I thought I’d share some more material with you with regards to the 250 Email Experts on Twitter blog. This material is more or less a byproduct that fell off while extracting the email influencers. That means the numbers are quick & dirty – mainly because the geo data append lead to inaccurate results. Yet, they are still good enough to end our working week: Let’s have a look.
First, what countries are popular among the top email influencers on Twitter – where do the experts originate from? You may have guessed it:
I used the geocoding API from Google to grab the coordinates from locations, which Twitter users specified in their profiles. Most results seemed pretty good, however some were completely wrong (Thailand, Namibia,…). That’s because the locations in the Twitter profiles were either ambiguous, or Google simply didn’t make it – like in this case of @minethatdata:
Obviously it needs some data cleansing before applying serious analytics.
Anyway, a big part of the email marketing opinions are shaped in the United States. 70.1% of the email experts originate from there within the network that I grabbed from Twitter. It’s even more, because most of wrongly geocoded users are from the US, too.
Can we locate the influencers even better? Let’s look at a states heat map:
Influencers mostly originate from Silicon Valley and California. In addition, there’s another rather big network located in the New York area. Both are probably the regions with the highest population density.
How does this look in the context of a worldmap?
(Note: The q’s on the map are glyph errors from a PDF export. They should have been circles, too.)
One could also explore how important tweets about email marketing would probably spread across the globe. Follower relationships indicate that:
For instance, the PDF below shows a little snapshot of where tweets from European email tweeters mostly land in California:
Yep, it’s a clutter. Perhaps it becomes a little bit clearer, when we focus on the more influential people:
No, it doesn’t. 😉 But that’s because the PDF is static. It’s more fun to play around with the data interactively in Gephi. That’s the tool of the trade here.
Last, but not least: Does the number of tweets correlate with the number of followers? This was another question that interested me:
At first, the answer seems to be: partly yes, and partly no. As time passes by and more and more tweets go out, a “natural” tweet volume attracts naturally more followers (see bottom-left quadrant). Gains from mass tweeting on the other hand are just incremental – with a few exceptions. However, the red curve can be misleading. This dataset – an excerpt of users, who got less than 30,000 tweets and followers – is extremely scattered. It’s difficult to see tendencies from a quick glance.
Finally, it would be interesting to also include “age of the account”* as a third dimension, and to look at the characteristics of those, who perform above average (top-left quadrant), and those, who perform below average (bottom-right quadrant or below the smoother). For example, how is it possible to have more than 25,000 tweets, yet not even 500 followers (bottom-right), if not using a bot. But all this would be another story.
Have a nice weekend!
The 3D plot below shows, how the number of followers (z-axis) depends on
- the number of tweets
- the number of people someone follows
- age of the twitter account
(color brightness – yellow is a fresh account, dark red means aged)
The regression plane explains about 56% of the variation. Click and drag your mouse to explore the relationship.