"How Data Failed Us in Calling An Election"--or some version of this headline--has proliferated articles all across the Web since Trump's win. Here's a newsflash: it wasn't the data's fault. While many people are searching for someone or something else to blame, I think it's abundantly clear who was really responsible for the situation: It was the media prognosticators and the professional blowhards who got it so wrong.
The truth is, data doesn't have an opinion. It's like the Dude in The Big Lebowski--it simply abides. What you do with it and how it's used makes all the difference. So let's not dump on the data and data-driven decision making--which is essential to the future of almost every business--just because a bunch of over-eager talking heads ignored the science, misapplied the mechanisms, and tried to manufacture a true miracle out of a mess of mixed messages which were never going to accurately predict the ultimate voting behaviors of millions of people. People who weren't prepared to share their true feelings with anyone, and especially not with paid political pollsters who aren't impartial.
I believe the most broken aspect of tracking this election was that pollsters were simply seeking affirmation of what they wanted to hear using a deeply-flawed methodology that everyone knows is meaningless in today's mobile and digital world. Just another ugly byproduct of the world of cable news where we only watch and listen to the people who support our outlook on the world.
I think about it this way: If you use a screwdriver to slice roast beef, who exactly do you think is to blame for chunky cuts of meat that look like they've already been chewed? Is it really the screwdriver's fault? Maybe you should have used a Phillips instead of a flathead for a more presentable platter. Or maybe you should have taken a moment to think about whether the instrument you chose would ever be up to the task that you set out to accomplish. We should know by now and freely admit that no amount of historical information is likely to accurately predict certain kinds of human behaviors. Voting isn't the same as buying a vacuum cleaner. You never know what the voters will do until after they've done it.
If a significant part of your prospective voter polling methodology is predicated on calling large parts of a population that no longer answers their phones or doesn't even own a land-line phone, what would lead you to believe that you had any real basis to predict anything that mattered? Then factor in the really depressing fact that the people desperate enough to waste their time talking to strangers on the phone aren't as likely to tell those folks exactly what they really think. The telephone might be the worst possible tool for eliciting the truth about anything.
The reality is, the numbers are neutral at best (except when they're just dead wrong, as in some of the state polls). It reminds me of a thermos: it doesn't really know whether it's keeping some beverage hotter or cooler for a period of time, it just sits there and does its job. The data is not to blame when it's tortured and twisted in support of answers and expected behaviors that it was never capable of predicting in the first place. If we want explanations for why the folks who were supposed to know didn't know anything in the end, we have to question them, not the numbers.