On Mixpanel, you can track visitor retention.
Basically, what percentage of people come back and do something some period of
time later. A simple question we can answer for Twitter would be: How many
users come back to Twitter and tweet 4 weeks later?
There are a few ways to calculate your retention depending on cohort groups:
Birth class based
Birth class based retention is a little complicated but usually what most
people expect in terms of retention analysis. We take all the users seen on a
given day and calculate what percentage of them come back to do the action
again. The difference though is we do not use those users again if they show
up tomorrow and do the action again to calculate retention. For example:
10 users showed up today
20 users showed up tomorrow (5 from the day before included)
15 users showed up the next day (7 from the day before included, 2 from the
5 users for the “tomorrow” cohort are not counted for the next cohort because
we already saw them but your day 1 retention from “today” would be 50%. On
your “next day” 7 showed up so your retention would be 35% however the 2 that
showed up from the “today” group and the 7 that showed up from the “tomorrow”
group will not be used to calculate the next cohort group. In that case, only
6 users will be used to calculate the new cohort group.
We call it birth-class based because it’s when users were “born” or first
seen. This is the type of analysis we did at Slide, Inc.
— the name just stuck with us here.
This type of cohort analysis is useful to see if your retention is actually
improving over time as you change your website/application.
This type of retention is similar to birth class based retention except we
don’t care if we’ve seen a user before or not. If 10 users from yesterday show
up today, we’ll count them in with the users you got today as part of the new
This type of retention analysis is useful for something else: learning about
your power users. Basically, as your site grows some percent of users who sign
up will continue to use it regularily. At some point, you’ll have 1000 people
who actively use your website but maybe only 100 sign ups a day. What you get
is power-user (sticky users) based retention.
Your retention in this category should be the highest you could ever imagine
especially compared to birth class based. These are people who love your
How do I improve this?
There are a few tricks to improve retention but ultimately the best thing to
do is: “Build something people want.”
- Email notifications (Example: Emails sent when someone tags you in a photo on Facebook)
- Game mechanics (Example: Get 10 experience points for logging in like Listia)
- Network effects (Example: More friends on Facebook you have the more useful it is)
- User generated content (Example: Dailybooth has users take pictures of themselves which drives a greater reason to come back to the site to see new pics)
Hopefully now you understand the difference in the types of visitor retention
analysis we provide. Enjoy.
By the way, we added the ability to segment your visitor retention further by
where a user came from. Check it out.