SAN FRANCISCO -- As the chief scientist for Inktomi (INKT) , Eric Brewer spends a lot of time tracking technology. And in the quicksilver Internet industry, where missing a trend can trash a company's business prospects, that job's no stroll in the park.
"My job is to see what's coming multiple years out," explains Brewer, who co-founded Inktomi with a former grad student of his, Paul Gauthier, who is now the company's chief technology officer. "That's what happened with
Brewer last year realized that San Mateo, Calif.-based Inktomi, which licenses search-engine and data-networking technologies, needed to bust a move in the fast-growing electronic-commerce space. So he pressed his case and built consensus within the company's management. Propelled by Brewer's vision, Inktomi last September acquired
, a developer of online comparison-shopping software.
The purchase marked a new strategic direction for the company. Today the shopping engine that grew out of that acquisition is in beta and is already in a preview release at
. "I don't always win, either," admits the 32-year-old computer-science professor. "There are things that I've wanted to do that we haven't done."
Brewer, who has a doctorate in parallel computing from the
Massachusetts Institute of Technology
, says he spends most of his time teaching computer science at the
University of California, Berkeley
. This allows him to keep close to his roots, since it was at UC Berkeley in 1994 that he co-developed the search-engine technology that later morphed into one of the Valley's most successful IPOs. But, he says, teaching also enables him to keep close tabs on the most innovative research, companies and ideas.
Right now, the most important technology for Inktomi is its traffic server, which is designed to ease the bandwidth bottleneck. Based on the premise that it's cheaper to store information than to move it, the traffic server stores, or "caches," copies of frequently accessed information nearby the user. Even though the market for network caching is extremely competitive, Inktomi managed to earn nearly $8 million from the traffic server in 1998 -- 39% of the year's revenue, up from 1% of revenue in 1997. (That's more than a 13,000% annual gain if you're wondering.) And traffic server revenue is up for the first quarter in 1999.
Over coffee and Coca-Cola in the lobby lounge of the
in Santa Clara, Calif., Brewer gave
the dope on forthcoming developments in e-commerce, search technology and the importance of broadband to the company's future.
Do you have any plans to roll out new services on Inktomi's traffic server?
The original value proposition was that the traffic server reduces user response time and saves bandwidth and operating costs. The key insight now is that the server is a general-purpose piece of scalable software that's touching all of the user base. And that's unique in the industry because most things that touch the bits are hardware-centric like routers. And they can't actually do much to the things that they touch. All they can do is route them. Down the road, we expect to offer lots of interesting value-added services on the server. For example, if you want to do audio and video streaming, the cache becomes the splitter point for live streams, and it can also do streaming right out of the cache.
As broadband rolls out over the next few years, won't that decrease the demand for caching services?
Ironically, it doesn't. It's actually the complete opposite. We have extreme uptake of caching in the high-speed last-mile sector. Basically, any high-speed last mile has to have a caching solution, and pretty much all the major players do.
They're only solving the bandwidth problem for the last mile. We can stream out high-quality video from the cache to the end user in a way that's not possible without some type of clever software at the other end of the pipe. By the way,
design from the very beginning was to have extensive caching in the architecture. And it's still one of their value-adds. In the long run, there's going to be a lot of broadband, and that'll be good for Inktomi because we can do a lot more with that bandwidth than anybody else can. We might even show off a bit.
How important is your new shopping engine to the company's strategy?
In terms of financial issues, this year it will not be that important. We expect it to have a fairly lengthy growth process into a fully featured version. So we'll have some deals and make some money, but this is an investment year. In two years, we'll start to see some financial payback. And it looks like it be could very large. If you get the traffic through it, you're getting a fraction of goods revenue. We're quite optimistic that it can be a substantial part of Inktomi's revenue.
Why should companies go with your shopping engine over others like
First, the core Inktomi values. It's a private-label solution that does not conflict with your brand. You control how it reaches your customers, which is not true for Jango or
brand. We also have a long, quite elegant track record of delivering high-quality data services. There's a trust factor that we have throughout the industry that we can do technically complicated things with low risk. Plus, there's the specifics of shopping. We have a very large number of merchants -- about 300 right now--and that number is increasing quite regularly with some very large names. For the merchants, they like us because they can reach a lot of portals with one relationship, rather than going through the somewhat painful process of establishing a transactional relationship with each merchant.
During our tour, you mentioned that you're going to be rolling out some new shopping services this year. What might those be?
Talk to Kevin
Brown, Inktomi's longtime director of marketing.
Today, there's a product finder on the sites using our commerce engine. You can do a search for a black leather belt, and you can order it by merchant, by price or product. So there are different ways to find and sort. We crawl out to all the different sites -- we've got over 1 million products today -- and bring it back into our database. There are other pieces of the shopping experience that we're building on top of that. For example, feature comparisons. The ability to select a given class of CD players and say, 'I must have the Dolby this and that, I must have four times oversampling.' Another one is that we have an exclusive relationship with
in which we can pull in and map all that data onto product searches from the Web. Those add-ons will be rolling out in spring and summer.
You said you made a 5% to 7% cut off each transaction?
A couple of revenue streams come out of shopping. If someone buys something, there's a finder's fee. The typical cost of customer acquisition is 30% to 40%. So if we can deliver them a qualified prospect direct to their site, 7% to 10% or even more -- it varies by product category -- it is very profitable. Another revenue stream is ads and merchandising. We could also do a per-search hybrid.
What is the thread that connects all of Inktomi's disparate businesses?
There are certain problems that are technically very difficult. To have something up 24 hours a day, to handle 10 million users a day, to handle a billion Web pages a day for
. A billion.
serves a billion -- ever. This is a billion a day. So when it comes to that kind of scale and that kind of availability and that kind of terabytes of data, there isn't anybody else that can do that like we can.
Search engines have been criticized for being dumb. Is Inktomi working on any new near-term projects that will help improve search technology?
I won't get into any specifics yet, but I will say that relevance is the single biggest area of investment in search right now. In the last year, we have gotten all the scale and availability and presence issues under control, and we've turned the guns on relevance. It's already having a big impact internally, and it's going to make a difference.
So do you anticipate significant developments in relevance this year?
Yeah. We started this quite a while ago. Some of them are so obviously better that we'll do announcements on them. Others are so algorithmic and they make a difference, but they're hard to quantify and hard to explain.
Do you guys want to stay independent?
There's no obvious partner to consolidate with. Plenty of people can buy either the portal half of our business or the network infrastructure half. There's nobody that wants both halves. Inktomi's Switzerland-like neutral status is a big part of our value proposition. The buyer would have to have a similar status.
We also have a fiduciary obligation to take any compelling offer seriously. I don't believe we've been approached post-IPO. There are many rumors about us being acquired and none of them have any substance.
Personally, I think we have a very large agenda, and we have more information and a better strategy and a better feel for how to execute it than anyone that could acquire us. But, again, I'll consider all comers. Maybe they've seen something we haven't seen.