What’s Next for Predictive Analytics After the Data Fails of 2016?

We know how poorly the future has been predicted via traditional approaches.  David Cameron’s return in 2015 as Prime Minister of Great Britain was considered to be a surprise, as he easily picked up 328 House of Common Seats.  Brexit was considered to be a no-brainer that it would be turned down as an option, yet 51.9% of voters voted yes.  Donald Trump was dismissed by many as a candidate, yet he picked up 306 votes or 36 more than he needed.

The good news is that what happens in politics often represents the tip of the spear for innovation. So we ask “why” and before you know it, we innovate.  Here is what our panel explored today.

I focused on five key drivers of change:

#1 — Subconscious behavior is more important to measure in highly emotional/partisan issues.  We won’t tell the truth if you ask us in a highly emotional setting, but our actions will tell the truth.

#2 — The “non-behavior” e.g. silence, apathy or a decrease in intensity is often more important than what we say.  If an important constituency starts to decrease its intensity or perhaps go silent, this may be far more telling than what we are reading or what people are saying.

#3 — Narrowcasting is leading to overinterpretation of what real trends actually are.  We are increasingly getting our information from the sources that are most comfortable for us

#4 — Highly partisan and/or even fake news has a cumulative impact even if we think it does not.  Advertising models taught us long ago that frequency matters.

#5 — A new set of peers are emerging as influencers (the interpreters).  As the 9% in the 1,9,90 model matures into a media force, what they do and say is often far more powerful than any set of media outlets.

Rebecca Haller, who leads audience insight for Politico, informed us that Politico just created a new department dedicated to understanding our audience two weeks ago.  This team is is focusing on what they can learn from their readers, subscribers and event goers, who are also their sources and advertisers.

Rebecca also said that “we are combining the best of first and third party insights to understand our audience’s lives outside of the Politico ecosystem.  We are looking at more ethnograpic research and combining the best of pyschographics with our basic knowledge of our audience, all to provide a better experience”.

This is real innovation at a major media outlet and is one to pay attention to in the months and years ahead.

Mark Stouse, founder of Proof said that “analytics is hard enough….predictive is fraught with peril”. He went on to describe seven key learnings:

We don’t first understand the past and present
We know what we want and that drives bias
We trust ourselves when we should not
We assume consistency v. inconsistency
We don’t understand the role of time
We like pretty pictures too much
We like large speculation v. small certainty

Dr. Alexander Krasnikov, assistant professor of marketing for Loyola University in Chicago focused on the value of brands and made several interesting points, such as:

We need to conduct continuous segmentation in real time. Continuous being the key word.

If we do this well, we start to uncover the customer’s hidden needs and preferences.  We see early warning signs.  And with time, we can start to become predictive of responses likely to occur in specific scenarios.

Just as important, finding “alike” consumers does not imply correct segmentation

We are entering a time where our ability to innovate in data science and behavioral models has never been more important.

I’ll conclude with the key message overall.  Major change leads to breakthroughs.  Yes, its often fun, even therapeutic to discuss what happened, but it is much more productive to evolve and change how we do business as a result of what we are learning.

Here is one example of what we are doing to get a better view of what is actually happening in the market place.

We have realized that we will now build multi-dimensional algorithms so you can get the full and real perspective for any market, avoid false positives and see how trends or movements or apathy is really occurring.

Here is an example of how we are approaching it.

We are building a new “Trump algorithm” that has six dimensions.  The first is the “brand”, in this case Trump and all of his followers.  Second, we look at his appointees and surrogates (the army). Third, we look at Congress and staffers.  Fourth, a wide range of normative data sets (the real secret sauce) ranging from normative sets of 1MM people are more per channel who represent the “average” to NGOs for a specific issue to all African American pastors who discuss politics in public to key journalists and more.  The fifth is time and motion related.  What is the duration for successful momentum and when do you know that a new idea or protest is taking hold for real? And the sixth relates to sensitization and desensitization to a topic.  We often forget to look at the rates of burnout for things we are passionate about or fail to see an ember turning into a fire early enough.

The result is a new way to look at how an audience is truly being built, shaped or redefined.  Not surprisingly, it is important to point out that a single group often does not automatically impact the audiences that matter.  They might…..they might not….and that goes for any one group.  Said another way, just because any group is vocal on a topic doesn’t mean that will ever correlate with success. You still have to win the hearts and minds of the right people.  In that respect, nothing has changed….but our ability to understand the psychology of the market via technology and how it is shaping our world is becoming a top priority for brands, companies and anyone in the world of politics.

Thank you to our leaders on today’s panel. Best, Bob

Bob Pearson
Bob Pearson
Vice Chairman & Chief Innovation Officer