Welcome to Step Three of our Getting Started with Amplitude Series. This series is intended to help you set up Amplitude in the quickest, most optimal way possible, by walking you through the Amplitude data structure and helping you identify the data your product should be sending to Amplitude. Specifically, we'll cover:
- Introduction & getting started
- Instrumentation pre-work: Things to consider before deciding what data to send
- Identifying your users: Requirements for properly tracking your product's unique users
- Event data: How to identify the events or user actions you should track
- User properties and event properties: The attributes you should send to upscale your analytics
- Cross-platform instrumentation vs. separate platform instrumentation: The differences between them, and when you should choose one over the other
- Dive into Amplitude and Snowflake, and explore powerful resources: Use Snowflake with Amplitude to answer key questions via SQL
(If you're a developer or product manager who will be responsible for instrumenting Amplitude, you should also read our getting started guide for developers.)
In this article, we'll be walking through the process of coming up with an event taxonomy. We'll go over best practices regarding how many events to track, which events you should be tracking, and effective QA, and we'll answer a few common questions we've received.
NOTE: This article is intended as a broad overview. For a more in-depth explanation of setting up a taxonomy, we strongly recommend checking out our Data Taxonomy Playbook.
What's a taxonomy?
A taxonomy is a set of hierarchical classifications and naming conventions for your data. It's a way to identify and categorize your event and user data so that Amplitude can generate relevant and valuable insights from it. The process of setting up a taxonomy in Amplitude will differ from organization to organization, but the heart of it is selecting the events you want to track, identifying event properties and user properties you want to track, and then naming them.
The decisions you make when setting up your taxonomy can have a significant effect on the value you derive from Amplitude, so think carefully before instrumenting.
How many events should you track?
Often, we see new customers try to track most—or even all—of the actions a user could possibly take in their product. Resist this urge. Though it feels logical to say that more data will lead to more insights, it usually doesn't work out that way. Too much data actually obscures insights, burying them under an avalanche of events and properties you don't really need to know about.
The real answer to the question of how many events you should track depends entirely on the complexity of your product. If your app has a more focused feature set, you might only need to track 20 events or so; if your product is feature-rich, on the other hand, 200 might be more appropriate.
Finding the right number will require you to really think about what kinds of insights you're hoping to find via Amplitude. What are the questions you'd like Amplitude to help you answer in the next quarter? The next two quarters? Identify those, and choose the events that will lead you where you want to go.
What events should you track?
No matter what, you should be tracking any actions that fit into these three categories:
- Those that are important in completing a process within your product, like completing a tutorial or signup process
- Those that guide a user through the main mechanics of your product
- Those that enable a user to make an in-app purchase
When you've decided which events to track, you'll be ready to plan your events in Amplitude.
Our Data Taxonomy Playbook takes a deep dive into this question, so be sure to read that as well.
One quick rule of thumb to keep in mind is if a process can be represented as a funnel, you should be tracking the steps in that process. For example, tracking an onboarding funnel will help you quickly understand if first-time users are successfully navigating your signup process. A retention funnel will help you understand which parts of your product your users find confusing or frustrating, so you can spot and fix them quickly.
If you need help deciding what events to track, please contact your dedicated Success Manager or reach out to us here.
QA your instrumentation
To verify your instrumentation works the way you'd intended, navigate over to Amplitude's User Activity tab. Fire some events using your test device, go to your project in Amplitude, and then watch as the device ID or user ID appears on the near-realtime feed. Clicking on that ID will take you to that user's event stream, which should be made up of the events you've decided to track. If you're not seeing the events you expect to see, it means something's wrong with your instrumentation.
How Amplitude handles duplicate events
Amplitude de-duplicates your data to prevent unique events from being logged multiple times. Amplitude checks the event ID, client event time, and device ID for every event. If the event isn't in the database, Amplitude writes it; otherwise, the event is dropped.
If you're using the Amplitude HTTP API, we recommend adding an insert_id field. Amplitude will ignore subsequent events sent with the same event ID/client event time/device ID or insert_id within the past seven days.
Event limits and data collection
If you go over your limit for the month, Amplitude will still collect your data as usual. However, you won't have access to that excess data unless you upgrade to a new tier or wait until the following month.
When you're ready to move on, just click this link to go to the next step: Setting user properties and event properties.