Jun 8, 2015

$$$ No Money, No Data $$$


In the previous post,  I used Nielsen’s “stadium experience” to kickstart a really bad metaphor.  This exhausting failure suggests that Twitter’s API only offers analytics to those with backstage VIP access.

For the low, low price of $14.99, I purchased data from a service that would generate data visualizations using Twitter’s API. I used this software to track three hashtags (#GameOfThrones, #Scandal, #MadMen) and spent several hours turning spreadsheets into Pivot Tables before finding an obvious flaw.



While filtering and sorting the raw datasets, I realized that a large segment of data was unavailable for timestamps synonymous with show’s live broadcasts.

tab 1: WHERE THE HELL IS ALL THE DATA BETWEEN T20:46:49 & T21:58:57???

This was the first of roughly one gazillion (est.) filtering methods applied across three sets, all revealing a major data gap starting around the time the episode began on the East Coast.

Using a relevant portion of my own twitter activity during Mad Men, which was not motivated by this study, I also located two tweets posted using #MadMen during the EST broadcast of The Milk and Honey Route on May 10th, 2015.

The dataset did display posts I made before and after the live broadcast, and yet this really silly tweet posted at 10:50PM EST was notably absent. I’m going to assume that the software isn’t making any judgements about my sense of humor.

tab 2: Oh, look at that! Seems like we’re missing something between T23:15:51 & T00:44:09??

Since a huge portion of Twitter activity was missing, I couldn’t use this data to make a legitimate assessment about user behavior or audience sentiment during a live broadcast.

While this may be a perfect illustration of why big data is not always reliable, it forced me to whittle down my topic into a more manageable size. In an attempt to reconfigure my research to fit the constraints of the data, the next post will explain how this failure led me to redesign a totally unacademic yet somewhat legitimate research project.

I did not mention the name of the service because while it took them over a week to answer a late-night e-mail from an over-caffeinated student, they did issue a refund with no questions asked.

After explaining my plight in a slightly more frantic manner, I was met with the following response:


taemailOkay,  so I didn’t exactly research the restrictions on Twitter’s API, but yes – it seems fishy.  But in my defense, Twitter’s extremely detailed developer agreement was updated on May 18, 2015, a week after I first contacted the service.  But seriously…

You can’t even get a quote from Nielsen Social without first requesting a demo by entering your relevant industry information.  The same is true for Hootsuite and Socialflow, two of the top software options for those wishing to track tweets in real time.  Trackr displays pricing options their site, and premium subscriptions range between $97- $447/month.

For now, I’m going to assume that Twitter is milking the value of social TV analytics, and $14.99 won’t even buy you nosebleed seats in this expensive stadium.

I promise to lose that metaphor.