Social Media Anlaytics with Twitter Example
Avinash Kaushik is one of my favorite Web Analytics professionals.
Just had a string of conversations with him recently- and I am sure he won’t mind my sharing this with readers of ChasingTheStorm.
The conversation string stemmed from one of his posts where he discussed analyzing Twitter. This is also interesting because usually Web Analytics professionals speak about “on-site” elements.
While I agree with many of things that Avinash has to say, I had some points of difference.
It is always a challenge to be speaking to someone of his calibre on Analytics- but this is something quite close to me and I decided to out forth my 2 cents.
I am including the conversation here as it happened:
Thread 1- My response to Avinash’s post (with references to points made in his post):
Avinash had mentioned (amongst other things) that
- “Number of retweets per thousand followers” is a good metric as a measure of efficiency and value provided and people voting with their clicks, all rolled into one.
- A metric like “Follower:Followed” ratio is not worth it
To this, my 2 cents (with humility and some apprehension)
*Number of retweets per thousand followers is a good metric.
However, due to the nature of twitter- this would probably not be fair to someone who tweets less Vs someone who tweets more often.(Assuming ceteris paribus).
So if A has 1000 followers and sent 10 tweets a day out of which 1 gets RT-ed, my ratio is 0.1 RTs per thousand followers; and if B has the same 1000 followers and send 100 tweets a day of which 10 gets RT-ed, my ratio is 1 RTs pe thousand followers
I think (number of updates / number of retweets*100) could be a good indicator.
In the above example, this percentage would signify that A is actually equal to B- because the propensity of his tweets to spread is equal to B- irrespective of the number of followers (that is 10%)
This number would signify more closely what is the propsensity of messages to spread and perhaps also the quality of tweets sent (a larger score would mean more of my tweets are found valuable)
“I don’t care about Follower/Follow Ratio-
I think it is disingenuous to follow everyone who follows you just for appearances sake when you have no intention of reading what they all say. Why be fake?”
Quite true. On its own self this ratio is probably not so useful. But maybe this is not a metric to be seen in isolation- rather in a combination with other metrics.
At the least, it could ‘flag’ certain cases:
*There could be some more ways to interpret this metric- significantly more people following you then you following them could denote less ’socialbility’ but more ‘authority’. I mean if I am a celeb/authority – I cannot meaningfully listen to all- so I have a high follower-follow ratio. On the flip side, a low ratio could imply a twitter-er who follows people recklessly (in the hope that they would follow you back)- and in turn follows a significantly larger number of people than other people follow them.
This metric- coupled with a metric like, say- “rate of follower adoption” could give more meaningful insights into the “genuine authority/sociability” IMO
It is quite simplistic yet not entirely useless IMO- if we know what are we using it for (like most other metrics)
Thread 2: Avinash’s response to this (This is not on the post- we communicated over email- but I am sure Avinash would not mind sharing this):
Thanks so much for the kind suggestions, they are wonderful good for thought.
The reason for making the metric per thousand followers is precisely to “democratize” the comparison, so someone with many followers and another with a lower follower count can be compared. The other goal is to set a higher bar for the person with high followers – if you have lots of followers and they are relevant then they would all retweet.
Your alternative is very nice, but it provides a different perspective (note: I am not saying a bad perspective, I am saying a different perspective).
To the last point, I am afraid by using terms like “sociability” and “authority” you are falling into the trap I don’t want to. Sure there is a difference between how many people a “celebrity” can follow vs how many a normal person like me can follow. But in the end the Dunbar Number dictates that there is a pretty hard limit as to how many anyone can follow, and listen. 150. That makes the ratio even less valuable.
Thread 3: My response to Avinash
I know you have a packed schedule and I don’t want to make it a long thread, but somehow this is taking me a while to get convinced. I am quite passionate about “Conversational/social media” metrics, and so will reply with some more of my doubts :
Avinash: The other goal is to set a higher bar for the person with high followers – if you have lots of followers and they are relevant then they would all retweet.
My doubt is this:
- Does the propensity to be retweeted depend on quality of followers or quality of your tweets? Probably it is the latter- and not the former.
- Also, due to the nature of Twitter- an average tweet stays in the stream for a few seconds. If I have a lot of followers from a different time zone (say I am from Singapore and have a lot of people from the US following me), the propensity of getting RTed would be reduced- because maybe many of my followers don’t even get exposed to that tweet.
Whereas if you compare this to the number of tweets I sent Vs the number of retweets, it might be a little more fair for comparison.
Yes I agree both of these achieve different objectives. But even then, I would contest that using ‘followers’ will make the analysis result a bit- well, wide.
Just because all people using Twitter are not the same. And this would make it a little more complex when diving deep:
- This would demand more answers like: what is the mix of followers- (a more homogeneous mix will react similarly to a message) , time zones, how influential are the followers themselves (are more influential people less likely to retweet you), how many people are the followers following themselves- and so on.
Avinash: To the last point, I am afraid by using terms like “sociability” and “authority” you are falling into the trap I don’t want to. Sure there is a difference between how many people a “celebrity” can follow vs how many a normal person like me can follow. But in the end the Dunbar Number dictates that there is a pretty hard limit as to how many anyone can follow, and listen. 150. That makes the ratio even less valuable.
You are being humble, but of course you know you are a celebrity.
Apologies on the terms. The idea was not to indulge in “buzzword play”- rather defining what exactly could this measure.
So how I would put it is this way – on channels like Twitter it is difficult to have a meaningful and personalized one to one communication. By the Dunbar number, anyone who has more than 150 people in their network is doing injustice to everyone.
Relationships on virtual networks are different from real life relationships, in my opinion. And realistically speaking, even in real life, we filter communication and respond to only a fraction of communication from even our close ones.
To that- the ideal “follower to follow” ratio need not be ’1′ . As you mentioned- you need not follow everyone who follow you.
But if we put “cut off points” it might have more to say.
Like if the difference is more than 30%- you either follow everyone recklessly (30% more following than followers) OR do not want to follow a third of people that follow you (which means either you are lethargic or plain “unsocial”).
Couple this with “rate of follower adoption”- and it could tell you whether you acquire followers organically or inorganically.
Just my 2 cents. Reminds me of that meaningless REM song (Losing my religion), where they quip: “Oh! No I said too much”
Think about it