It’s been a couple of months since we’ve posted anything but rest assured
we’re going to be kick starting this blog with new posts every week. We wanted
to let you know, we’re still here, we have a heart beat, and we have some
huge changes coming for Mixpanel in the next few
weeks so please stay tuned.
Regardless, I want to touch on something: Cohort Analysis and one specific way
that web companies today need to think about it. In most user centric web
companies, they have at least two types of users: new users and power users.
How these two types of users function is very different and it’s a mistake to
neglect either one at certain stages of your business.
The new user
This is what newer companies tend to focus on aggressively. How do we get more
traffic, how do we acquire more new users? Some websites we see have great
engagement for users who “get” their product, and those users interact heavily
with it so much so that it almost creates an illusion that they are doing
Tracking how certain types of users interact with your software is immensely
important. As you build and iterate your product, new features that you add,
remove, or update could quite possibly change the way your users behaved
previously. You may be surprised to find that your newest users interact with
things differently, they may play more songs then your power users but add
less of them to their playlist for example. They may drop off in a different
way then you expected in some of your viral flows. Maybe, that ad that only
“free” users now see, decreases your overall retention or conversion of sign
ups. Either way, certain changes may not align with your immediate business
goals and need to be reverted.
Often the analysis of new users generally happens to be the trend you see on a
regular basis simply because they outnumber your power users.
The power user
These are your most passionate users, they come back.
Understanding how a specific set of users behave is going to help you learn
more about them than just looking at an entire set. We’ve seen companies that
have users do x, y, z actions on average of 10 times and suddenly realize the
standard deviation (how spread out their data is) is so high that it’s clear
that there is more than one kind of user they have to cater to.
Some games have users that play it literally all 24 hours of day. Keeping them
happy is very important or they will just leave which is likely the case of
many of the new RPG browser games that are popping up everywhere. To grow, you
need to compound on visitor traffic and to do that you need great visitor
Generally though, you want to cater to your power users, watch how they
interact while you first start out. You need to find out what makes them want
to use your product, what hooks them. New user acquisition is hard and it
takes time, don’t just watch everything at an “on average” case–it’s a
mistake. Align your goals with making a small set of users happy and then use
what you’ve learned to convert more new users into power users.
Watch how a specific set your users interact with your product, figure out the
best way to cater to them based on your companies’ overall goals. It doesn’t
just need to be new vs. power, you can divide groups of users anyway you want.
In order to evolve your product you must start looking outside the “average”
case. At Mixpanel, we’re happy to help you figure this