#yay data!

So I saw some posts yesterday with several Nonnie's and OPs angry/mad/worried about certain trends lately and I thought it would be a fun exercise to explain, using math and my handy dandy Python scripts (fun!), why trend attempts beyond just showing general appreciation for the show/actors/your ship are completely meaningless and why. 

 As a person with severe anxiety, who sees these types of things and in her brain says "What if some big shot at ABC sees this and changes the whole story!" despite how ridiculous and irrational that line of thinking is, I am comforted by real numbers and data and find that a lot more reassuring than a thousand people telling me it doesn't matter.

I hope others in this same boat, or those who handle the comforting of those in the same boat, find this helpful. 

And there’s a chart!

So on average trend attempts generate between 20,000 - 50,000 tweets for the OUAT fandom. The highest I've ever seen one get to, after several years here, was the OUAT 100th were all fandoms came together to celebrate and together we generated over 100K. Regular weekly/bi-weekly trend attempts don't even come close to that. Basically people are leveraging the quieter times of day/week in order to get noticed and it's less about the number of tweets than it is picking an opportune moment to start your trend.

 So let's dream big here, let's say we have a trend attempt that generates 50K tweets which is the higher end of our average.

 Most trend attempts are scheduled for 1 hour. They last far longer than that, especially if the trend works and people want to keep it going, but we are simulating perfect, best case scenario conditions here so let's put 60 minutes on the clock. 

 Most people participating in the trend are tweeting very rapidly, or have scheduled tweets to occur at very rapid intervals. 

 Twitter itself only allows a daily limit of 2,400 updates which is broken down further into an hourly limit of 100 tweets per hour (including RTs). This is per their API documentation, but other factors can come into play if you or the application you are using, are making too many calls (basically too many actions within a specified timeframe).

 In order to hit the threshold you would need to send a tweet or RT a tweet every 1.6 minutes. So the maximum I would say any one user will be tweeting during a trend attempt is once a minute. However, let's say a user only sends a single tweet every 5 minutes or so. They are pretty casual in their trend attempt participation. (The reality is people participating are usually spamming tweets as fast as they can).

 But since we have a mix of people participating at different levels of intensity, some just sending one or two, some sending as many as they can, we'll look a few different intervals and we'll also look at an average based on some a Twitter search data that makes up a representative sample. We’ll also talk about some general demographic information.

 So let's say we have 50,000 tweets, being sent at an average of 1 tweet per 5 minutes for each user. Which is only 12 tweets per user participating for the whole attempt duration.

 With those conditions an average of around 4,167 users could be participating in this trend. That's all. And this is under ideal, fairly mild conditions.

 The provided chart should show you how completely inconsequential these numbers are:

   But let’s look at a specific and recent trend!

 So there really isn’t a way to see every tweet that participated in a trend with the restrictions on Twitter’s REST API, of which the Search API is apart of, but what we can do is use the search to get a pretty decent representative sample of users. The Search API queries against an index of tweets, so it searches against a sampling of recent tweets published within the past 7 days. So sadly I can’t compare trend to trend, but we can at least get an excellent random sampling to work with.

 Using a recent trend I pulled as many tweets as I could (around 2285 before I hit the call threshold) and then totaled up the number of tweets per username.  We had a total of 698 unique usernames in our master data set (without the time restriction).

In our sample data, taking a random 1 hour interval we had users that sent as many as 87 tweets (in one hour! Out of a SAMPLE dataset) for that time frame down to as little as 1.

Within that hour we had 298 unique usernames in the sample sending out tweets.

Overall we averaged 6 tweets per username.

 Looking back at our handy dandy chart, if each user is sending an average of 6 tweets for a one hour trend attempt that’s around 8,333 users. Again, not exactly a staggering number.

 If we’re talking demo, thinking the show averages around 4 million viewers, those 8,333 users only account for .002 percent of the overall viewers.

 BUT the demo is based on US households in the 18-49  age group. Our sample data is not. Our sample data includes international viewers who watch via stream as well as users outside the ideal target age range.

 Internationally, 31% of all Twitter users are under the 18-49 demo. So we should adjust our sample data accordingly.  That leaves around 5,750.

 And only 62% of all Twitter users are in the US.

 So that leaves us with 3,450 users. That’s about .0008 of the demo. Most people on Tumblr have more followers than 3,450 users. It’s a completely inconsequential number of people and NOT one that any network is making decisions based on.

 Not only are a lot of these users international viewers who watch via stream and contribute exactly zero to the numbers ABC needs to sell ad space, some have multiple accounts, a few aren't even viewers of the show anymore, and far, far less of even the American viewers within he demo watch live or during the C7 timeframe they are looking at. And ABC knows this. They have far more sophisticated tools, metrics, and market trend info than Steffie with her Python script at home.

 And honestly only a handful of those users left after we account for all that have a Nielsen box, DVR, Hulu account to actually count.  All of which the people who matter know. These users and their desires/whims are absolutely useless to them.

 Trends are still a great way for the fandom to interact and to show appreciation for our favorite parts about the show, and I both encourage and participate in them from time to time, but no one is making decisions off this data, and no one of importance is paying attention to it.