Colleges Rehab Their Web Sites for Major Payoffs

Colleges spend dearly to maintain vast, ever-expanding Web sites. They tweet. They blog. They podcast.

But most colleges have no idea just how much bad Web design can cost. Kafkaesque online forms and pages that nobody visits, for instance, can have disastrous effects: A quarter of prospective students decide not to apply to a college because of a bad experience on the college’s Web site.

That loss (documented in a survey of 1,000 high-school seniors conducted last year by Noel-Levitz, an enrollment consulting firm) can add up to a lot of money. “Generally, higher education hasn’t ever had to think about that before,” says Shelby Thayer, a Web strategist at Pennsylvania State University’s main campus. “How much does bad design cost us, and how much does good design save us?” As colleges do more crucial business online, “that’s kind of my burning question.”

For answers, a number of institutions, including Penn State, are now turning to Web analytics. Going far beyond superficial measures like counting visitors or hits on their Web sites, they track who their visitors are, what they are looking for, why they fail to find it, and—a crucial measure to gauge advertising spending—how much a successful Web visit is worth.

Many of the techniques, such as closely monitoring prospects, are standard practice on e-commerce Web sites and among for-profit colleges, but they are just gaining a foothold in most of higher education.

The Chronicle talked to officials at several colleges that have set up sophisticated analytics operations in admissions, audience tracking, and public relations. They warned that data can be misused, and collecting them can be hard because responsibility for college Web sites is often spread among departments. Plus, many goals in higher education—such as improving reputation—are not easily measured.

But the officials also said analyzing their Web data to drive online decisions brings enormous rewards. “We spend a huge amount of time in higher ed maintaining content that has little return on investment,” says Michael Vedders, director of Web services at Bethel University in Minnesota. Analytics has helped Bethel spend money in the right places, he says.

Funneling Admissions

Mr. Vedders is blunt about Bethel’s old Web site: It looked horrible. But more important, the site for the liberal-arts college, in St. Paul, made it difficult for prospective students to find information that would encourage them to apply.

Many private universities spend upward of $2,000 to recruit each student who enrolls, and their Web sites often form prospective students’ first impressions. The critical path leading from prospect to applicant to paying student is known as the “admissions funnel,” and Mr. Vedders’s goal is to optimize it.

An analysis of Bethel’s Web data, drawn from Google Analytics, showed Mr. Vedders that the college’s funnel had some problem areas. Prospective students tended to drop off right before starting an application, at the point when they were required to create a new account. The main admissions page, which was supposed to route applicants to different Web pages based on the program they were applying for, had a high “bounce” rate, meaning that people left without selecting any option at all. For those visitors, the “next step” wasn’t clear.

“We need to get our 17-year-olds to the undergraduate school and our 65-year-olds interested in the seminary to the seminary Web pages,” Mr. Vedders says.

When Bethel redesigned its Web site last year, it clarified where different prospects should go and added a new explanation about the steps required to submit an application. Since the redesign, visitors spend half as much time moving from the main admissions page and are much less likely to leave the page without selecting any option, efficiencies that will help Bethel get more applicants.

Other colleges use search data to make sure prospects’ questions are being answered. During the financial crisis in 2008, visitors to the Web site of Hamilton College started searching about financial aid more frequently than they did about admissions. J.D. Ross, director of new media, asked Hamilton’s dean of admissions to write a letter to parents explaining why the college was affordable, and he put a link to the letter at the top of the main admissions page.

Effectively using analytics is about “being familiar with the data that you have, and knowing what normal looks like, knowing what abnormal looks like,” Mr. Ross says. “Then you can see if there’s a problem.”

Wayne State University uses an even more rigorous approach to optimize the very start of its admissions process, on the college’s home page. Nick DeNardis, associate director of Web communications, tested the effectiveness of that page’s “apply” button. Using a strategy known as A/B testing, Mr. DeNardis tested which version of the button led visitors to click through to Wayne State’s admissions pages. Changes based on the testing led to a 62-percent increase in traffic to the main admissions pages this year over the same period last year.

But Mr. DeNardis says he is hamstrung by a common limitation of college-analytics programs: Bureaucracy. He works on the main pages on the Web site, but the department that manages the application itself has not installed analytics software, precluding him from tracking how many of those students who hit the “apply” button actually follow through.

“It’d be awesome if we could see the entire funnel from prospect to applicant throughout their student life cycle, down to the actual donors,” Mr. DeNardis says, “but we do as much as we can.”

The 5-Headed Monster

College Web sites must do much more than get prospective students to apply; they must speak to many audiences at once. John Drevs, manager of Web services at Loyola University Chicago, calls the quintet of main university audiences—students, prospective students, faculty, staff, and alumni—the “five-headed monster.” It is, he says, difficult to create general Web pages that have the right information for five different groups.

“All universities are challenged with what to put on the home page. Right now, our home page is a shotgun approach” that tries to do a lot of different things at once, he says.

So Loyola Chicago has stopped trying so hard. Instead, the university is building software to identify in advance which audience group a visitor falls into. Mr. Drevs estimates that the university’s future Web site will be able to identify the audience group of 60 to 70 percent of visitors by making intelligent guesses. For instance, a visitor logging in from a campus computer is likely to be a faculty or staff member (or possibly a student). A visitor who signs up to attend an open house will be considered a prospective student, and the Web site will remember the distinction if the prospect returns using the same computer.

Automatically tailoring content to visitors is common on e-commerce sites like Amazon.com, but it is rare in higher education. Avinash Kaushik, a leading expert on Web analytics at Google, praises Loyola at Chicago’s effort and says colleges could reap big rewards from adopting its strategy. Delivering a good Web experience, he says, depends on finding out what visitors want and then giving it to them.

Mr. Kaushik says learning more about Web visitors will help colleges determine one of the most important metrics, called a task-completion rate: How many visitors were able to complete the tasks they came for? “You let the people who use the site tell you: ‘I came here for this. This is how much you stink. Now fix it,’” Mr. Kaushik says.

Loyola University Chicago’s first foray into custom content is a home page, called Inside Loyola, designed for internal audiences like faculty and staff members. On a recent Thursday, the college’s main home page asked admitted freshmen to make their deposits and advertised a campus tour. But the Inside Loyola home page introduced its new men’s basketball coach and noted an on-campus colloquium on Jesuit music.

The internal edition initially worried some employees, who felt like they were missing something on the “real” front page, Mr. Drevs says. But another metric revealed that Inside Loyola was broadly liked, he says: Most visitors came back.

One persistent limitation of Web analytics at most colleges is that the most important transactions—say, campus visits by applicants or big donations by alumni—take place offline. Web data can track an applicant’s online movements, but how can they measure what might ultimately drive an applicant or donor’s decision, such as a college’s reputation?

In short, they can’t. Some colleges use services like Klout to judge their influence on social media like Facebook and Twitter. Each college is given an influence score that reflects how many people follow and share the college’s messages. And one company, the Global Language Monitor, scours the Internet to rank colleges based on which is mentioned the most. But those measures are grossly inadequate for determining whom a college’s messages reach and whether those messages persuade anyone to donate or enroll.

Getting Good Press

But some colleges are finding that less grandiose measures of social-media usage, when combined with other Web data, can at least help them promote themselves. James Madison University gives each of its news releases a “shareability index,” a score that combines how many people read it, how long they stay on the page, and how much the release is shared on social media.

The index, generated by software called Vocus, has helped James Madison rethink what kind of news releases it puts out, says Andrew Perrine, associate vice president for communications and marketing. “We need to provide content that actually influences our audience,” Mr. Perrine says.

Colleges naturally like to promote news of their successes: a high ranking for a chemistry program, or a program that helps high-school students. But those stories tend to get little attention. “It’s always about: The world is beautiful and it’s because of us, and there’s no reason to read that,” he says.

By contrast, releases with the elements of a good news story—a little drama, a person fighting for what is right, a villain—have scores four to five times higher than those about the success of a program, he says. As a result, James Madison has started to tell more human-oriented stories.

For instance, a writer recently turned in a story about nursing students doing community service in rural areas. “I was bored within the first paragraph,” Mr. Perrine says. But he edited the piece to highlight a buried fact that roughly half of the homeless people in the surrounding county were children. “Bingo, there’s your lead,” he says. “And then our nurses are the ones trying to solve the issue.”

Mr. Perrine admits that good storytelling predates having the data to track what works. But he says that in a university with competing priorities and faculty who want their research promoted, being able to show what is effective allows him to make the case for a consistent message. “Then I can defend my decision,” he says. “It’s not just me trying to trade on my good looks.”