Friday, 2 April 2010

Twitter network strength calculated - for real



Most daring title ever, I think - and I've had a few so far
Starting with a few disclaimers before people try to bite my head off and spoil a good case:
  1. I used my stats from dlvr.it
  2. I picked new blog posts being announced by people themselves
  3. I compared Umair Haque, Tim Kastelle, Capgemini and myself
  4. I'm comparing direct reach, extended reach, clicks and retweets to Twitter network size
  5. Clicks form the very basis of my argument. I know, I know, but a click is the closest you can get to (measuring) a blog post
I don't have the time nor the means to undertake this on a bigger scale, but I'm sure there are people out there
who do. This post is meant as a case-study and I'm trying to make a point so the figure-fetisjists (like myself) can have another go at quantifying quality as much as possible without actually taking the quality out

Having said that, here's what I did.
I use dlvr.it for posting tweets and blog posts, for my own blog posts as well as for others, like my company's. Some of my posts I also post on my company's blog. Not all, as I don't see all of them fit for that. Out here I'm much more present than when I represent my company
Dlvr.it is a very handy tool that gives you location, clicks and retweets of something you posted. So, for each post, I can see how many clicks it got, from which location and application clicks were made, and who retweeted a tweet, as well as the direct, indirect and average reach a post got. More than enough stats to play with...

After a while, something strange occurred to me: there was actually a  big difference between the figures my posts got when posted on my own site, and when posted on my company's site. Of course my site is much better-looking than my company's site and has a great lot more functionality and 'goodies', but the posts were (almost always) exactly the same in content and title - and that's what makes them perfectly comparable

Twitter network size
@Capgemini has 3,800 followers currently, I have 540
With those figures in mind, it's easy to calculate that my company's network is 7 times as big as my own
That would easily lead to the assumption that my company's blog posts get 7 times as many clicks
#Wrong

 Average direct and extended reach
On average, blog posts from my company have a direct reach of 2,500 people, and an extended reach of 7,000 people. Mine have a direct reach of 625 people, and an extended reach of 1,800 people.With those figures in mind, it's easy to calculate that my company's network is 4 times as big as my own
That would easily lead to the assumption that my company's blog posts get 4 times as many clicks
#Wrong

Average (re)tweets
On average, blog posts from my company get 10 (re)tweets. Mine get 4.
With those figures in mind, it's easy to calculate that my company's network is 2.5 times as big as my own
That would easily lead to the assumption that my company's blog posts get 2.5 times as many clicks
#Wrong

Average clicks
I could reveal the secret straight away, but like to keep you in suspense a bit longer.
Let's just do some more assumptions based on the figures calculated above. You do know how I feel about assumptions, I hope, but these are good assumptions: we're going to verify them all the way, right away.
My company has 7 times as many followers as I do, a reach that is 4 times as big, and gets 2.5 times as many tweets for a post. So, assuming away, my company's posts would get in between 2.5 and 7 times as many clicks as mine. Right?
#Wrong
Here it is:
On average, blog posts from my company get 41 clicks. Mine get 65.
With those figures in mind, it's easy to calculate that my network is 50% stronger than my company's
#Right

Example
My biggest post got 165 clicks, with 11 tweets, a direct reach of 1,885 and an extended reach of 8,739
My company's biggest post got 97 clicks, with 11 tweets, a direct reach of 3,850 and an extended reach of 8,652

So, my company's network has a reach that is 5 times as big as mine. But their blog posts only get retweeted 2.5 times as much as mine. That's an indication that the bigger the reach, the weaker the ties: not every follower is in it for the attention. I suspect some are just following my company 'out of support' so to say
And of course, their blog posts only get 2/3rd of the clicks that my blog posts get. Even if I put my posts on their blog, they get half the clicks or even less. That last bit is distorting the picture a bit, as I only put (some of) my finest posts on my company's blog

My conclusion: my network is way stronger, dense, close, tight-nit, interesting and attracting attention than my company's. Way. It's at least 150% as strong (mere clicks), but possibly 1,000% (clicks multiplied by Twitter network size ratio), or anything in between, it depends on how you measure

What do you think? I'm really anxious to get your opinion on this!

Now, on to Tim Kastelle and Umair Haque. I ambushed them on Twitter, waiting for them to announce a new blog post. Yes, #disclaimer this is going to be based on one blog post for them only so here they are: Tim's post and Umair's post
Umair has 27,600 followers, Tim has 850
I just counted the (re)tweets, Umair got 425. That's huge. Tim got 6 tweets
According to dlvr.it, Umair got 16 clicks, and Tim 26

Needless to say, I'm lost here. 16 clicks is nothing for someone like Umair. Compared to 425 tweets, it's inconceivably little. It would mean that over 90% of Umair's Twitter followers mindlessly retweet his blog posts without even clicking on them? That really can't be true of course! I hope Umair can help out here...
In the meantime, I'll look for another "click-tracker" and might update this post later. As for now, I'm off to the beach with the family

Update April 5th 11:05 AM CET:
Back now and having recovered from Trance Energy 2010, here's the final math:

This post got 206 clicks and has now "run dry" I think. Direct reach of 1,200 and extended of 275,000. That last is because @Twitter_Tips (175K followers) picked it up which led to an extra 17 tweets, and 11 clicks! There's another blog post in that, of course, as that's exactly what I mean with the retweet-to-click ratio inverting at some point when a person turns into a brand or organisation (see comment below). Using my own Twitter Search tool, the original @Twitter_Tips tweet triggered a few automated retweets, like you can see here (check the seconds in the timestamp):

2010-04-03 15:02:37 Nice Post: Twitter network strength calculated - for real http://j.mp/dhs830 (via @Twitter_Tips ) sahilmalhan (Sahil Malhan)
2010-04-03 15:00:45 RT @Twitter_Tips: Twitter network strength calculated - for real http://j.mp/dhs830 datalore_tv (Datalore)
2010-04-03 15:00:29 RT @Twitter_Tips: Twitter network strength calculated - for real http://j.mp/dhs830 tashamiel (Tasha Sefrida Dimas)
2010-04-03 15:00:20 Twitter_Tips: Twitter network strength calculated - for real http://j.mp/dhs830 : Twitter_Tips: Twitter netwo.. http://bit.ly/ckIgwq RTSamPolanco (Sammy Polanco)
2010-04-03 15:00:19 RT @Twitter_Tips: Twitter network strength calculated - for real http://j.mp/dhs830 lettersandessay (Ley Marie)
2010-04-03 14:59:41 Twitter network strength calculated - for real http://j.mp/dhs830 Twitter_Tips (Tips, Tools, Status)

8 reacties:

Steve Keifer said...

Very interesting analysis. I have many debates on the topic of whether people want to follow individuals or corporations online. I think your statistics reflects the fact that in the web 2.0 world, people are more interested in reading blog posts and following tweets from interesting people versus interesting companies.

Martijn Linssen said...

Thank you Steve! And long time no chat :)

I agree with you, and see something very interesting happening.
Out here, as well as IRL, we still "want to be seen" "with the rich and famous". That's why we follow and retweet, or dress like and do like, our icons
However, our real interest will be expressed in actually reading, or listening to what they have to say or do
In web 2.0, that interest is measurable in page views or clicks, I think
I found some other ways to look at the stats:

Capgemini Me
Follower-reach ratio 1:3 1:3
Follower-retweet ratio 380:1 135:1
Follower-click ratio 93:1 9:1

but I'll probably write another post on that. By the way, as we speak, this post is now my most-clicked ever. I wonder what would happen if I'd post it on my company's blog...

Tim Kastelle said...

Very interesting post Martijn! I'm a stats junkie too, so it definitely spurs a few thoughts.

First off, the one negative - I am very skeptical of the numbers from services like dlvr.it. Ken Gillgren and I have done some comparisons of the stats reported by ow.ly and bit.ly and both of their click counts seem to vastly over-rate the number of people actually clicking through to our blogs according to our google analytics stats. Neither of us is really sure why this is. But personally, I find the analytic sites like GA and clicky.com to be much more reliable indicators of traffic and sources. If you're not using one of those yet, I recommend them both highly.

Second, I'm not sure about your experience, but mine is that on my blog, neither retweets nor hits are normally distributed. Both have skewed distributions - they're not quite power law distributions, but they are heavily right-skewed. For example, over the past month I've averaged 13 tweets per post. But 21/30 posts in that time have had less than 13 tweets. That group is normally distributed around an average of about 7 - but the overall number is skewed by the 'hits' - there have been a couple with 40+ tweets, and a handful more with 20+. Same story with page views - the highest is a post with 345, the lowest is 29, and again, 21/30 are below the mean. So in some respects, blogging is like publishing books or making movies - the whole thing is driven by hits, but it's really hard to know in advance which ones will work.

Finally, with regard to Umair's numbers, those just don't seem right to me at all (which goes back to my first point!). I know for a fact that a lot of his followers click through on his his tweets - 2 out of my 5 highest traffic days ever have come on days when he has tweeted one of my posts (including that 345 hit post mentioned above!). The last time he did this, I had 90 visitors to my blog in the next 40 minutes. So only 16 clicks on his latest doesn't sound right to me at all.

With regard to your posts here versus on the Cap Gemini site, I think that Steve has it exactly right - people are a lot more interested and engaged with other people than they are with firms.

Thanks for the mention here, and thanks especially for the interesting analysis!

Martijn Linssen said...

Thanks Tim, knew you were a junkie :)

I appreciate the "yes but" as I have a few too but I don't understand what you mean by 'services like': do you mean free short-url services?
However, if they over-rate for one, they should over-rate for all, and that wouldn't matter for the relative ratios - would it?
I compared 55 posts from Capgemini with 30 posts of myself - that should give a nice basis
Anyway, I've signed up for GA now and it's on my site, let's wait and see. Will be great to see how GA compares to dlvr.it for the same content!

I'm not a mathematic so deviations and normal distribution don't make sense to me. But I can tell you, like the little blog post experiment we both had, that it's not very predictable at all "to tell clicks and tweets up front" regarding a certain post

Like I said, Umair's "stats" don't make any sense at all, I wholeheartedly agree with you here. I checked back and Umair's first post announcement was at 2010-04-01 16:19:40 by Antonio Figueiredo. It wasn't until 2010-04-01 16:26 that it started to get mass-retweeted, however. I tweeted my tweet at 2010-04-01 18:01:43, considerably later. In between, there had been 162 tweets about the post that I missed
So, that would mean that 290 (now at a total of 452) tweets would have led to (now) 20 clicks - I don't think so...

I agree with you and Steve about interaction, but am now thinking that people turn into icons at some points. Is Armano a person, a brand, a company? Zappos? Scobleizer? There's a grey area, and in that grey area ties get weak and at some point I firmly believe that the ration turns inside-out; there will be more tweets about a post than actual clicks on it: it has become a live gossip stream...

Tim Kastelle said...

By 'services like' I mean anything that counts clicks through a proxy address. I suspect that their errors are regular enough that the relative numbers will be meaningful. But, because it is not at all clear what exactly they are counting (as the Umair example shows), I'm not 100% confident of that.

Your last point in the comment is really interesting - I agree that at some point people become brands and that it's then difficult to figure out what is happening with tweets & clicks...

In any case, it's all an interesting discussion!

Lee Provoost said...

One reason that might also explain this is the fact that the background of the people inside Capgemini's network is quite differently than the people in your network (and I'm not talking about the ones that just support Capgemini).

Think about journalists, competitors, vendors, etc. I have a hunch that your network is of a higher quality (engagement wise) than Capgemini's, also partly because your network is bit focused around the things you are interested in. @Capgemini follows more of a "whoever wants to follow me is fine", whilst you will probably spend much more effort in engaging with people that you are interested in and build your network like that.

One last thing that wasn't completely clear to me but probably can have a big impact: was it only you that used the divr.it for posting your own's and capgemini's posts in your tweet stream or did @capgemini do that as well for its own posts?

But regardless of what method, I agree with Steve: humans like to deal with humans.

Bill Flitter said...

Martijn, this is a great analysis. Have you added click tracking with Google Analytics to your dlvr.it links? You can find out how to add it here: http://gsfn.us/t/wsib

Cheers,
Bill Flitter
Co-Founder/Dlvr.it

Martijn Linssen said...

Tim, I now have Google Analytics, and (thanks to Bill Flitter!) added click tracking with Google Analytics to my dlvr.it links

Lee, thanks for the compliments :) - I used my own dlvr.it stats, but I'm (almost) always the first to tweet Capgemini blog posts

Post a Comment

Thank you for sharing your thoughts! Copy your comment before signing in...