Still dressed in black from a memorial for their CEO Dave Goldberg, SurveyMonkey executives were confronted with a confounding piece of news. The results of their first big public push into political polling were in -- a long-awaited moment of their beloved CEO -- and they differed wildly from their competitors'.
The credibility of the company, which Goldberg, a high-profile Silicon Valley personality and husband of Facebook COO Sheryl Sandberg, had headed since 2009, was on the line. So was a bigger question: Could Silicon Valley crack the code to calling elections, an undertaking long dominated by major media and data outlets, and one that has become less and less reliable in recent years?
“We were dressed in black from the memorial, and we said `do we do this?' We do run a risk if we're wrong, and what if we are,” Bennett Porter, head of marketing, recalled. “But we knew that we weren't, and we knew that Dave would've wanted us to do it.”
SurveyMonkey's executives decided to go out on a limb. They gave their predictions to the Washington Post, which published them the day of the election, showing the UK Conservatives trouncing their Labour rivals. Other polls showed the race neck-and-neck. The bet paid off: Conservative Party Leader David Cameron was re-confirmed at the country's prime minister with a sizable majority.
The episode wasn't just a triumph for SurveyMonkey. It was an inflection point for the broader polling industry, which has traditionally relied on calling thousands of homes until they reach enough people to stand up a statistical analysis. SurveyMonkey's strength comes in about 3 million people using its online do-it-yourself questionnaires each day for purposes as mundane as scheduling dinner meetings. Digital technology makes it easy for the company to tack political polling questions onto the end of those surveys.
Meanwhile, technological advances have been causing problems for traditional pollsters. With the advent of cellphones and near-universal caller ID, traditional telephone surveys have become more expensive and less reliable. By the mid-1990s, the rate of people responding to telephone surveys had dropped to about 30 percent from more than 80 percent. According to a 2012 Pew study, the rate continued to fall to less than 10 percent.
That's pushed some of the established players away from classic horse-race polling. Gallup, which had Mitt Romney beating President Barack Obama in its final 2012 survey, last month announced it wouldn’t do its signature polls for the 2016 presidential primaries. Instead, Gallup editor-in-chief Frank Newport says the organization will focus on surveying public opinion on specific issues.
After long rejecting online polls as a cheaper alternative, pollsters are thinking again, prompted by SurveyMonkey's success.
“The field has had a significant change of attitude about this from one of ‘No way’ to ‘Well, let’s take a look,’” says Scott Keeter, director of survey research at Pew, who has been working with SurveyMonkey to better understand the differences with online polls.
SurveyMonkey's business is about getting people to pay for more sophisticated surveys, instead of just using the simpler free ones. The company also gathers data for paying clients. A technology accessory company, for example, might pay SurveyMonkey to get feedback on its latest iPhone case design from female iPhone owners.
“There are customers who are doing very sophisticated, very time-sensitive market research and other data-gathering projects with our platform, but it's not that widely known,” Tim Maly, SurveyMonkey's chief financial officer, said in an interview.
While the company doesn't disclose its financial results, it raised $250 million in December, which reportedly placed the company's value at about $2 billion.
“Polling at some point may be a business in and of itself, but we really view it more today to enhance and evolve the brand and open the door to selling our higher-end solutions,” Maly said. All of which “would be very helpful” if the company were to go public, Maly acknowledged, though he wouldn't say whether that's the plan.
The company's push into politics began in 2012, when executives noticed the demographics of their users mirrored that of the broader U.S. population. They quietly experimented with their own political polls and accurately predicted the winners of the 2012 presidential elections. At the time, they weren't quite sure why they were right, but it gave them hope that they could provide an alternative to traditional telephone polls that were failing.
For Goldberg, whose death last spring rocked Silicon Valley, solving that question was key. He wanted to expand SurveyMonkey's market research business. Key to that was establishing credibility for measuring public opinion. In Goldberg's view, there was one way to do that: calling elections.
So, in 2014, Goldberg hired Jon Cohen, a former Washington Post pollster, to figure it out. Cohen quickly dived into the 2014 mid-term elections but didn’t feel confident enough to widely broadcast his results in advance of the elections. “Dave wanted to put out the midterm data, but he let me have my approach for one time,” Cohen said. Goldberg was clear, though. The next time—the UK elections—would be different.
Rather than trying to cold-call respondents, SurveyMonkey chooses some people who are already filling out its surveys—such as one Cohen recently sent his family asking about their preferred Thanksgiving sides—to take part in political polls. The company includes questions about age, race, gender, and education level. Just as in telephone polls, Cohen’s team weights responses to reflect the general voting population.
Critics say SurveyMonkey's approach—and that of other online surveys—excludes people who either don't or infrequently use the Internet. “Not everyone is online,” said Cliff Zukin, a professor at Rutgers University who specializes in measuring public opinion in politics. “This is particularly problematic in voting studies, since it is the elderly who are the least likely to be online and the most likely to vote.”
Jonathan Mellon, a research fellow at the University of Oxford who is examining why most polls failed in the UK election, said it's too soon to tell whether this is the new frontier for polling.
“At the moment, we don't know if they're doing something different and it will consistently lead to the right results, or if it's doing something different and it will just systematically have different results than everyone else, which sometimes might be better and sometimes might be worse,” Mellon said.
SurveyMonkey's success in the UK has become a proof point for the company as it tries to work its way into the conversation around the 2016 presidential election. They partner with NBC News on polling. And this month, when the Los Angeles Times turned to the company for political polling, the newspaper cited SurveyMonkey's success in the UK elections.
“One can effectively deliver telephone surveys today, but to do it well is extraordinarily expensive therefore the approach has become a luxury good and the need for good polling is such that we can't afford to make it be a luxury item,” SurveyMonkey's Cohen said. “The reality is that people are responding to surveys. We need to meet people where they are.”
SurveyMonkey is already outgrowing its shimmering new corporate headquarters in Palo Alto, California. In October, the company announced it hired Mark Blumenthal, a Democratic pollster who created the popular pollster.com blog, now owned by the Huffington Post.
At a weekly meeting in early November, the Survey Monkey staff celebrated the poll conducted for the Los Angeles Times and, above the din of employees playing ping pong and billiards nearby, teased one of the team members who used the occasion to purchase his first actual newspaper.
As the meeting broke up, Cohen marveled at his surroundings—the game room, the smell of catered lunch wafting through the hallways—and how different it all was from his old days of conducting telephone polls. “I was at the Post the year they got rid of the free coffee,” he said.