Cohort Analysis - Is this useful for community development?


(Angelo Luciani) #1

I know many folks use cohort analysis for SAS applications, mapping how many people drop off after a certain amount of days, etc. I was wondering if this is something useful for community tracking. Has anyone done this? Any examples of how to perform this?

Thanks


(Richard Millington) #2

Hey @aluciani,

The short answer is yes. Very important. But it’s really hard to do without paying for professional tools (which I’d recommend if you have the budget).

If you can’t do it for the entire cohort then pick a stratified segment of that cohort and track as many as you can. Perhaps that’s 50, maybe 20. Using virtual assistants can also help here to see where and when people drop out.

Not sure if many others are doing this yet, but you’re definitely leading the charge.


(Bas van Leeuwen) #3

Hey @aluciani,

It is an extremely powerful tool; we’ve been busy integrating it into our product for the past few weeks and it is giving us tremendous insights. Happy to share below :slightly_smiling:
Currently we base a cohort on someones yearly-quarter-date-of-first-post (so if you posted your first contribution on 12 March 2015, you’d be in the Q1 2015 cohort), and we show the active-members metric (number of members that have posted in the past 28 days).

This is a typical one (Discourse Meta)
You often see that the first (couple of) cohort remains active; that makes sense since this is the group that includes the founders and the early members. After that you often see that cohorts generally decline.

Here we see a rapidly, steadily growing community; every cohort adds some members that tend to stick around. If your chart looks like this, you’re doing something right*.

This is a typical support forum, you can see that the vast majority of people only sticks around for a single quarter; note that the first two cohorts remain active (they are probably company staff).

I really really love this one, it shows a very seasonal community (gardening focused), but it also shows that they have new members participating nearly regardless of the season. It has a healthy base and a portion of nearly every cohort stick around.

* assuming increased engagement is one of your goals


(Bas van Leeuwen) #4

I’d love to hear what other people use as measurements and as cohort definition!
The experience we’ve had so far just leaves us with way to many options for cohort definitions, currently we’re experimenting with

  • First post date
  • First post month (as in, group all those that post in august, no matter if it’s 2014 or 2015)
  • First post subforum
  • Topic of first post
  • First post in introduction topic yes/no
  • Join date
  • Date of first visit (this one is technically difficult)

And some others :slight_smile:


(catykobe) #5

@Bas_van_Leeuwen These charts are awesome! Your product is really coming along nicely… great work! :slight_smile:


(Richard Millington) #6

Jesus…this is amazing @Bas_van_Leeuwen.

Mind if I share these elsewhere?


(Bas van Leeuwen) #7

Share away! :slightly_smiling:


(Alessio Fattorini) #8

Woah! This product is amazing! Monthly price is high for my budget, otherwise, I would have already bought it!!


(Richard Millington) #9

$35 per month is too high? :slightly_smiling:

We thought it was too low!


(Richard Millington) #10

I also really like this as a benchmark of comparing the skill of different community managers. The ability to retain members (and increase retention rates) is a pretty great skill to have.


(Rebecca Braglio) #11

this is fascinating. Extremely interested in the types of tools being used to measure this…


(Rebecca Braglio) #12

I’m sorry -what is the name of the software/tool you are using?


(Bas van Leeuwen) #13

Hi @rebeccabraglio, it’s a tool we’re building ourselves, it’s called Community Analytics and it’s used to analyse communities (what’s in a name :stuck_out_tongue:).


(Rebecca Braglio) #14

lol fabulous! wonderful work!

I just checked it out - could this work on a platform built on cold fusion?


(Bas van Leeuwen) #15

In a word, yes :slightly_smiling:

(I’ll send you a private message, don’t want to come across as spammy)


(Richard Millington) #16

Does anyone do a cohort analyses at the moment?

Just curious. Thinking of writing up a post on it.


(Nick Emmett) #17

A post would be great @richard_millington - I’ll be honest it’s a fairly new concept to me.
I have a Beta feature in Google Analytics but don’t really know what I’m reading about!
Can already see from this thread though that there’s clearly a ton of value - @Bas_van_Leeuwen tool looks great.


(Angelo Luciani) #18

Any trial accounts?


(Bas van Leeuwen) #19

Thanks @Nick_Emmett!
The cohort analysis in Googel Analytics is interesting as well, though it is a bit too short term for my taste. You can see in the UI however that they have bigger plans (there are a lot of dropdowns with only a single option at the moment); it’s very interesting to see where they’ll take this :slightly_smiling:

@aluciani, please get in touch via a PM, we can probably arrange something


(Alessio Fattorini) #20

That’s my 2015 cohort analyses, looks great :slight_smile: we started in February, thanks to @Bas_van_Leeuwen
I can’t get why Q3 2005 was so bad in terms of long-term engagement