions League.
The Challenge
Shakhtar has Android and iOS applications for fans, where you can buy tickets, watch matches, or get additional content about the team. The client-side development team is really strong, they themselves have integrated Firebase with Google Analytics 4 and e-commerce modules; built a system of conversion tracking, event tracking, and other useful settings.
But there was a problem. Firebase did not track traffic sources in applications. This affected all related marketing activities because when they ran the ad, they could not track the sources of the conversion – all visitors were marked direct / none.
So Shakhtar asked us to understand why traffic is noticed direct / none, and how to set up correct traffic tracking, with the transfer of conversions, their value, and sources.
The second task was to set up e-commerce modules. It was necessary to teach the system to take into account refunds and cancellations, as well as to test whether the data is duplicated somewhere.
And additionally asked to check for potential difficult areas. So, we had to check the existing analytics system, test its work and clean up any errors.
Our Approach
Firebase Attribution
First, we started solving the attribution problem in Firebase. To track traffic sources, we suggested using dynamic deep links.
Deep link is a link that leads to a specific page or section within the application.
Dynamic Deep link is an in-depth link to which parameters such as UTM tags are added but takes into account the specifics of the applications.
We were surprised that the client used the deep links, but stopped, so we offered to experiment again. During the experiment, it was found that the data from the deep links are delayed, and this delay can last several days. That is why the client first decided that the deep links do not work. But after several test runs, we were able to get all the necessary data and show the client how to use this system properly.
It is easy to implement dynamic databases, they can be created automatically: either directly in the Firebase interface or by API. Read more about their attributes in the Firebase Help. The difficulty for us was to diagnose the problem, to find out what the links were.
So, we set up deep links for existing campaigns and developed a technical task for the client, which he will be able to work with subsequent campaigns. After that, we saw that the system began to show traffic sources.
By the way, we have previously encountered similar issues with Firebase tracking. How we solved them – read in the case of Firebase VS Adjust: how media campaigns can suffer because of data.
E-commerce Refund
When a user cancels a ticket or part of an order, the amount of the cancellation must be correctly transferred to the analyst so that it can take into account the cost. For the client to work with this, we developed a task for programmers, in which we wrote the proposed script and the configuration of tags that should link the return to a specific transaction.
Other Issues & Audit
Next, we examined all other potential vulnerabilities that could be in the client’s analytics.
- Do the events work correctly?
- Are conversions and costs from Firebase to Google Analytics 4 correct?
- Aren’t conversion costs duplicated as the funnel passes?
In fact, no other critical problems were found – everything was set up excellent. But here we want to share the life hack. We noticed a difficulty in identifying our actions in the application.
When you go to the app for the test, you choose the geolocation “Kyiv”, it is difficult for you to recognize yourself, because in Kyiv at the moment there are 20 more people from the same iPhone as yours. Usually, we use special filters and parameters in this case, but it does not always help. In this case, we were very helped by the fact that the application had a test version, which we installed, and then we were able to find ourselves on the version of the application.
In general, we did not find any critical errors during the audit.
Result
After we introduced the use of dynamic deep links in-application advertising, Firebase began to correctly track traffic sources and channels.
And the implementation of the terms of reference for the e-commerce module can improve the already clear digital analytics of the client.
If you’re looking for an analytics solution for your mobile apps, you’ll be interested in our Firebase & Adjust comparison case study.
