One of the over-arching narratives in the big data movement is that data can provide the necessary insights for businesses to make good decisions. I’m not saying at all that isn’t true.
However, the theory is being put to the test because of the recent U.S. Election, which saw a surprising upset victory for billionaire real-estate mogul Donald Trump. And, to a lesser extent the Brexit polling too has led to many questioning the validity of analytical data.
Questions that should be asked are how is the data being collected, is it good data and how is this data being interpreted?
Pollsters may have woken up on the morning of Nov. 9 to claim their numbers were accurate in the popular vote which Hillary Clinton did win. But that too is saying the Cleveland Indians should be awarded the World Series because they scored more runs than the Chicago Cubs in total, during their seven game series. It doesn’t work that way.
The popular vote discussion does not hold water because voter data was captured from each state and interpreted to lead several of the pundits predicting Clinton would win battle ground states such as Wisconsin, Ohio, Michigan and Pennsylvania, her so-called “firewall”, and therefore win the White House. Those states were all captured by Trump.
Both Sam Wang and Nate Silver, two of America’s most successful pollsters had Clinton winning the election. Wang had it at 99 per cent for Clinton, while Silver had his numbers at around 70 per cent. Even in the 2015 Canadian Election polls had the three parties who could form a government in a three-way tie. As we all know Justin Trudeau’s Liberals won convincingly.
Related Blogs
And, even those who were right before can’t take it for granted they will be consistently correct in the future. Take for example Qriously, the only research firm to correctly predict the Brexit outcome. This London-based firm used “in-app polls” in those same battleground states mentioned above along with Florida, North Carolina, Arizona and Colorado and came to the conclusion that Clinton would be President-Elect. Qriously captured its data via smartphones.
This method is not too similar to the way most businesses doing research for important buying decisions. I’m not knocking Qriously because they too capture data not just by demographics, but also behaviour and B2B sources. Even Qriously, on its Web site, says “current opinion polling methods are broken.”
And, maybe that’s what we should be doing here. I think what the Trump victory proves is that data should be analyzed at its core and the method is equally as important as the net result. Business people, large and small, are going to continue to look for data sources to make informed decisions. An Acquity Group study on B2B procurement found that 94 per cent of business buyers make decisions via Google search, business Web sites, reviews and other third-party Web sites.
Another study from Forrester Research shows 74 per cent of business buyers do more than 50 per cent of their purchasing decision from online information.
So given the vast majority of the pollsters used data insights and came up with the wrong answer, can data be trusted – as much as it is today – with helping business make the right decisions? I think so, but those vendors who are bringing the data narrative to market and presenting it as gospel will now have to revisit their stance on data analytics being the be all and end all of good decision making.
One quick hit before I go. Broadvoice, a hosted voice and unified communications vendor has appointed Ryan Ficken to the role of Channel Chief.
“Can data be trusted…”?
I believe it would be easier to solve this by going back to the very questions at the start of the article.
“Questions that should be asked are how is the data being collected, is it good data and how is this data being interpreted?”
So, in fact it is not ‘data’ that needs to be trusted but those who collect, assemble and interpret it.
When I bought my first computer, in the early ’90-ies – I live in Romania – I came across the concept of GIGO.
It fits here perfectly.
It is very hard to reach good decisions based on ‘garbage’ data.