I thought I’d share thoughts about election night that are relevant not just for politics but for the overall business of insights and strategy. Three failures (or successes, depending on your perspective!) were laid bare in the 2016 election results:

1) Predictive polling failed, big league.

For those who rely on polling to provide certainty for uncertain events, polling failed to deliver. Pollsters across the aisle envisioned a Clinton victory. They were wrong, and this is a serious problem. We’re going to read lots of different autopsies about this: were people lying to pollsters? Were pollsters “herding” their data toward the result that other pollsters had found? It’s all hard to know. But we do know this: we need to ascribe more uncertainty to polling than we have in the past, and the “gold standard” may no longer exist for outright prediction.

Client polling is NOT primarily about prediction – it’s about building strategy and messaging, and understanding how different groups react when faced with key choices and information from different perspectives. But in the end, polling needs to pass the test of accurately reflecting a broader population’s attitudes and beliefs. For that reason, it failed.

2) “Big data” and predictive modeling failed – but innovative “little data” has great potential to help improve that work.

Every “big data” model had Clinton winning with a high probability. Even the most pessimistic modeled outlook had her winning odds above 75%, and (more important) winning the popular vote by 3 or more points.

Big data analyses are based on pre-existing information about individual voters and assumptions based on past behavior and incoming data. The data sets are getting larger, which creates illusions of certainty based on standard statistical parameters. But if incoming data and the assumptions are flawed, probabilities will be off – sometimes dramatically so. Computing power is cheap, but human intelligence and insight remain absolutely essential and are critical inputs that need to be more influential.

I believe that “little data” has been squeezed out too readily, and should be taken more seriously as part of all big data efforts. Our most insightful work for Bloomberg Politics over the past 2 years was not state polling, or any of our slice polls. It was a focus group we did in July, 2015 among Trump voters. It was the first public research to show that Trump supporters weren’t just a fringe of the GOP – many were Romney voters who were sick of the status quo and had great admiration for Trump, his message, and his tone. More than any poll at the time, it provided texture, nuance, and a real grounding for what was to come.

We’ve gotten similar insights in “little data” work for a number of our clients. Focus groups and in-depth interviews are the standard tools, and we have implemented new tools to tap into this stream of information. These new methods can provide ongoing, continuous feedback without pigeonholing people into simple response categories. All opinion researchers need to remember that little data – combined with human intelligence and big data analytics – are essential and complementary tools for unlocking real insights.

3) The Beltway elite failed.

The “insiders” of both parties did not predict this outcome, nor do they fully understand the sentiment behind it. Democrats are paying the price today, but there are many uncomfortable Republicans as well.

The sad fact is that we all talk about bipartisanship and “coming together,” but there has been only one issue on which leaders of both parties have joined hands for the last 25 years: free trade agreements. And that is precisely the issue that – more than anything – energized Trump’s base and electrified Sanders’ supporters. This is not a call to reject free trade (far from it!) but rather to reflect on how the parties may want to find additional ways to help solve the real problems that are affecting voters across the political spectrum.

At Purple, we are dedicated to developing insights for our clients in a smart way that transcend traditional partisan and brand conflict. Echo chambers and opinion “bubbles” are dangerous for candidates, elected officials, and companies. The most effective strategies are built on a diverse, innovative, and thoughtful approach to understanding the audiences you care about.

In that way, this is (of course) not the “death” of polling, but rather a reminder of the importance of constant innovation and evolution in gathering and interpreting insights.

about purple

Jean Walton

What *really* happened in the 2016 election

Purple Strategies@PurpleStrats

Alec Jacobs

Scotch Plains, NJ

Three failures and one HUGE upset

Purple Strategies@PurpleStrats

What really happened in the 2016 election