Big data may be the technology that everyone’s talking about, but that doesn’t mean it is flawless. Big data has created havoc in some cases—and the reasons can be anything, such as detection of false positives, lack of tools, technical glitches, low quality data, wrong data, or unnecessary data.
With such errors, it may be possible that the results may be completely different from what you expected. Moreover, the results are sometimes not analyzed, which can lead to unpleasant results.
Let us take a look at some examples where big data went bonkers.
Google is feeling under the weather:
Probably the biggest and the most well-known big data failure was Google Flu Trends. This web service was started in 2008 with the aim of predicting flu outbreaks in about 25 countries. The logic was simple—just analyze Google search queries about flu in a particular region. This was compared to a historical baseline of flu activity level in that region, and based on the results, the activity level was reported as low, medium, high or extreme.
Sounds cool, doesn’t it? But it wasn’t.
At the peak of the 2013 flu season, GFT failed—and how. It was off by a whopping 140 percent! How did this happen? The algorithm was flawed and did not consider several other factors. For example, searching for terms like “cold” and “fever” did not necessarily mean that people were searching for flu-related diseases; they might just be looking at seasonal diseases. In 2009, it also missed out on predicting the outbreak of H1N1 entirely. GFT could not recover from this flu, which ultimately led to its untimely demise in 2013.
Targeting the wrong audience:
Big data is good and sometimes it can give you the right answers all the way. But sometimes, you have to pay the price for predicting everything right if you don’t use the results wisely.
Well, Target learnt this the hard way. It ran a lot of algorithms and analysis on customer information, such as shopping trends, what they’re buying, where are they buying from and personal information like anniversaries, birthdays, and marital status. With extensive data crunching, Target targeted (pun intended) expectant mothers through their buying patterns and offered them personalized pre- and post-natal items.
The idea seems wonderful from a marketing point of view, but one instance caught them off guard. A furious father stormed into Target for offering his teenage daughter these deals via mails. He was shocked as to why Target would give such recommendations to his innocent daughter. But what he didn’t know was that his teenage daughter was indeed pregnant and was hiding it from her parents.
Privacy and sensitivity are important, and being smart about how you use the results of analytics is crucial—otherwise, you will compromise your client’s trust.
Big data playing with emotions:
Analytics requires a lot of data. Agreed—but is everything necessary? Probably not. OfficeMax, an American office supplies retailer, sent a letter to a certain Mike Seay in Illinois; however, the contents on the letter were not called for. The letter was addressed to “Mike Seay, Daughter Killed in Car Crash.” Seay lost his daughter in a car accident almost a year ago. And sadly, just before leaving his home to attend a counselling meeting for grieving parents who lost their kids, he received this letter.
This incident raised several debates about how companies use data and if data is checked before running analytics programs on them.
Nobody can keep calm:
When you do not pay attention to your codes and their results, things can go awfully wrong—big data wise! Amazon was in the middle of a furore when it published offensive T-shirts on its site such as “Keep Calm and Rape A Lot” and “Keep Calm and Punch Her.”
Obviously, Amazon was forced to take them down. Solid Gold Bomb, the seller company, put the blame on poor programming and analytics. These phrases were automatically generated using a scripted computer program running against a huge dictionary. The results were downright offensive, with the bigger blunder being that nobody checked the results.
Analytics is bound to give you answers—but, in this case, obviously not the right ones.
Mitt Romney’s failed bid for presidency:
In 2012, Republican presidential candidate Mitt Romney used big data with big dreams of sitting in the White House. His campaign tech team developed Orca, a big data platform, for giving insights about what was happening at polling stations. This would then later assist volunteers in going out and getting the vote.
The plan seemed wonderful—except for the fact that the execution wasn’t. Blunders were plentiful—technical glitches, bugs, lack of resources to use the big data platform, and lack of training. The dependence on Orca led to a lot of resource and money wastage, and strategies based on the outcome of this program were nothing but futile.
Big data cannot be solely responsible for Romney’s loss, but one can’t help but wonder what would happen if big data was used in the right way. President Romney may have been a possibility.
But he made a big mistake with big data—and this meant that he couldn’t sit in the big office and make big speeches. He is probably doing a facepalm using a baseball ‘mitt’!