FreakyFriday: Weird of the Day – Google Doesn’t ‘Get’ the Flu

Looks like Dr. Google misdiagnosed this year’s flu season. A widely circulated article posted by Nature this week describes how Google’s flu tracking application ended up overestimating this year’s epidemic.

If you’re not familiar with Google’s flu tracker, it’s one of a number of projects falling under Google.org, which seeks to leverage Google products for social good. Check out their site – not only are the projects really cool, but they show how the data generated by search goes far beyond marketing. The flu tracker attempts to measure the spread and severity of outbreaks based in part on Web searches – that is, people searching for flu symptoms and other related topics.large water slides for sale

Neat, right? And, actually, the project has historically been fairly accurate, at least enough so that medical researchers planned to take a serious dive into Google’s numbers this flu season.

The flu was bad this year; just not as bad as Google predicted. The search giant’s numbers doubled what the Centers for Disease Control actually observed, according to Nature.

It doesn’t take Nate Silver to figure out some of what went wrong. Among other things, Google didn’t account for the – excuse the pun – viral nature of this year’s flu season. Media coverage about the predicted severity of this year’s flu – including stories about Google’s incredible flu tracker – boosted the number of Web searches for flu-related topics, throwing off Google’s algorithms.

Nature posits that this is a temporary setback for a promising approach, but GigaOM and others point out that the whole issue raises important questions about the reliability of Web data.

My takeaway – and the takeaway for search marketers – is that search data, click-thru-rates and all the other numbers we pull out of the Web are incredible, powerful tools, but they still need real-world context to be used effectively.

The freaky thing is, numbers don’t always mean what we think they do. And neither can they be divorced from facts on the ground. For a blunt example, Carnival Cruise Lines is probably seeing a spike in Web searches this week coinciding with its well-publicized fiasco in the Gulf of Mexico. And anybody who is familiar with cyberchondria knows that not everybody who searches for disease symptoms on the Web is actually sick.

Digging into the numbers still requires human expertise. Google alone can’t cure what ails you – or your business.

Freaky Friday: Weird of the Day – #drunknatesilver Ruins Freaky Friday, Debunks its Freakiness

By: Alex Dalenberg, iCopywriter Blogger

“Statistically speaking, Friday isn’t freakier than any other day of the week.”

OK, New York Times über stats geek Nate Silver didn’t actually say that, but we imagine that he might, especially after putting back a few. Welcome to our favorite Election Week meme, #drunknatesilver. Twitter is having statistically unprecedented amounts of fun (unprecedented except to Nate Silver) imagining the math man carousing about the town to celebrate his dead-on prediction in the presidential election.

This isn’t that surprising. Silver was more or less dead on predicting the 2008 presidential election, missing just one state. And while some see Silver as a wizard, he uses a fairly straightforward forecasting model that averages the results of numerous polls and gives more weight to the ones with a better track record of picking the eventual winner. The idea is that, this way, outliers have less of an effect on the prediction.

A few of the best #drunknatesilver tweets.

@davelevitan: Drunk Nate Silver stumbles through traffic on the Jersey Turnpike, screaming out what time each driver will get home. #DrunkNateSilver

@kelkulus: Drunk Nate Silver stumbles through the streets, shouting obscenities at the future ex-wives that he has yet to meet. #DrunkNateSilver

@copyblogger: Drunk Nate Silver says “Call me maybe? I’ll know it’s you because your number is …” #drunknatesilver

All kidding aside, Nate Silver is a model of viral success for every blogger seeking to create compelling, vital content. He started as a relatively humble contributor at liberal political blog Daily Kos http://www.dailykos.com, but because his work focused less on the ideology and more on the numbers, he found a much wider readership.

Silver turned his musings and number-crunching into the smash-hit website FiveThirtyEight.com which was subsequently picked up by The New York Times. According to the executive editor, it’s now one of their biggest traffic generators.

His new book, “The Signal and the Noise: Why so many predictions fail, but some don’t” is also worth a read. Not just for political junkies, but any business decision maker who wants to better understand how to sort good information from the bad.

As we see it, here are three lessons to takeaway from sober Nate Silver.

1) Use content to address an unmet need.

Silver saw that most political coverage struggled to rise above the day-to-day minutiae, spin and partisan emotion dominating the news cycle. Silver found a way to let the data speak for itself. Readers looking for a better way to make sense of the political climate flocked to his approach. What need does your content fill?

2) Don’t just aggregate, interpret.

Polling is a mainstay of modern politics, but Silver isn’t a pollster. He’s an aggregator but, more important, an interpreter. The numbers aren’t his, but he explains them. Don’t just retweet and repost. Give readers context.

3) Make your content indispensable. 

Easier said than done, but there’s a reason the bleary-eyed hordes of political junkies keep Silver’s website bookmarked: his take on the polls are can’t-miss content. These days, if Silver isn’t part of your repertoire, love him or hate him, you just don’t follow politics. If only all of us could say that about our industry blogs.

#drunknatesilver says be like him and you’ll have at least a 72.3337492 chance of content success.

Have you checked out iCopywriter lately?

Photo Credit: joewcampbell