Lessons From Yahoo in Database Marketing

Granted, it was an out-and-out sales pitch for Yahoo. But Usama Fayyad, chief data officer for the online giant, awed database professionals with the statistics he presented this week.

He said, for example, that Yahoo collects over 10 terabytes of clickstream behavioral data each day, the equivalent of all the information within the Library of Congress.

“There’s no database on planet that can handle that 10 terabyte load,” he boasted.

And it was clear that he thinks Yahoo leads the pack with its DATA OS, or data operating system.

“Most data warehouses are useless because they were designed to support reports meant for humans to consume,” he told the audience of CRM and database execs at the NCDM conference in Chicago. “Most practitioners do not understand the needs of data mining algorhithms.”

That’s big talk, but Fayyad does have the numbers on his side. Yahoo is the top Web destination in the U.S., snaring visitors for 282 minutes apiece per month. It serves up 38 billion pages views a month and has an average daily reach of 44 million people, “a Super Bowl a day,” Fayette said.

And that provides the firm with thousands bits of data from which much can be inferred.

For example, Yahoo can tell who is shopping for a car simply by tracking who visited Yahoo Autos or who searched online for car dealers, buying guides or manufacturers, and target those consumers for messages.

“Seventy percent of the people we identify have intent to buy in three months, and 24% buy within a month,” Fayyad said. “That accounts for roughly 250,000 leads a month, 30% of the total U.S. car market.”

And the firm can predict through data mining and modeling what its online visitors will buy offline, he said.

That data mining ability also helps Yahoo solve its own problems (yes, it has them). For example, it noted a few years ago that it had much higher search usage on Netscape than on Internet Explorer.

The company ran tests and determined that the problem was the placement of the search box on the Yahoo home page. Netscape’s appears in center of the screen, Internet Explorer’s on the side.

Likewise, Yahoo found, after testing color and other variables, that Instant Message usage increases with the number of friends on the Instant Message Friends List, five being the “tipping point.” So it started the Just Add Friends campaign to get consumers to the magic number five.

“Imagine tons of patterns like these lurking in the data,” he says. The challenge is to “fish them out and make them come to the surface.”

Some marketers might respond that they can perform similar feats. But Fayyad had a larger point —that advertisers are missing a massive opportunity online, of which Yahoo is only a part.

Americans now spend 15% of their weekly media time on the Internet, but only 6% of all advertising dollars are being spent on it, he noted.

Search alone delivers 150 million leads per day (and Yahoo has a new strength in that market thanks to its purchase of Overture). And why is it so potent? “Because you have a glimpse of intent at an instance in time,” Fayyad said. “Clickthrough is 100 times more likely.” (He implied, though, that intent can also determined from behavioral analysis, and that ads served on that basis may be undervalued).

Companies are equally dim about online metrics, he continued.

“Total traffic not a good performance measure,” he said. “High-traffic referral sites often produce poorer quality clickthroughs, and the best response is not always the most effective.”

One site might generate four times the traffic of another site—and the same level of conversions. According to Fayyad, site two is the one you should invest in, although “the response would lead you to a completely conclusion.”

And conversions are a more effective measurement than clickthroughs.

“Firms know how to drive traffic, but they don’t think about what happens on that landing page,” Fayyad said. “That’s why we typically end up measuring only the clickthroughs. Some sites are really good, some are bad. Although you have expressed desire, they drive you away.”

In the end, it’s all about linking the brand impact of ads with classical direct response.

To illustrate this, Fayyad described a funnel, with brand at the top and direct marketing at the bottom. The top represents awareness and the bottom purchase.

What happens in the vast area between awareness and purchase? Consideration.

And which tactics do marketers use to reach someone at that stage? “My answer, after research: Essentially, none,” Fayyad said. And yet consideration takes up the majority of the time. “So there’s a huge business at the top, and a huge business at the bottom, and nothing in between,” he added. It’s by influencing this period that the narrow drip of customers at the bottom of the can be widened.

Say one thing for Dr. Fayyad: I didn’t see a single database marketer walk out.

To comment on this opinion, please e-mail me at [email protected]