Creating Meaningful Charts in Amplitude
Recently, I have been working on more analytical side of web development, and have really been trying to understand what drives different user behavior.
Specifically, I have been focusing on one question that I find extremely interesting:
“Which user action is more correlated to user retention at the 1–months and 3-month marks?”
So basically, what behavior that a new user partakes in is most indicative of their long term retention? Is there a certain action that users perform in the first session after account creation that is a clear indicator that they will be users for the long haul? If so, how can we encourage more users to take this action so that we can increase retention overall?
What Amplitude is Used For
If you’re not familiar, Amplitude is a great analytical platform that allows you to perform event tracking and user behavior analysis. My company has added it to it’s React web and mobile apps in order to anonymously track things such as user sign up activity from marketing events, daily, weekly, and monthly active users, number of posts and comments per user, and a multitude of other useful data.
Right now, our 1-month and 3-month user retention rate is lower than we had hoped. This led to us deciding to do a deep dive to see what causes the users that do stick around to keep using the app, and see if we can encourage more users to follow suit. Here is how you can use amplitude to track events and try and deduce this information.
Creating a Chart
The first thing you’ll want to do is make sure event tracking is added to your web or mobile application. I already wrote an article describing how to get started with this in React.
Once that is done, you’ll want to navigate to your dashboard and press the New button to create your first chart.
Next, you’ll want to choose the events to track in order to form your chart. Here is an example of what yours may look like, but they will all be created by adding code to your application that tracks each event.
Without showing you too much of the specifics of our user stats going forward, I will try to explain how to build a proper chart and extract useful data from it.
Let’s say you have a messaging board app. One interesting thing to chart would be the numbers of users that make a post, versus the number of users that reply to a post. In a healthy app, you would ideally be getting lots of replies on each posts, but it could also be interesting to look at how many people are doing both of these behaviors versus strictly replying.
You can also monitor how many users are “lurkers” who read posts without interacting and are more content consumers than creators.
In our case, the number of posts and post replies are pretty even, which can be from one of 2 things:
1) Power posters who create a lot of different posts without responding to other people’s
2) Low engagement rates on posts where you are only getting 1 or 2 replies to posts.
In our case, we actually seem to be skewing towards the former, where we have a lot of engagement on posts, but there are a strong contingent of people who posting daily and don’t really engage with others, so this is skewing the numbers toward even on posts vs post replies. We are coming up with a plan to encourage engagement on other people’s posts by these power users as I write this.
Gleaning Insights from Charts and Tables
Now you know how to set up event tracking and you know how to create charts and tables, but how can you figure out how to gain meaningful info from all these fancy graphs?
You want to begin by asking yourself a question and then trying to answer it using Amplitude’s tools.
In my case, i’ll refer back to the inspiration for this article:
Which behavior is the most clear indicator of long term user retention?
This is a very complicated question, but in our example case I will only use 2 behaviors:
- Creating a post
- Sending a DM to a user
What you want to do here is set up tracking for each of these events and then chart it against the user retention at various intervals. In our case we will use 1 month of data, but in your case you may want to extend it to a longer timeframe.
This chart depicts the user retention rate at each point for users after making their first post. After 1 day, there is a steep drop off for users that are still posting after making their 1st. Also, it shows that around 20% of people who make a post are also sending DMs at the time. It would appear that only 18% of people who make a post are still active at the 1 month mark.
Conversely, here is the chart when you use a DM as the trigger:
It appears even fewer people who send a message request are still making posts at the 1 month mark, while the sending of DMs actually picks up towards the end.
This is a very simple concept, but gives an idea of what you can do with Amplitude, and when you add it to your app you can start answering these type of questions to see which actions lead to certain results.