5 Mixpanel Pro Tips
that will make you look like a data scientist
(Note: this post was written a few years ago and is likely outdated. Read with caution! Check out my more recent posts.)
I know there are fine-grain analytics alternatives out there that people swear by, like Localytics, and KISSmetrics, but I got started on Mixpanel first, and they deliver the kind of value that I can work comfortably with. Everything is easy to set up, and their customer support team has been very understanding, even when it came down to being charged big money for an implementation mistake.
The thing about Mixpanel is that it works best in the hands of someone who understands the nuances of the interface. It’s not as easy to use as people give it credit for, but, like Google Analytics, Mixpanel won’t stop you from getting as detailed as you want when the time comes.
Let’s be frank: Mixpanel won’t tell you anything useful if you don’t know how to set up your tracking correctly. I managed to make our internal funnels into total garbage for the first month of use because I couldn’t properly tie users to the same unique ID after they authenticated. Sometimes I start experiments that don’t get enough user exposure, and therefore create serious false positives (which luckily I ignore). Mixpanel will show you everything you feed into it, and then it will break those things down into single 1's and 0's if you ask it to. But you know what they say… garbage in, garbage out.
Okay, let’s get to the good stuff.
Five Advanced Mixpanel Tips:
1. Funnel completion time
There are some types of actions that shouldn’t be allowed a full month to happen. If a user signs up on your app and you want them to do something (say, add their profile picture) in the first visit, change the settings to “1 Hour” or “1 Day”. This will give you a much more accurate snapshot of typical user behavior. By leaving the setting on 30 days, you allow “trickle-back” users to skew the displayed conversion percentage. It’s not important that they did it two weeks later, you want to know if users did it in the first visit (plus, returning user behavior is typically much different from that of a brand new user).
On another note, it really helps when you know what you are trying to find out. Define your experiments first. “Looking for insights” in your unstructured pile of raw data is the same as “waiting for your big moment”. Pretty good chance you’re not going to find anything.
2. Multi-level segmentation
Segment your users based on an event. Segment those users again based on what country they are from. Segment one more time based on device type.
That is just one example of how you can slice and dice your user base into something more meaningful. In the above images, you can see that I segmented my selection by “are they an expert-type user?” (at my company mosaicHUB) and then again by “what badges did they get?”.
As you can see, for this selection (I can’t remember what date range this was, so it’s probably not reflective of the current state of the site) the badges that are earned most are “Coach” for experts and “Reviewer” for non-experts.
We have a many more badges that users can earn, so something we might have done after looking at a graph like this is dug into how close users are to earning those other badges. Are they too hard to get? Do users even care? Mixpanel isn’t good at saying why something happens, but it tells us that it does happen. Otherwise, we might not have considered something like, “users are only getting these badges… why?”.
3. Retention — “Addiction” mode
This is a more recent addition to Mixpanel that tries to wrangle a concept that is pretty challenging to measure: short-term retention cycles, aka addiction.
Regular retention measurements will show you how many days or months in a row users do a particular action. The “addiction” mode, on the other hand, will show you how long users spend doing those actions. If a user listens to music on your app for more than 2 hours, but never for more than 5 hours, you will see that curve appear on your chart.
The “addiction” mode, at it’s core, is simply a segment filter laid over a static x and y-axis. It shows you how much time users spent on your site doing anything crossed with users that fall into the segment of “did the action” that you defined. (e.g. users that upload their third photo regularly spend more than 2 hours on the site per day). This is different from the two other modes, which track the frequency of individual events (e.g. users that uploaded a photo who came back the next day and uploaded another).
4. Segmenting your funnels
Many times I found myself asking:
“Okay… so a percentage of the user’s completed the action. But this doesn’t tell me anything about which users did it. What should I do next?”
The answer is built right into the funnel viewer. Click on either one of the bars in the chart, then scroll down to see the familiar segmentation tool (which is peppered all over Mixpanel).
In the above picture, I took my funnel of “Viewed Question” and only showed users who were on the default Android web browser. Turns out that not only were there a small number (31) for the selected time period, but only 19% of them viewed a question after looking at their dashboard. Do a couple more and you can draw conclusions quickly.
(Note: when you segment a funnel, it doesn’t matter which bar you clicked on. The segment will apply to the whole funnel, starting from the very beginning.)
5. Bookmark everything
If you are serious about analytics, you will be looking at a lot of different displays. It will become time-consuming to punch in all of the segmentation rules, click all of the events, and change the dates when showing other team members your results. Screenshots work for a while, but they cannot be up-to-date as time goes by.
Enter bookmarks. On most screens in Mixpanel, you can freeze and save the current display for later viewing by clicking the + button in the top right corner. Google Analytics has this option as well, in the form of “Report Shortcuts”.