Distorted Data: How Were Polling Predictions So Far Off?

2016-12-22-1482435223-5947472-distorteddata.jpg

 Melody Gambino, Marketing Director, Grapeshot

It has been over a month since the conclusion of the Presidential election and the tremors from the Trump-quake are still being felt as D.C and the U.S political system are “shaken-up”. Whether you were pro or against Trump, there was a wave of shock experienced the night of November 8th. We all experienced this surprise in different ways of course, some celebrated with shouts of “woohoos” and others stung with a five finger slap to the face. There were many disagreements among this election but one thing we can all agree on is the media’s predictions were very, very wrong.

The extreme inaccuracy of the polling predictions served as a wake-up call to the market research industry. How could the polls get it so wrong and what are the lessons for us all in the wider marketing industry with a healthy data dependency?

First up, did the polls really get it so wrong? After all Clinton won the popular vote. It was the interpretation of the data rather than the numbers themselves which was off.

This is a sharp reminder for us all to question and challenge the quality of our data. What questions were asked of whom and when?

Furthermore, we must consider what sort of characteristics the sample size represent. So for example, how many people would have taken the time to answer your questions? Response rates are currently set to 5% but is this 5% representative of the general population. (How many people do you know would stop to answer detailed questionnaires?)

Then it's always a smart idea to check one set of data against another to see if there is a correlation.

Evaluation of social media engagements indicated that Trump would win the Presidential race earlier in the year running contradictory to the predictions of some polls. Yet, us marketing folk understand that social media can be a bellwether to predict changing mood and opinion.

However, relying on social media alone is folly too. Just because someone engages with a social post regarding a presidential candidate it does not necessarily mean they are going to vote for said candidate. With added granularity this can be a more viable way to measure/survey the country due to the prominence of social media users and access to data.

Nevertheless, these issues existed long before the big data era, in which we live, began. The biggest change is in our classical definition of thought leaders and influencers. While a rise in the application of quantitative methods like big data and marketing technology has dominated the collective minds of marketers, a more qualitative shift has also occurred, highlighting the changing trend in consumer trust that marketers must address: Influencers are now more powerful than traditional celebrities.

In this regard it’s important to know that we tend to ‘digitally hear’ what we want to hear. A lot of people didn’t want to deal with the reality that Donald Trump may win the election. News feeds on social media were skewed by Facebook algorithms. People consume their news via social media now as sources have multiplied, and we were fed stories that aligned with our likes – politics included. Additionally, the waves of unfollowing and defriending due to disagreeing viewpoints was unprecedented, furthering the polarized opinions that plague this divided nation.

Sounds crazy, right? However, it makes sense when you consider that modern consumers expect authentic brand relationships, and therefore value the input of influencers they trust over celebrities whom have a perceived monetary relationship to promote products and brands.

Last month, AdWeek wrote about the rise in user-generated content (UGC) as a powerful marketing asset in a piece entitled: “Move Over Kardashians: Why Average Joes Are More Influential Than Celebrities on Social Media,” the author thoughtfully argues that a well-executed influencer program can yield higher conversions at a much lower cost for brands than spending enormous money on traditional celebrity endorsements. For example, consider that to get Kim Kardashian or Kendall Jenner to post ONCE on a social platform can cost an excess of $200,000, which not only is cost prohibitive to many brands, but is hard to justify given the changing consumer sentiment around authenticity and transparency.

For advertisers, while there are certainly digital influencers that themselves have jumped into the celebrity category, the key to success is understanding the effect of influencers appears to be less about single initiatives and more about absorbing a brand’s total digital signature to create content on behalf of the brand.

We live in a time where every voice can be heard via social networks. For marketers this means, you should not only learn to shut out environmental noise but also know who your influencers are, which provides a unique window to the current market trends. It’s beholden on us all, marketers and politicians alike to carefully question the quality and correlate multiple data sources to have a better chance of predicting future behavior, avoiding shocks and surprises.