What I Learned Switching from GA4 to Amplitude
Longtime Google Analytics user, Ani Lopez, Senior Data Engineer at Unity, made the move to Amplitude. Like all change, the move had its share of ups and downs. In this How I Amplitude session, Ani talks through why the Unity team migrated, and a few of the things they learned along the way.
“You can define all your channels, and replicate those in Amplitude. But you have to really understand how channel definitions work, and not to take the definitions of other tools. That's a little example of how users have to get deeper and put their knowledge to work.”
Two years ago, Unity was in a state of constant evolution, which impacted our measurement tools.
We had around 40 web properties, several e-commerce sites, desktop apps like Engine and Hub, and web apps like cloud.unity.com. So we have a complex ecosystem with information coming from different back ends and systems.
We were already using Google Tag Manager (GTM) and used the Enterprise version of Google Analytics.
Why Migrate?
So, why did we migrate to Amplitude? The trigger was Google's announcement to sunset Universal Analytics. Their erratic roadmap and lack of partnership were significant concerns. We needed a tool that could cover all our data consumers' needs, from basic charts to advanced analysis in BigQuery.
At the same time, we needed robust governance, not only in terms of data access, but in terms of managing taxonomies, too, as they are very difficult things to govern. This prompted us to re-evaluate our tools.
Pre-Migration
If you think about how much we had to track and the complexity of our setup, we needed time to lay a solid foundation. For us, the pre-migration process took around nine months. We gathered requisites and concerns, talked to stakeholders, and evaluated potential solutions. The key things we needed from a solution were:
- Serve many different data consumers
- Robust Governance
- Excellent Support
- Ability to measure marketing performance, including attribution, ecommerce, and segment creation.
- Integration with other tools, like CDPs, Experimentation, Session Replay, etc.
We chose Amplitude because it best met our needs. We already had an account, so were able to jump right into the platform and start tinkering and understanding how the tool was going to be.
Here is the implementation setup we added Amplitude to:
Technical Migration
We tackled the technical migration in two phases:
- Phase 1: Measuring Needs
The first phase took about four months. We wanted to understand what aspects of our setup was still working for us, so we could keep, as well as identify areas of improvement.
We ended up redefining our event taxonomy to make it more flexible and scalable. - Phase 2: Technical Implementation
We wanted to avoid as much engineering work as was practical, so we used GTM heavily. We ended up with all events, whether user generated, pageviews, or back-end responses being passed into GTM as a central control node.
This works alongside a privacy protection solution, and helps us ensure an agnostic and scalable approach. From there data is then sent along to downstream tools, like marketing and analytics platforms. This all made switching from GA to Amplitude relatively simple.
Once we migrated, we then:
- Ran GA and Amplitude in parallel for a couple of months for comparison. This was mostly for our team’s purposes, we’d already made clear to end consumers that things are not going to be exactly apples to apples - don't waste your time trying to get there, just learn the new tool.
- Documented the taxonomy in a spreadsheet and created a clear process for tracking implementation. We established a ticketing system for requesting marketing tracking implementation tasks, and a Slack channel where folks could ask all of their questions.
- Reduced the support overhead of people learning a new tool by leveraging Amplitude's educational resources.
Elevating our Measurement Culture
One sticking point was recreating the ready-made reports people were used to in Google Analytics in Amplitude.
GA treats their channel classification as something of a black box - what’s considered search vs. social vs. organic, etc. This is something that many marketers don't think too much about, they're used to seeing these figures. But when you leave GA, you have to come up with your own definitions. So we had to define and classify the domains that made up those channels in Amplitude using channel classifiers. Setting up those rules is a combination of fields like the referring domain, UTM medium, source, and campaign, and you can use regex.
Below you can see an example report in Amplitude, re-creating a default GA chart.
This process is painful and ended up being more difficult than we had expected. Our first attempt at implementing them ourselves did not match up closely to what was in GA already. We ended up creating a spreadsheet, assigning owners to different channels, and getting their input on exactly which sources should go into the different channels.
Here's the spreadsheet we used, which you can use as a template for your own definitions.
However, going through the process is an eye-opening experience. It levels-up your game and actually improves your measurement culture because it brings a lot more awareness of exactly what it is that you're tracking.
Some more advanced stuff we did included things like user journey analysis by personas.
We’ve adopted Amplitude Experiments across the board, replacing Google Optimize, which they also deprecated. Finally, we’re using Portfolio to enable us to perform analysis across a mix of different data sources: not only websites, but applications, too. They send data to different Amplitude projects. With Portfolio on top, you can mix and match those data sources.
2 Years On
After two years, we've found that Amplitude is not just a product analytics tool; it also serves our marketing needs. The partnership with Amplitude has been strong, and their support has been invaluable. There is constant communication about ‘what is missing?’ or ‘what we can do better?,’ which you don’t get with GA.
We used to have an array of tracking tools, and now we’re all in on Amplitude. Not only does that decrease costs, but it’s also been positive for improving our data governance, and to grow our culture of measurement. You can’t do that if you’re using 20 different tools.
Join the community!
Connect with Ani and other practitioners in our community. The Cohort Community exists to help people become better analysts. We focus on actionable programs, sharing best practices, and connecting our members with peers and mentors who share in the same struggles.