Reference this appendix for further explanation of instrumentation review and how to validate your data.
Start with the right amount of data. Have you just started tracking your data? Now’s a good time to make sure your events are instrumented correctly and the data you’re collecting is in good condition. We recommend having 3 months of data at minimum before you start with the analyses presented in this playbook.
Before analyzing metrics, the first thing to do is to make sure you’ve instrumented your analytics correctly. It’s tempting to rush through this part (we know it’s boring), but this would be a mistake.
Sending optimal event data to your analytics platform is the single most important step toward understanding how your users are engaging with your product. It’s worth the upfront time investment to get your instrumentation down correctly.
Organize your event taxonomy
In event-based analytics, the term event describes any action performed by the user or any activity associated with the user. Opening an app, making a payment, adding songs to a playlist — all of these are examples of events a user can perform, whereas things like receiving or interacting with push notifications are examples of activities associated with the user.
It’s critical that your event taxonomy reflects your business objectives. This means taking time to understand the type of company you are (e.g. the vertical you’re in, your business model), what success criteria you care about, and the type of metrics that are important to you.
Do a quick check of your event taxonomy by asking yourself the following questions:
- Are the events you’re tracking aligned with your analytics goals?
Think about your analytics goals in terms of business objectives. How will you use analytics to measure the value you deliver to your users and vice versa? Are you able to track revenue, retention, conversion? What experiments and funnels would you want to run in the future? Are you tracking the relevant events that would allow you to run those experiments?
- Can everyone understand what each event is and why it’s being tracked?
Make sure you understand the context around all the events that are currently being tracked, as well as when you expect them to fire. Having an organized event taxonomy document listing each event currently being tracked, as well its corresponding name and properties in a central location is critical if you want everyone to be able to derive data insights. Check out Amplitude’s Event Taxonomy Template for more information.
- Are you tracking events aligned with your critical path funnel?
You’ve probably envisioned an ideal path to conversion that your users flow through—one that perfectly syncs with the product’s core value. Make sure you’re tracking all events along this path. If you’re an ecommerce product, for example, you should be tracking all events leading up to the user clicking ‘checkout’ and completing a transaction. We discuss how to determine your critical event in Section 2.1 and how to set up your critical path funnel in Section 4.3.4.
- How are you defining an active user?
Do users simply need to open an app to be counted as “active,” or are there specific actions they should take to be considered active?
Validate your data
To analyze your users’ behavior, you need to have a deep understanding of how their actions are reflected in your analytics platform. The easiest way to check your instrumentation is to be your own user:
- Check your onboarding:
Download your own app and and simulate your app’s first time user experience. Go through the onboarding process. Identify yourself as a new user in your analytics platform, which should be possible with a unique user ID. Then check to see that events are firing correctly and all behaviors are being captured properly as you complete each step of the onboarding process.
- Check your critical paths:
Simulate an “ideal” user flow through your app, from start until conversion, and make sure those events are being captured correctly.
- Do rigorous error testing:
Bugs and crashes can be major detractors of retention and should be resolved before making product optimizations (more on this in Section 6.6). Try “breaking” your app and forcing it to throw errors, and make sure you track events that correspond to those errors and crashes.
Taking the time to do a comprehensive audit of your event taxonomy and your data quality will ensure you have a solid foundation on which to do more granular analyses. Doing this legwork upfront is critical if you want to trust your data.
We strongly recommend naming your events as human readable strings. This is because if someone on your team wants to look at the data, they won’t understand it if the event names are written in mysterious shorthand.