Chris Wilson, CEO of WPA Intelligence spoke with TechRepublic's Dan Patterson about why the traditional advertising model isn't working anymore. Here's their conversation:
Patterson: We met while you were the director of analytics on the Ted Cruz for President campaign, and we are talking about the digital transformation of politics. Not really politics, though, we're kind of talking about campaigns and tactics, and what businesses can learn from watching how these hyper-focused campaigns operate in a kind of, binary win or lose scenario, through this election year.
So, Chris, thanks again for your time today. I wonder if we could kind of stop or start with the real high level digital transformation, and how political campaigns have evolved, over the last, even since the days of direct mail, which really was technology in the '70s, '80s and '90s, to, now we have automation machine learning, predictive analytics, and artificial intelligence.
Wilson: Right. Well, gosh, to back to, and that's really the beginning of the keeping of lists that, today, drive the revolution we live in as direct mail. It's probably the only area of the things that we talk about, that Republicans took the lead in, back in the early '80s, under Ronald Reagan, and people like the Jesse Helms machine, down in North Carolina, where they started assembling these huge lists, for the purposes of just going out. It was kind of like the version, the 1980s version of spam, where they'd just send a letter to anybody who'd ever responded to anything, that made those people think that they were conservative. I said, "think," because that was like, people learning, instead of machine learning. So, now that we've moved from the days of those lists that may have had a few different opinions on, or because they subscribed to Reader's Digest, that there was an assumption that they might learn conservative. Or they got the Paul Harvey newsletter, to really go old school here, that they might be conservative.
They add them on to a mailing for Ronald Reagan, or National Conservative, NCPAC, were some of the leaders in this. Or I already said, the Jesse Helms' conservative caucus, down in North Carolina, and the ability to begin to assemble those lists. But you take those lists, that may have had two or three names on them, and June and I discussed, when we got together in Cleveland, the evolution of the files that allows us to utilize machine learning, for the purposes of crunching data, and moving to the modern day, and the implications of 2018.
It has grown exponentially, to where, even as late as 2010, a well-appended voter file would have, maybe 200 pieces of first-, second- and third-party data on it. And then, in '14, it grew to about 400, and in '16, it grew to about 800, and then, 2016, whenever I played the role, that you've mentioned for Ted Cruz, as director of analytics, we had almost 5,000 pieces of first-, second- and third-party data on most of our voter file.
And I'm saying, mostly, it'd be a Republican file. We didn't make it out of the primaries, sadly. I'm still going through counseling for that, but I'll get over it soon. Moving into 2018, when some important decisions have been made by Google, that allows us to match against the voter file, by several other different entities that have allowed us to grow our files by the proliferation of quiz apps. It's like, everybody is going to spend time answering trivia, and doing quizzes, and all that.
SEE: How data analytics has forever changed political campaigns and elections (TechRepublic)
Or, not all that, but a lot of that comes back, and the ability of us to utilize all of those data that are now, in some cases. I mean, we have some files that are exceeding tens of thousands, and even in states where we would do a lot of work, we build models on top of models, hundreds of thousands of pieces of underlying data, and utilizing machine learning, for the purposes of finding the trends, and looking for correlations, and no longer caring about causations that allow us to really move into the next level, is, it's kind of exciting stuff.
I was actually having a conversation with another journalist this morning about this. I mean, the ability to make predictions in 2018 is good for a framing campaign, even down to the state rep level. This used to be only done for the presidential level, and now, we do it for state rep campaigns. On March 6, is the Republican primary, or the primaries in Texas. I work for a lot of state representative candidates who are really utilizing levels of machine learning and predictive analytics that were not used by the Romney campaign in 2012, just because of Moore's Law, and massive amounts of data have ended.
Patterson: Yeah, it is, when you zoom out a little bit, it is easy to kind of follow, campaign-to-campaign, and see the iterative advances make huge steps. I wonder if we could open up a voter file for a second. I mean, you are not exaggerating, when you talk about tens of thousands of pieces of data that's kind of appended onto an individual.
Even when I use a tool like L2, which lets me see quite a bit of information, what can campaigns learn about a voter, and how do they convert that information into action? I'm sure the viewer can kind of understand, well, if you run a business, and advertise your own marketing, you'd kind of want the same types of data, and to convert your audience into some sort of action.
Wilson: It's difficult for me to think outside of goals, because my clients here me to reach a specific goal.
Wilson: In 2016, it was to get Ted Cruz elected. We won the Republican primary. We did okay, we came in second, but we didn't win. So, I guess I failed at that, but then, when we go into 2018, my clients hire me to either get elected, or get re-elected. And so, my goal in 2018, and I'll talk about this from a corporate angle, separately.
My goal in 2018, on the Republican side, is to find those voters that turned out in 2010, and 2014, and 2016, that voted for Donald Trump. There's a lot of new voters that cast ballots in states like Ohio, and Pennsylvania, and Iowa, and Wisconsin, and Florida, for Donald Trump, that had voted for Barack Obama, or not voted at all in past presidential elections.
So that is my goal, to find those. To give you an example of how minute our attention is to that, as you do know, I do a lot of work in Texas. Still work for Ted Cruz, work for Greg Abbott. I can tell you, in the state of Texas, there are 2,068,746 voters that do not currently plan to vote in 2018. That, if they do vote, will vote for Ted Cruz, Greg Abbott, Will Hurd, John Culberson, Pete Session.
Those are three seats that Hillary Clinton won, that Republicans that are incumbents have to get re-elected in. That's the level of minute detail that we get down to. Now, the file, and the underlying data behind that, is what allows us to make those predictions. The difference between making those predictions in 2018, and this is what's important to my clients, I think, in 2016, is the accuracy and the granularity of the prediction.
Now, there's far more that goes into it. How do we motivate them? What do we say to them? What is it that we know about them? How do we reach them? One of the things that we just can't, it would be criminal not to talk about, on a podcast, broadcast like yours, is to talk about the evolution of how people receive a information.
By 2020, by the next presidential election, 54% of Americans will be what are known as cord cutters. That means they receive their television from Sling TV, DIRECTV NOW, PlayStation, Hulu, things like YouTube TV.
Then you've got 37% that subscribe to traditional, and streaming, and then, 9% that have never, that will have never been a cord, who have never had cable or satellite at all. So you got cord cutters, you got cord nevers, and you have those who are kind of making the evolution. You add those up together, that doesn't leave a lot of people that you could only reach through traditional advertising. I think the traditional advertising model is breaking, and anybody who does not adapt to that, be they political or corporate, is going to find themselves left behind, and we spend a lot of money on reaching not a lot of people.
SEE: 60 ways to get the most value from your big data initiatives (free PDF) (TechRepublic)
Patterson: Yeah, and some of these technologies, like OTT, are coupled with targeting tools. One is built by Zach Moffitt, and others who have a pedigree in political technology, and have kind of bridged the gap into the private sector. Chris, I wonder if you could leave us with some insights into the technology, the actual tech tools that are being used on this year's midterm election, and what sorts of technologies we can expect, as voters make decisions, and as they pull the lever in November?
Wilson: Well, I will tell you what will work in all. Because, unfortunately, all of my friends at BlueLab, and Synthesis Analytics, which are the versions of us on the left, don't invite me into their office to look at their new tools. But I know they're probably doing the same, if not more, over there, because they tend to have a little bit larger budgets than we do, believe it or not.
So, some of the things that we're working on is, our goal is to be able to be, sort of a Netflix or Amazon type of environment for Republican campaigns. What I mean by that is, in the same way that Amazon can predict whenever you're about to run out of shampoo, or Netflix predicts what you want to watch next on television, they get pretty good at it.
That's what we're trying to do, is building in artificial intelligence into our database, which is what we call, we've named Archimedes: "Give me a lever, and I can move the world." We're building, and we have, building APIs, with all the right of center of tech programs. You mentioned Zach Moffitt.
At Targeted Victory, Michael Beach, his former partner at Cross Screen, who has another type of utility, there's deep-root targeting, all of those that we can build and then, gestate it into, and book, send data back out, and be able to make predictions about voters on a real time basis. Because, and with the same way that product needs or television watching desires change quickly, so do what positions, what issues, are going to motivate you to turn out and vote, and vote for a specific candidate.
So that is our goal, to be able to build those same sort of predictive analytics, those same sort of predictive technologies, the same sort of artificial learning technologies in the world we have, so that we're able to work back with campaigns, ingest all of their data very quickly, to be able to turn around and give back to them a love of prediction down to the single, inverted voter level, that will allow them to make decisions about who they're talking to, and what they're saying to that voter.
I feel like, if we can do that, then we will have gone a long way toward bridging the gap, between where we are today, and where we need to be, going into 2020.
- Google: Russian groups did use our ads and YouTube to influence 2016 elections (ZDNet)
- Facebook: Cambridge Analytica took a lot more data than first thought (ZDNet)
- Facebook, Twitter aim to dodge regulators by regulating themselves (CNET)
- Google: Russian groups did use our ads and YouTube to influence 2016 elections
- Facebook, Google, Twitter execs to testify at Russia hearings (ZDNet)
- Cambridge Analytica's Facebook game in politics was just the beginning, the enterprise was next (TechRepublic)
- Did Russia's election hacking break international law? Even the experts aren't sure (ZDNet)
- Trump backs down from 'impenetrable cyber unit' with Russia (ZDNet)
- Bitcoin laundering suspect caught in US, Russia extradition spat (ZDNet)
Dan Patterson has nothing to disclose. He does not hold investments in the technology companies he covers.
Dan is a Senior Writer for TechRepublic. He covers cybersecurity and the intersection of technology, politics and government.