"The Presidential Election of 2016 will go down in history for many reasons - but one of the top will be the failure of the polls to see the eventual outcome. In fact, averaging the major polls projected Clinton's chances of winning at 90% the day of the election." Reuters 11/7/16

Pundits had Clinton's best-case showing at 355 to 182 Trump. Trump got 68% more than projected with Electoral 306 Trump vs 232 Clinton.

"The election is in 15 days. And the electoral map just keeps looking grimmer and grimmer for Donald Trump. We are making three changes to The Fix map this week, all favoring Hillary Clinton." Donald Trump’s chances of winning are approaching zero Washington Post 10/24/16

What happened?

Predictive Analysis & Polling

"If the election were today..." gives the impression that polls are predicting the outcome. But the election wasn't today - and they did not have a crystal ball. All data is historical - there is always an epistemological chasm between the past and the future. There is NEVER enough data to make a decision or to make an infallible prediction. Further, the polls were not just about observing sentiment, they were hoping to allocate resources to influence behavior.

"...while Obama’s analytics team discreetly competed by way of predictive analytics to win the election itself–as Hillary for America is now doing. This is a form of of quantitative prediction that transcends forecasting the outcome to actually exert an effect on it." Scientific American

Here's the catch, when you're looking in the rear-view mirror, your time horizon is short, and you only have one confirming event (the election), you can easily make your data models self-confirming. To avoid this, you must make micro-projections, hypotheses which you can test, confirm and/or disconfirm. And selling hats and gathering donations can fit the bill – if you set it up correctly. Working with the Blood Center in Kansas City, I saw that a disproportionate number of donations came from upscale women - it made sense but was too strong. I asked, "Where do you make your appeals?" “Upscale women's magazines... our board likes to see the ads or they don't think we're doing enough." The board was mainly women, they appealed to women, the profile showed that women gave so the conclusion was drawn that more should be done to attract their key market. In hindsight, Trump's numbers among traditional white male republican voters were lower than Romney. Trump's big gains were with Blacks & Hispanics. In his rally speeches - even in the whitest geographies - he consistently talked about his vision for helping the inner cities. Unlike the blood center (above) he realized that often there is a larger market where you've NOT marketed than where you have. Too often the traditional data models (profile our best people and find more like them) show you where you’ve been marketing – but fail to identify the much larger untapped opportunities.

Data Quality Polling Data vs Buying Data

The biggest MISS in the polling process is thinking, "people do what people say". Poll answers are NOT behavior (neither are 'positive' tweets, facebook 'likes’, information requests, free subscription signups, or focus group discussions). Polls ask people how they feel and what they expect to do. Trump's campaign used buying behavior instead. In Direct Marketing, we have ALWAYS known that what people say they like or will buy can be FAR DIFFERENT than what they BUY. Obviously there were many factors in this election but one consistent element was massively positive and consistent Clinton polls. Trump tweeted "Major story that the Dems are making up phony polls in order to suppress the the Trump . We are going to WIN!" Twitter 10/24/16. The question remains, why did Trump believe what turned out to be true?

Testing Your Conclusions

The echo-chamber of the pollsters was deafening. What did Trump look at?

From an AP interview in May, “In a break from recent major party nominees, Trump does not plan to invest heavily in a data-driven effort to court voters in the fall campaign. The businessman said he'll spend "limited" money on data operations to identify and track potential voters and to model various turnout scenarios that could give him the 270 Electoral College votes needed to win the presidency. He's moving away from the model Obama used successfully in his 2008 and 2012 wins, and which Clinton is trying to replicate, including hiring many of the staff that worked for Obama.” AP 05/11/16

The Clinton data juggernaut was repeatedly pointed to as her secret weapon. The Democratic party had a huge lead in data science applied to the political process.

“Big data firms claim they can generate as many as 1,500 data points on voters collected from public records, cookies that track web browsing, and voter registration databases. By putting all of this data together, they can create a personalized profile of every voter and target messages based on what each one cares about. This was especially important in Barack Obama’s 2012 reelection, where this data was used to model “persuadability” in order to find the voters most susceptible to Obama’s message. This year, Hillary Clinton’s campaign is reported to have spent almost $500,000 on Timshel, a startup data operation backed by Google co-founder Eric Schmidt. The company created a tool called the Groundwork that organizes the massive amounts of data collected from donor tracking systems, marketing analytics databases, and mass e-mail programs.” Trump vs Clinton: Analytics and the Presidential Race 06/13/16

Following on the BIGdata successes of Obama, it was almost certain that Trump was off track. Not only did he publicly minimize the value of BIGdata, it seemed that he was hiring ‘friends’ instead of true ‘data scientists’.

“...one part of the story I wanted to drill a bit deeper on - in part because it shows what a mess Trump's campaign remains but even more because it illustrates a very basic Trump MO: friendship and loyalty over experience and competence. The issue is who's doing Trump's digital work.” TalkingPointsMemo.com 8/22/16
"At first Kushner dabbled, engaging in what amounted to a beta test using Trump merchandise... The Trump campaign went from selling $8,000 worth of hats and other items a day to $80,000, generating revenue, expanding the number of human billboards–and proving a concept." Forbes Nov 22, 2016

Kushner pushed Trump’s empowering encouragement to a political unknown, Brad Parscale. He is a San Antonio marketing entrepreneur who taught himself code while working at a silicone-valley startup. He launched is own company but it was a casualty of the dot.com bubble. Returning to Texas, he went door-to-door through industrial parks, looking for clients. He frequented bookstores, aproaching people pulling web books off the shelf if they needed help. Trump International Realty asked for a website bid - he won at $10K (they were probably shocked it was so low). Before Trump decided to run, Brad built the pre-announcement website for $1,500. As the campaign became a reality, his authority grew.

“After a March event in Chicago devolved into a melee, Parscale decided to stop relying on the ticketing service Eventbrite and build his own tool to accept RSVPs. He says he coded the program himself in two days so eventgoers would have to confirm via mobile phone. The added layer would weed out fraudulent requests placing tickets in protesters’ hands—and also collect supporters’ phone numbers.” Bloomberg 10/27/16

I called Giles-Parscale and confirmed that ‘direct marketing’ was part of their ‘corporate DNA’. Their testing-driven system also became one of several key elements which allowed Trump's team to understand the demographics or his rally attendees and the intensity of their interest. Parscale began applying internet sales techniques, more precisely, he consistently employed the proven principles of direct marketing. Forged in the crucible of more than 100 years of direct mail practice, it was the capability of direct marketing practice to isolate causal variables in field valid experimentation which gave the Trump team better decision-making data than all the opinion polls combined.

"...in late June, Trump sent out his first e-mail solicitation, it ended up in recipients’ spam folders 60 percent of the time. Typically marketers in that situation would have begun quietly blasting less important messages from a new server to familiarize spam filters with the sender’s address. Parscale shrugs off the ensuing criticism from technologists. “Should I have set up an e-mail server a month earlier? Possibly,” he says. “We also raised $40 million in two weeks. Woo-hoo, spam rating.” Bloomberg 10/27/16
“...The Trump campaign has had a comical running story about spamming foreign parliamentarians with cash asks to take on "Crooked Hillary" and at least started with massive and totally unprecedented spam problems.” TalkingPointsMemo.com 8/22/16

The emails could have been done better, but they got done. And as true direct marketers, they tested more broadly in order to learn fast. Trump's disdain for political correctness did not just extend toward the left and the Press. His own team was charged with getting his message out fast... and in as real a manner as possible... even if it got a high spam rating. They were not just fund raising. They were also testing offers and watching geo-demographic data on the buyers and the non-buyers.

"...in August, Project Alamo’s data began shaping even more of Trump’s political and travel strategy—and especially his fundraising. Trump himself was an avid pupil. Parscale would sit with him on the plane to share the latest data on his mushrooming audience and the $230 million they’ve funneled into his campaign coffers." Bloomberg 10/27/16
Parscale said on FOX, “Jared Kushner is Ivanka Trump husband. Jared was instrumental in being kind of an overlay in kind of bringing Trump's genius down to the all the different parts of leadership. You know, Steve Bannon was there. It's kind of that strategy person. But the data operation right from the middle and I think for the first time in history, the data operation ran everything from TV buying to where we were on the ground to all of the different operations. And so, and having that data right there, we could start to where the persuadable targets are, GOTV (PH), everything we needed to know -- “ RealClearPolitics 11/16/16

The pundits were very certain that the wheels were coming off the Trump team’s wagon.

“...the Trump campaign is blowing a lot of money on poorly targeted advertising. It's also really damaging its reputation with the Icelandic, Finnish and Danish political communities. These are the kinds of things you would expect if the operation were being run by people who don't have experience doing digital work for political campaigns. And - surprise, surprise - it is being run by people who don't have experience doing digital work for political campaigns.” TalkingPointsMemo.com 8/22/16

You will never make money with data - nor (these days) will you get elected President, unless you make decisions with it. The Trump team understood that donations and purchases and rally attendance were far better than measured opinions. The Trump team started using donor data... and hat sales... to decide where to fly and where to place their advertising. As we have covered above, opinions multiplied by gigantic data sets and then shaped by estimating algorithms can be little more than garbage out. But when millions of people participate, drive hours or take out their wallet... that speaks volumes - if you set things up right... according to those time proven direct marketing principals.

"...statistical models isolated likely supporters whom Parscale bombarded with ads on Facebook, while the campaign bought up e-mail lists from the likes of Gingrich and Tea Party groups to prospect for others. Some of the ads linked directly to a payment page, others—with buttons marked “Stand with Trump” or “Support Trump”—to a sign-up page that asked for a name, address, and online contact information." Bloomberg 10/27/16

One of the challenges with the information which has come out about the Trump team's work, was that the principal authors of the Bloomberg article, Josh Green and Sasha Issenberg are not direct marketers. Josh is an American journalist who writes primarily on United States politics. Sasha Issenberg is the editorial director and chief strategist for VoteCastr. He is a contributor to Bloomberg Politics, a former columnist forSlate, and the author of The Victory Lab. Issenberg's blogs included titles like, "A Vast Left-Wing Competency - How Democrats became the party of effective campaigning—and why the GOP isn’t catching up anytime soon". Bloomberg asserted if there is anyone who knew about campaigning, it was Issenberg. Neither these authors, nor those of Forbes or others who commented on the original information, understood at the time, the superiority of both hard buyer behavior or the heuristic power of direct marketing testing. In many ways, the Bloomberg article suggested that while Trump didn't care much about data, or perhaps even winning... he may have stumbled into a way of selling stuff and raising money or even perhaps building a new TV network.

"... Parscale invited a variety of companies to set up shop in San Antonio to help determine which social media ads were most effective. Those companies test ad variations against one another—the campaign has ultimately generated 100,000 distinct pieces of creative content—and then roll out the strongest performers to broader audiences." "... those whose algorithms fared worst in drumming up donors lost their contracts." Bloomberg 10/27/16

The principles of Direct Marketing are a lot of work to implement... in many ways far far more difficult than placing ads on main stream media. Direct marketing is about setting up the advertising so that you not only generate customer behavior, but that behavior moves you ahead in your decision making. Direct marketing is applying the Scientific Method to the marketing process. Combining a variety of offers with list segmentation and measurement, the practitioner can isolate causal variables - answer questions, prove and disprove theories and thus conclusively understand how markets are connected with your offer. Gary Coby, director of advertising at the Republican National Committee, worked on Trump’s campaign.

"Coby’s team took full advantage of the ability to perform massive tests with its ads. On any given day, Coby says, the campaign was running 40,000 to 50,000 variants of its ads, testing how they performed in different formats, with subtitles and without, and static versus video, among other small differences. On the day of the third presidential debate in October, the team ran 175,000 variations. Coby calls this approach “A/B testing on steroids.” The more variations the team was able to produce, Coby says, the higher the likelihood that its ads would actually be served to Facebook users.
“Every ad network and platform wants to serve the ad that’s going to get the most engagement,” Coby says. “The more you’re testing, the more opportunity you have to find the best setup.” Wired 11/15/16

Translating this into the political arena, the Trump team was able to not only raise money, sell hats, signs, offer credit cards and influence voters, they were also able to learn how what was said and offered resonated with voters - independent of opinions and expressed intentions. Who would have thought spending more on selling hats could be more powerful than spending on polling information (I'm not going to call it 'data').

"He (Brad) developed rudimentary models, matching voters to their Facebook profiles and relying on that network’s “Lookalike Audiences” to expand his pool of targets. He ultimately placed $2 million in ads across several states, all from his laptop at home, then used the social network’s built-in “brand-lift” survey tool to gauge the effectiveness of his videos, which featured infographic-style explainers about his policy proposals or Trump speaking to the camera. “I always wonder why people in politics act like this stuff is so mystical,” Parscale says. “It’s the same shit we use in commercial, just has fancier names.”" Bloomberg 10/27/16

Because Brad came at it from a commercial perspective, he had little time for the slow-moving generalizations of mass advertising. He clearly understood that much more was to be gained by setting up a systematic process to test the tides of public opinion with concrete cash-generating offers. In my opinion, Brad had the heart of an old-fashioned junk mailer - a high compliment indeed.

"...while the RNC had spent years trying to identify as many generic Republican voters across the country as possible, Trump was attracting a different kind of person: sporadic or unreliable voters disillusioned with politics in general and with both political parties. That was the kind of voter that Cambridge had focused on identifying in the electorate and was the focus of the voter universe models it built." Yahoo 11/9/16

Since 2012 the RNC spent more than $100 million in creating their BIGdata voter analysis system. The Trump team ERP was self-funded by hat sales (etc.) and used it to leverage the RNC list system by adapting it to their testing and experiment setup. I'm sure much more will come out in the next few months but one thing to be certain of... buyer behavior beats BIGdata in the real world of generating action. I would go so far as to say that BIGdata failed Hillary in two huge ways: First, it substituted opinion and projection for actual behavior. Second, so much expended effort fueled the hubris - blinding the Clinton team from the actual world (rather than how they wished it to be). Are you piling data higher and deeper or are you planning every piece of communication, generating the data to learn what your market really desires? FREE Report: 5 Tricks to turn your BIGdata into a Decision-Making Powerhouse.