How to Evaluate Predictive Analytics for Your Business
Did Big Data blow it in the election, and if so, what does this say about the technology’s ability to deliver all manner of insight into complex business and technological processes?
The question has been circulating throughout media circles since the electoral map went from blue to red last Tuesday, casting doubt on the transformative power of analytics and automation that is supposed to propel the enterprise into the next-generation digital economy.
As usual, however, the reality is mixed. The fact is that the Big Data engines crunching the poll numbers worked exactly the way they were supposed to work – it was the expectations as to what they could actually deliver that were inflated.
TopTechNews’ Matt Sedensky hit the nail on the head the other day by noting that the key problem with the predictions was that the data was wrong. In the classic garbage-in/garbage-out scenario, even the most state-of-the-art technology will produce poor results if both the data and the assumptions used to condition that data are faulty or incomplete. This is the reason why people get into trouble basing major purchases on online reviews: They can be either too rosy due to manipulation by the reviewee or too negative because customers tend to post only their negative experiences. In all cases, whether political, commercial or social, the quality of data must be assessed carefully before faith in the results can be established.
The other thing to keep in mind is that most of the polling that got the election wrong hardly qualifies as Big Data, says GVA research analyst David Garrity. The National Council on Public Polls identifies a common sample size of about 1,000 voters, while the typical Big Data sample is a minimum of 100,000 – and by 2020 they will be in the millions. In fact, it was Team Trump’s reliance on expanded data sets provided by London firm Cambridge Analytica that caused them to re-weight their polling data to reflect an increase in populism among the electorate, much the same way the Brexit vote in the UK revealed a markedly different demographic than traditional polling methods were showing.
Other more data-intensive analytics models were also getting it right prior to the election, although these were largely ignored due to the consistency of the traditional models. As RTInsights points out, firms like MogIA pulled together more than 20 million data points from social media to predict a Trump victory, while other models that look beyond mere voter data and into things like economic and foreign policy factors, internal party politics and historical trends were also skewing toward Trump in the final days of the campaign.
So what does all this mean for IT? According to InformationWeek’s Jessica Davis, the takeaway for enterprise executives is to maintain a healthy skepticism for any solution that purports to predict the future with any degree of accuracy. This can be a difficult square to circle for most business leaders because, as KPMG’s Nadia Zahawi notes, the Big Data infrastructure being implemented today has a “black box” quality to it: Data goes in, the analytics wizards do their thing, and results come out, but there are few mechanisms to question those results or challenge the underlying assumptions that produced them.
On top of all of this, however, is the bigger issue regarding the relationship between humans and technology as the technology becomes smarter and more autonomous. The fact remains that no algorithm, no matter how advanced, can replace that spark of intuition and the capacity for judgement that is uniquely reserved for the human brain. At best, technology can crunch data faster and with greater accuracy than people, but that should in no way imply that it is better or can provide a more accurate version of the truth.
In the end, it wasn’t Big Data that failed in the election, it was the hype surrounding it.
Arthur Cole writes about infrastructure for IT Business Edge. Cole has been covering the high-tech media and computing industries for more than 20 years, having served as editor of TV Technology, Video Technology News, Internet News and Multimedia Weekly. His contributions have appeared in Communications Today and Enterprise Networking Planet and as web content for numerous high-tech clients like TwinStrata and Carpathia. Follow Art on Twitter @acole602.
