Feel free to follow Student Activism on Twitter or Facebook, if you like. You can also read this essay in German, if you like.

December 8 Update: Twitter released an official statement on the Wikileaks trending controversy this afternoon. I’ll have a full response soon, but for now I’ll just say that it doesn’t seem to me that it fits the data I’ve presented here.

December 11 Update: This has been an absurdly busy week in the world of things-this-blog-is-interested-in, but here it is at long last: How Twitter Kept Wikileaks from Trending, and Why.


Okay, this is a little ridiculous.

A week ago, I wrote a piece dismissing the idea that Twitter was actively working to keep Wikileaks out of its trending topics lists. This morning, I wrote a followup in which I continued to express skepticism that any monkey business was going on, but acknowledged that the data were really kind of weird.

Now I’ve gone back and compared long-term traffic patterns for “Sundays,” one of today’s big global trending topics, with those of “Wikileaks,” and I have to say I’m kind of flabbergasted. If the data I have are accurate, something very very strange is going on.

Here. Let me show you.

(click each chart to view full size)

This is the last 180 days of Twitter traffic data for “Sundays,” taken from the Trendistic website. We can see that the word peaks every weekend — unsurprisingly — and that it’s grown only slightly in volume since mid-July. With the exception of today’s large spike, the biggest weekend bump for Sundays was only about double the volume of the smallest.

Looking at trending topic data from Twend It, however, we see that Sundays has trended four times in the last two months — on September 26, October 18, November 21-22, and today. The first two of those “trend incidents” took place on completely ordinary days for Sundays (measured by total volume), and the third, which lasted much longer, took place on a weekend when traffic for the phrase spiked over a longer period of time, but no higher, than it had in the past.

Why is this significant? Because, as I wrote this morning, Twitter claims that a phrase’s novelty is a major predictor of whether it’ll trend or not. But “Sundays” is the opposite of novel — it’s a term that spikes once a week, every week, in pretty much the same way at pretty much the same time. By the criterion of novelty, Sundays should be at a huge disadvantage versus hundreds of other terms — including, yes, Wikileaks — in making the trending topics list.

Take a look at the Trendistic chart for Wikileaks over the same 180-day period and you’ll see what I’m talking about.

The first thing that jumps out, obviously, is that Wikileaks’ traffic pattern is far less regular than Sundays’. There are a few large-ish spikes, some extended periods of lower but still significant activity, and a bunch of long stretches when nothing’s going on at all.

If we look a little closer, we see that there’s very little happening here until the first spike, which comes on July 26. Activity then tails off gradually over the next month, with a second, smaller spike on August 21, and then almost nothing for two months. There’s a third spike — slightly smaller than the first, but much longer lasting — beginning on October 24. Traffic begins rising again in mid-November, spikes on November 29 at a volume four times that of the August peak, and has since settled in at a plateau about two-thirds the height of the November 29 spike ever since.

If you asked me, I’d say — just off the top of my head — that there are three or four obvious places on this chart where I’d have expected Wikileaks to trend, with the most recent being the most obvious. But if we look at Twend It, we see that it trended for thirty-eight hours over the course of three days at the time of the July spike, for less than two hours on August 21, and never again since.

Let me repeat that. Wikileaks trended on Twitter for three days following the first spike on the chart above, for two hours at the time of the smaller spike right next to it, and not at all for the third similar spike or the huge one that’s still going on now.

Weird, right?

But no. That’s not the weird part. This is the weird part.

This chart tracks Sundays and Wikileaks over the last 180 days, with each term’s volume drawn to scale. The red spires? Wikileaks. The blue dust at the base of those spires? Sundays. The tiny blue uptick in the lower left hand corner of the chart has had Sundays trending for more than twelve hours today, while Wikileaks has been completely dark since August.

What the hell is going on here?

As I said in my last two posts, I don’t care that much about trending topics lists. I’m a big believer in online organizing, but I just don’t think getting your cause to trend is all that important in the grand scheme of things. But this, like I say, is ridiculous.

December 6, Morning | Be sure to read the full exchange between myself and Josh Elman of Twitter in comments — it’s illuminating in its own way. There’s a bunch of other important new info and analysis in the comments thread, too. I’ll be continuing to follow this story on Twitter over the course of the day, and I’m hoping to have a full new post up sometime this afternoon.

Also, you know, feel free to check out the rest of the site. Welcome!

December 6, Late Morning | Blogger Bubbloy has a post up that covers similar ground to this one in a complementary (and complimentary!) way. Be sure to check out the discussion of the “oil spill” trending topic in the second update.