Filter by tag

The Unfriendly Skies: A Data-Driven Breakdown of What We Hate About Major Airlines


Every major American airline maintains a customer service twitter feed. It seems like a thankless job, consisting largely of replying to endless grievances about weather-related cancellations and spotty in-flight wifi with “We’re sorry you’re upset. Please DM us your reservation number.” Then you get yelled at.

That said, it’s a tremendous source of sentiment data. We looked into some 20,000 tweets from February 1st until today to uncover not just whether twitter users were relentlessly negative towards their air carriers (they were!), but exactly what it is they were grousing about in the first place. And as it turns out, when we complain to customer service, we’re mostly just complaining about customer service itself.


About 36% of all negative tweets airlines receive are about the service they’re giving their customers over the phone, on Twitter, in chat, et cetera. You know, tweets like this:

Seems pleasant. At any rate, you should note that the categories in this sentiment analysis job allowed for problems with the airline’s website, issues at the airport, with flight attendants, and of course, the classic gripes about lost luggage and cancelled or late flights. In other words, those categories were not included in the customer service bucket. That note in the first paragraph about this being a thankless job seems suddenly more accurate.

The Delta Between Airlines (Sorry)

Let’s start with something simple: the positive, negative, and neutral interactions. This is fairly basic sentiment analysis, but it helps inform the other information we’ll be covering, so starting here makes sense. Here’s breakdown of positive (green), negative (red), and neutral (blue) tweets sent to each airline:


You can see that US Airways was swamped with negativity, with 77% of all conversation being something akin to the tweet we shared above, often times with far more unhinged invective. Virgin America, meanwhile, had a fairly equal distribution of positive, negative, and neutral conversations with their passengers.

Now, here’s where it gets really interesting. Since we can ask our contributors to not just note if a tweet is mean and bad and crabby but we can also find out what the passenger is whinging about. That way, we can get a better look into what problems these individual airlines have.


For example, American scored lowest as to their actual customer service, with some 43.5% of all negative tweets centered around that issue alone. Delta had the largest percentage of complaints about late flights (but the fewest about cancelled flights, strangely enough), while United hears most about lost luggage. Virgin users had a higher instance of problems booking flights and more people complaining about the flights themselves, though some of those were actually fliers who really, really find that Virgin “Safety Dance” song a repugnant earworm.

Sorry if we got that in your head. We regret the error. Complaints may be sent to @crowdflower on Twitter.

 How We Did It

We scraped Twitter for about 40,000 tweets to the official airline handles. Then, we created a sentiment analysis job where we looked asked contributors to read individual tweets (and conversations with customer service agents) and determine whether they were positive, negative, or neutral. After that, we had our contributors categorize the negative tweets into the main sources of online kvetching. Those were:

  • Cancelled flights
  • Late flights
  • Problem with the flight itself (bad wifi, turbulence, overhead bin space, etc.)
  • Lost luggage
  • Damaged luggage
  • Problems with flight attendants or airline staff at the gate
  • Long lines in the airport
  • Issues with the booking process or frequent flier miles
  • And of course, problems with customer service itself

If you want to learn more about how to run one of these sentiment analysis jobs, well, you’re in luck! We’re hosting a free webinar tomorrow (Thursday, February 26th at 10 a.m.) with Edelman, a global PR firm. You can get more details about what we’ll cover in this post, or just register here.

And if you’d like to take a gander at the data above, you can poke around our Tableau visualization or download the data set.

But if you’ve learned nothing else, be nicer to the poor people who run airline Twitter handles. We doth protested enough already.

Justin Tenuto

Justin Tenuto

Justin writes things at Figure Eight. He enjoys books about robots and aliens, baking bread, and is pretty sure he can beat you at pingpong.