Election campaigns recognize need for analytics in politics
Candidates are now using data to build profiles of potential voters and target their outreach in much the same way enterprises use BI tools to personalize marketing efforts.
With five days until the high-stakes midterm elections that will determine whether Democrats or Republicans control Congress, the use of analytics technology in politics is pervasive in campaigns across the ideological spectrum.
And just as the deployment of analytics can be the difference between success and failure in the corporate world, it can be the difference between victory and defeat in certain elections.
In 2020, incumbent Sen. Ed Markey battled Joe Kennedy III in the Massachusetts Democratic primary. Given the state's historical lean toward Democratic candidates, the September primary was a de facto election for one of the state's two seats in the Senate, though the general election against the Republican nominee was still two months away.
One year before the primary, a Suffolk University poll showed Markey nine points behind Kennedy.
Traditionally, Markey campaign workers would have called voters who historically vote Democratic and tried to convince them to elect their candidate. But that wasn't going to work, given Kennedy's popularity.
Instead, the campaign determined that it needed to find a hidden constituency to mobilize in order to overcome its deficit, according to Scott Ferson, president and CEO of Liberty Square Group, a PR and consulting firm, and a member of Markey's communications team in 2020.
Using analytics, the Markey campaign found that hidden constituency in environmental voters. The bloc traditionally does not show up at the polls but could potentially be mobilized, the campaign decided, given Markey's commitment to environmental causes.
"Environmental voters are really bad voters," Ferson said. "But in a tight race where we started out behind, given Markey's stance on climate change, we felt we should target them."
Analytics -- using data to develop profiles of voters and figure out how to target them -- enabled the Markey campaign to find those environmental voters and push them to the polls.
"The analytics told us that they may not have voted, but they donated to environmental causes and have showed up to events, and then we were able to target them," Ferson said.
In concert with targeting young voters -- another bloc that is passionate but doesn't historically show up on Election Day -- the Markey campaign's emphasis on environmental voters helped sway the election in his favor. Markey ultimately earned a decisive victory over Kennedy, winning by eight points, and went on to rout Republican nominee Kevin O'Connor in the general election.
More than just polling
While Markey's defeat of Kennedy is a recent example of how analytics can be used to help sway an election, the use of analytics in politics is nothing new.
Election campaigns have long used data. That data, however, largely came from polls, which are conducted infrequently throughout campaign cycles and can miss shifts in voter sentiment, and sources such as the U.S. Census and election boards that look at the past but reveal little about what is happening in real time.
Now, the analytics used in politics -- at least at the federal level in campaigns for the presidency and U.S. Senate and House of Representatives -- is just as sophisticated as the data analysis used to inform decision-making in the corporate world.
The intent of a campaign is to find out how to attract enough votes to ensure victory. And they do nearly everything within their means to do so.
That means using data to inform strategy.
Campaigns use analytics to allocate what can be scant resources depending on funding, optimize who to contact to raise funds, inform messaging and learn as much as possible about potential voters in their district to make decisions about how to get those potential voters to vote for their candidates.
"You're really trying to puzzle together" enough votes to get elected, said Michael Cohen, founder and CEO of Cohen Research Group and author of Modern Political Campaigns: How Professionalism, Technology, and Speed Have Revolutionized Elections. "That's the goal, more than anything else, and you're looking at a variety of factors to get there."
Anna CarmichaelTeam lead of solution architecture, Civis Analytics
Meanwhile, the analytics technology now used to manage data and inform decisions in politics is similarly as sophisticated as the technology the corporate world uses, according to Anna Carmichael, team lead of solution architecture at Civis Analytics, a Chicago-based vendor that builds analytics tools used across industries, including political campaigns.
While election campaigns at the local level usually lack the resources to develop data models and build data pipelines, well-funded campaigns have access to analytics stacks that can inform decisions from moment to moment, from the time a political campaign begins through its conclusion on Election Day.
"Analytics gives campaigns the power to make decisions, and to also evolve those decisions," Carmichael said. "You're able to … re-assess assumptions about what is happening. That brings more efficiency, and is more equitable by showing what is really going on with the American voters beyond 10 people in a room."
So much of what a political campaign can accomplish comes down to financial resources.
In a sense, political campaigns are akin to startup companies, Cohen said. And just like a startup, much of what political campaigns can accomplish depends on how much cash they have to spend:
- Getting their message out there and reaching their goal of victory on Election Day takes money.
- They need to hire staff, and the quality of the people they can hire often depends on how much cash they have to spend.
- At the local level, it may mean just hiring as many people as possible to knock on doors and canvass for votes.
And at the state and federal levels, it may mean delving into advanced analytics like predictive modeling and hiring consultants to develop data models that can drive decisions from how to raise more cash to what areas and voters might be most receptive to their message.
"The main decision a campaign has to make is allocation of resources," said Kimball Brace, president of Election Data Services, a research and consulting firm based in Manassas, Va. "Resources are always scarce, and the question is how to allocate them."
Similarly, Alex Carabelli, senior political partnerships manager at Civis, said decisions about how to allocate resources are critical right from the start of a campaign.
But the choices don't end there. As funds flow in over the course of a campaign -- or dry up -- decisions about how to optimize spending need to be made.
And analytics can play a role in maintaining that balance, Carabelli continued. Just as analytics enable corporate enterprises to find operational efficiencies, analytics in politics can be used to optimize every resource.
"There's a need for efficiencies in everything we do and every dollar that we spend," he said. "It's about figuring out how much it costs to get a vote -- from zero to a vote, how much does that cost? You need to look at the techniques to get there, and follow them and measure them and test them against one another over time."
Meanwhile, campaigns need to evolve the way they think about allocating resources, according to Ferson.
Historically, political marketing efforts have centered around television, radio and print advertising, which is expensive. Online ads are cheaper, he noted. But even less expensive is an investment in analytics tools that can identify who to target, as in the case of Markey's victory in 2020.
"Facebook ads are dirt cheap," Ferson said.
Regarding analytics, he added there are firms such as nonprofit Bluebonnet Data, which provides volunteer data scientists for progressive political campaigns.
Historically, polling data has been crucial for political campaigns.
Polls can inform candidates which issues are most important to their constituents, and whether candidates are ahead of their opponents or trailing were the election held at a given moment.
They can also reveal where candidates are doing well, where they are so far behind it's not even worth sending people to canvass for votes, and where there may be large numbers of voters who might be swayed.
But polls have their drawbacks.
They're snapshots of a moment in time before Election Day, rather than an accurate portrayal of what will happen on Election Day -- the closer to the election they're conducted, the more predictive they can be. They're also expensive, so polling is not constant and therefore doesn't always reflect what's happening within an electorate.
Census data, party affiliation and voter histories are also informative, and have historically been used by campaigns to make data-informed decisions based on demographic data and voter patterns.
But analytics in politics has evolved. Now, campaigns collect reams of data about potential voters based on their online activity, and just as online activity enables marketers to understand more about individual customers than in the past, it also gives election campaigns greater insight into the public as individuals.
"The big distinction, starting in 2012, is being able to campaign at the personal level," Carmichael said. "You can understand demographics about people, decisions they're making, and incorporate data from a lot of different sources about those people."
Campaigns can see what topics are bubbling up on Google Trends or Twitter in a particular area. They can also monitor how many clicks their ads are getting, and who is clicking on them.
And at a more personal level, they can track email and text exchanges they're having with constituents to understand more about those people, glean trends among voters and view the social media activity of potential voters.
Even door-to-door information gathering is more advanced than it was just a few election cycles ago. Now, canvassers are equipped with tablets, and the information campaign workers collect can be uploaded immediately into a database, as opposed to a couple of decades ago when all that information was written with pen and paper and perhaps never put into a database.
"There are a lot of different inputs now," Cohen said. "It used to just be polling, but now you can bring in a whole bunch of other data."
The ultimate goal is to essentially create as complete a view as possible of each potential voter and use that "Voter 360" concept to personalize a candidate's message to a potential voter, according to Andrew Tavani, executive vice president of data at Aristotle, a political consulting and software development firm based in Washington, DC.
Campaigns with enough capital to invest heavily in analytics put each data point about an individual into a model to build as complete a picture of that person as possible. The model -- based on machine learning -- then informs the campaign how best to target that potential voter.
Or whether to target them at all. Given the limited resources of a campaign, if someone is unlikely to vote for a particular candidate or donate to a campaign, it may be just as prudent to ignore that person and target others more likely to be swayed.
"We've been building voter files for a long time, and we have a lot of demographic information -- we have 1,000 data points," Tavani said. "I know what magazines people subscribe to, whether they're evangelicals, if they're an environmentalist, if they're giving to a campaign, their age, ethnicity, voter turnout. I know if they play golf, tennis or pickleball."
While the use of analytics in politics closely mirrors the use of analytics in the corporate world in the sense that data is used to inform decisions about resource allocation and how to sway potential voters, the corporate and campaign worlds diverge in the way they deploy analytics tools.
Campaigns exist for only a limited amount of time. Once an election is over, so is the campaign.
"It's different than any other business cycle in that it has a hard end where you either win or you lose, and that's it," Carabelli said.
Therefore, it doesn't make sense for an election campaign to use a traditional analytics vendor like Qlik, SAS or Tableau, whose pricing structures are geared toward long-term use. Nor does it make sense for them to build a data warehouse or store data in the cloud with a vendor such as AWS, Google or Microsoft, whose platforms are also designed for long-term use.
And while large enterprises tend to have data teams including data scientists, data engineers and data analysts to model and analyze their data, campaigns don't have the resources to keep a team of data experts on salary.
Instead, those that can afford it tend to hire firms like Civis and Aristotle that offer consulting services and possess both data science expertise and analytics platforms built for political campaigns.
"Some campaigns might have someone out of college who can help do some data analysis, but most are going to a consultant or a consulting firm," Tavani said.
The exception is the presidential level, at which campaigns have data scientists on staff to do their data modeling, he added.
The analytics pipelines developed by consulting firms, meanwhile, actually mirror those of corporate enterprises.
They're every bit as sophisticated as those deployed by a Fortune 500 company, integrating with other technology platforms to access data, ingesting data from those various sources, managing large data sets and feeding that data into models that lead to decisions and actions.
"The political users have often been the ones pushing our product forward the most," Carmichael said. "They tend to have an extremely innovative mindset and are always trying to do things that have never been done before."
Despite the quality of the technology, analytics in politics faces some of the same challenges as analytics in the enterprise.
Large companies tend to have the capital to invest heavily in analytics, while small and midsized companies sometimes don't have the funds to deploy an expensive analytics platform. Similarly, not only are local campaigns likely unable to afford investing in analytics capabilities but even many campaigns for the U.S. House of Representatives lack enough money, according to Carabelli.
"The biggest challenge when you're getting down to a state senate race, or even sometimes a congressional race, is the scale," he said. "Some of the analytics tools haven't evolved to a point that they can be built cheaply enough that these campaigns can afford them."
But that doesn't mean smaller campaigns have no access to advanced analytics, Carabelli continued.
Democratic and Republican party leadership can help by giving smaller campaigns access to analytics applications that can inform decisions.
"They can pool resources," Carabelli said. "Some of those tools are trickling down. What they don't have are customized models that can test their candidate against an opponent."
Meanwhile, buy-in remains an issue.
Despite the growing pervasiveness of analytics in politics, not all election campaigns place a high value on analytics, or pay close enough attention to what their data is telling them.
In 2016, Hillary Clinton lost to Donald Trump in part because her campaign made poor decisions based on data it was receiving as Election Day neared, according to Cohen, who works for Republican campaigns.
He noted that canvassers were reporting back to campaign headquarters that Clinton's messaging was not resonating with voters in what became swing states.
"They kept telling [campaign headquarters in] Brooklyn that they had the data and the messaging wasn't working, and the campaign said, 'Nope, we feel really good about our communications, and we're going to stick with it,'" Cohen said. "That was a fundamental mistake. They had the right tools but didn't use them well."
And often, it takes losing an election to learn a lesson, he continued. The Biden campaign in 2020 learned from the Clinton campaign's mistake, but some candidates -- just as some enterprises -- resist the use of analytics to drive decisions.
"Over the past 10 years, you see a lot more data science as part of the team, but it's not necessarily driving the team," Cohen said. "There's a lag between the data science that we can do at this point and the people that run campaigns and are willing to accept it into their strategic thinking."
Future of analytics in politics
It's been a bit more than a decade since political campaigns used analytics beyond polls and publically available census and local election board data.
According to Carabelli, the use of analytics in politics had its proof of concept with the election of Barack Obama in 2008 and became more widespread by the time of his re-election in 2012. Now, 10 years later, while not valued by every campaign, analytics continues to grow with campaigns below the federal level.
Meanwhile, advancing technologies like natural language processing and no-code process automation that reduce the need to hire data experts, plus the financial assistance of party leadership, are helping analytics to trickle down from upper-level campaigns and grow.
However, in order to truly advance the use of analytics in politics and make data accessible to a broader array of campaigns, there needs to be more standardized data input and collection at the local level, according to Carmichael.
Much of the data available at the local level is input by town and county clerks and secretaries of state. A lack of standardization when inputting that data makes curating usable data sets complex and time-consuming.
"The biggest data issues are around data cleanliness," Carmichael said.
Meanwhile, Ferson noted that as time passes, the gap between so-called "election people" and "analytics people" will narrow. And as that happens, analytics will become a more pervasive part of decision-making during political campaigns.
As recently as eight years ago, when working to elect Massachusetts Democrat Seth Moulton to the House of Representatives, Ferson said analytics was disconnected from decision-making. By 2020 on the Markey campaign, analytics was a determining factor, and by the time another few election cycles have passed, Ferson said he expects there to be convergence between those responsible for messaging and those responsible for analysis.
"There are people like me who love politics but wouldn't know a data point to save my life," he said. "Then there are analytics people who can deconstruct something in numerous ways but have no idea how to connect that to messaging. The next growth area for politics is for people to be political analytics people who can marry the art and data."
Similarly, Cohen said analytics in politics will grow in the coming years as a result of what is taking place in education.
As data science and analytics become a greater part of the curriculum at colleges and universities, the use of analytics in politics will grow, according to Cohen, an adjunct professor at Johns Hopkins University in addition to his consulting work. Data science and analytics are not yet commonly incorporated into political science, but as it starts to be, a transition will occur as a new generation of candidates and campaign leaders arises.
In addition, like Carabelli, Cohen noted that the use of analytics will continue to trickle down from the top.
"What you'll find, probably over the next 10 years, is that analytics will become a standard part of every campaign at the federal level," he said. "You'll find it also at the state level for governor's races. And then, you'll know it's been fully adopted when you see it in the state house races and races for the state legislature."