<img alt="dcsimg" id="dcsimg" width="1" height="1" src="//www.qsstats.com/dcs8krshw00000cpvecvkz0uc_4g4q/njs.gif?dcsuri=/index.php/c/a/Projects-Processes/Harris-Interactive-Reading-the-Internets-Mind/1&amp;WT.js=No&amp;WT.tv=10.4.1&amp;dcssip=www.baselinemag.com&amp;WT.qs_dlk=XtQygtW6xL7z25Y3CvIBAwAAABI&amp;">

Survey Says

By Mel Duvall Print this article Print

Their market research system mines the collective thoughts of netizens. But can the Rochester, NY company accurately interpret its polls to represent the world at large?

Survey Says

Operating in the world of market research, Harris Interactive helps companies determine what products consumers buy, or are likely to buy, their physical health and medications they may be taking, and opinions on a host of subjects. While the company earns the majority of its money by performing surveys on behalf of corporations, it's best known for its weekly Harris Poll, a public opinion survey that has taken the pulse of Americans since 1965. Topics range from whether the Masters Golf Tournament should be moved from Augusta National because of its refusal to admit female members (45% say yes), to who is the nation's most popular TV personality (Drew Carey).

Founded in 1975 as the Gordon S. Black Corp., it took its name from the current CEO. The firm was renamed Harris Interactive following the 1996 acquisition of Louis Harris & Associates, then a subsidiary of the Gannett Corp. It was a relatively small player in the market-research business with annual revenue of about $20 million. Most business came from firms in upstate New York, including many in the pharmaceutical industry. Until the early '80s, polling was generally conducted by mail or face-to-face interviews in people's homes, but improvements in telecommunications and call-center technologies shook up the market research industry. Using call centers, it was possible to vastly increase the sample size of surveys, and the geographical and socioeconomic spread, while at the same time lowering costs and the time involved in producing a survey.

With the rise of the Internet in the mid-1990s, Black recognized that another technological and economic revolution in the market research business was on the horizon. "The Internet offered another way to do it better, faster and cheaper," he says. "Replacement technologies don't come along very often, and when they do, you can't afford to be sitting on the sidelines."

On the surface, using the Internet to conduct research sounds like a slam-dunk. Once the technology is in place, there is very little cost difference in e-mailing a survey to 1,000 or 1 million people. Costs rise significantly with traditional polling methods as the number of respondents increases because of the involvement of human interviewers. The Internet also offered much more versatility in conducting surveys. They could be answered at any time of the day, include pictures, graphics or advertising copy, and could be modified on the fly according to a respondent's answers. One of the biggest benefits was in the area of productivity. Telephone surveys typically take six weeks from design to completion. In contrast, Internet surveys can be completed in two to seven days.

Others weren't so quick to buy into the idea. The chief concern among pollsters and corporations requiring research was that the Internet did not represent the public at large—particularly in 1997 when only 40 million Americans (about 16% of the population) were online. In addition, those online were highly skewed toward white, high-income men. Many of Harris' clients also required very specific panel groups. Drug companies, for example, wanted panelists suffering from a particular ailment or disease, or in a high-risk age group, and the industry was skeptical that Internet polling would deliver the quality of results required.

"There was a battle raging in the industry over this," says Bob Lederer, an industry analyst and editor of Research Business Report, a newsletter focusing on the sector. "There were people saying, 'Why are you even wasting your time on this? It's useless.'" John Fair, vice president of customer insights for Telephia, a San Francisco company that conducts research on behalf of telecommunication clients such as Verizon and AT&T, recalls bracing for meetings with clients. "It was a long education process," he says. "Companies were legitimately concerned about whether [the Internet surveys] would adequately represent their clients." A Harris client, Telephia now does most of its research using the Internet.

Black doesn't dispute the fact that the Internet in 1997, and to some extent even today, does not deliver a balanced version of the American public. But he is also adamant that modern polling methods, which apply algorithms or propensity weightings to accommodate deficiencies in a sample size, could overcome that challenge. "It's not necessary that your panel be proportionate to the population—the industry has been applying algorithms to deal with that for years. We were confident you could apply the same kind of science to the Internet," Black says.

Overcoming industry bias was one of three major challenges the company faced. The other two were just as imposing: building a technology infrastructure robust enough to simultaneously deliver and process thousands of surveys, and developing a system to amass and maintain an Internet panel of more than 2 million members. For this, Black turned to Leonard Bayer, the company's chief scientist and leading technology executive since 1978. Bayer formerly taught mathematical statistics at the University of Rochester School of Medicine and has degrees in mathematics, statistics and astrophysics.

Black called Bayer into his office and told him what he wanted. His only question was, could it be done? Bayer went off for a few weeks and researched the problem, including conducting a statistical analysis of an election survey the Internet provider Prodigy had conducted in 1996, and came back with his answer. "I said it can work, but it's going to cost a lot of money," Bayer recalls. Initial estimates were in the range of $20 million to $30 million—an enormous amount for a company with $20 million in annual revenue.

The private company's board approved the proposal. Initial development funds were raised from private investors, including directors, with an eye toward eventually taking the company public. In the era of the Internet IPO, the investors had their eye on a big payoff.

Bayer then recalls having a meeting with his small engineering staff and telling them something that almost seems unbelievable in today's climate. "I called everyone together and said, 'I'm going to tell you something now that you're never going to hear again in your careers. Time is more important than money. Until I say otherwise, the sky is the limit—time is our enemy.'" Time was critical because Harris was not alone in pursuing an Internet research strategy. Rivals, most notably Greenfield Online, NFO WorldGroup and the NPD Group, were either already doing research using the Internet or getting in the game. From an engineering staff of four or five people, Harris grew its developer staff to as many as 45. It included hiring a new CIO, Peter Milla, from Response Analysis Corp.

This article was originally published on 2003-02-01
Contributing Editor
Mel Duvall is a veteran business and technology journalist, having written for a variety of daily newspapers and magazines for 17 years. Most recently he was the Business Commerce Editor for Interactive Week, and previously served as a senior business writer for The Financial Post.

eWeek eWeek

Have the latest technology news and resources emailed to you everyday.