The most important step before you begin sending data to any event-based analytics platform is identifying which events to track and send. The importance of good, well-planned instrumentation cannot be overstated. When you skimp on instrumentation, issues like omitted events, poor naming, and misused properties can needlessly complicate your analyses and prevent you from fully maximizing the full value of your data.
NOTE: This article is part one in Amplitude's data taxonomy playbook series. When you're done here, you can read part two here, and part three here. You may also find this course on designing a taxonomy in our Academy helpful.
To make it easier, we've created this three-part playbook to creating your data taxonomy. In part one, we'll take a high-level overview of how your data taxonomy will guide and shape your experience using Amplitude, and in turn, the experience your users have interacting with your product. In part two, we'll take a more detailed look at events and properties, the two key elements of your data taxonomy. Finally, in part three, we'll offer a few industry-specific suggestions for building your data taxonomy from the ground up.
Before you begin instrumenting your application on Amplitude, we recommend that you read our Quick Start Guide.
A note on data quality
It's extremely important that your company establish and enforce a set of data governance rules when it comes to instrumenting events. These rules will ensure that your taxonomy stays clean and ensures that people within your organization use Amplitude, understand the taxonomy, and are able to derive meaningful results from the data.
Common things to consider when creating business rules:
- Make sure new data aligns with your goals/use cases and KPIs
- Create a consistent method of naming events
- Event names should have consistent casing and formatting guidelines
- Consider shortening longer strings so they are easier to read
Above all, keep your nomenclature consistent. For example, if you track two events like 'Checkout Submitted Order' and 'Checkout submitted order', they will be considered two separate events and cannot be merged automatically. Furthermore, you will confuse other teammates who are using Amplitude.
You can learn more about our data governance best practices from this helpful video.
Define your business objectives
What are you and your team working towards? What is your overall goal? Why are you using Amplitude in the first place? There are all sorts of business objectives that Amplitude can help you achieve.
Some typical goals Amplitude customers pursue include:
- Setting product strategy
- Improving acquisition ROI
- Targeting engagement
- Optimizing conversion
- Increasing retention and LTV
Once you've identified your organization's overarching goal, the best way to begin working toward it is to establish a set of key performance indicators (KPIs). These are metrics you'll focus on improving in order to achieve your goal.
Let’s say your business objective is to optimize conversion. Your KPIs may be:
- Improving onboarding conversion
- Improving checkout funnel conversion
You should define these before you start thinking about your data taxonomy, so you can send the right events in order track your KPIs and achieve your goals.
Understand the user journey
As you begin to create your event taxonomy, don't forget about the end user's journey through your product. Your taxonomy should allow you to split your analysis into three different levels:
- Event counters: Daily / monthly average users, total purchases, etc.
- Funnels and conversions: Retention rates, funnel conversion, power curve chart
- Behavioral analysis: Impacts of performing actions on your other metrics
You should understand your product well enough to have a good idea of what a typical user's journey from new user to power user should look like. The idea here is to understand the factors that cause a user to transition between these states. Tracking clearly-defined events for these key areas will help you better analyze patterns that help (or hinder) users from progressing in your product.
Understand your product's critical paths
Every application has critical paths. Critical paths are sequences of actions a user takes that align with your product’s purpose. For example, a critical path in an e-commerce product could be:
Search → Browse Products → Add to Cart → Checkout → Order Confirmation
For a gaming product, a critical path may begin when a user opens the app, is prompted to register, and then taken through a game tutorial.
You can break this onboarding process or path into a series of events: 'App Open', 'Registration - Personal Info Populated', 'Registration - Avatar Selected', 'Registration - Complete', 'Game Tutorial - Started', and 'Game Tutorial - Watched'.
Once that's done, go through and determine the type of tracking you'll need to best understand that critical path and the flows surrounding it. It's not enough just to know how many people placed an order or finished the tutorial, for example: you also need to know what factors drove them to trigger that critical event.
Amplitude projects and data taxonomy
In Amplitude, a project is a destination where all the event data from your product is collected. You can have multiple projects within your Amplitude organization. However, it is extremely important that you have a minimum of two projects in your Amplitude organization: a test project, and a production project. You should always QA your instrumentation in your test project before sending it to your production project, to catch any instrumentation bugs and keep test data separate from production data.
All data sent to one project will be independent of any other project. Your total number of projects will depend on your products and your business objectives. For example, if you want to analyze users across platforms—maybe you want to see a user's full story and understand what it means that they triggered one event on an iOS device but a different, related event on the web—you'll want to instrument everything under a single project.
NOTE: Amplitude does not currently support out-of-the-box cross-project analysis. All data is kept separated between projects unless your organization has the Portfolio add-on.
However, if your company:
- Wants to track data for multiple products
- Has fundamentally different experiences across platforms (web vs mobile)
- Has individual, independent teams responsible for each platform or product
Then you should set up multiple individual projects for each silo (product, platform, etc.). This will keep your data taxonomy clean and easier to understand.
NEXT: Events and properties