(FORTUNE Magazine) – Kim Polese has never had a problem drawing a crowd. When she spoke last December to the Churchill Club, a nonprofit tech business group, some 350 people packed a room at the Santa Clara Marriott to hear her thoughts on "The Silicon Valley Phenomenon." For 90 minutes they munched on chicken or beef tacos and listened raptly as the co-founder and CEO of Marimba touched on thought-provoking, topical matters like Internet stock mania and the dawning of political consciousness in Silicon Valley. The speech was measured and articulate and covered wide ground, but Polese made scant reference to her three-year-old company.

Which inevitably got me to wondering how many of the 350 knew anything about Marimba, once one of the hottest startups on the planet. As Polese left the podium to hearty applause, I buttonholed a few people to find out. The results were not good. Of the four people I queried, one hadn't the foggiest. One recalled that Marimba used to be a "push" company. "They do something with Java," another guessed. And the fourth answered correctly that Marimba is doing "application distribution and management," though he admitted that he knew this only from reading the program, and that he didn't really know what it meant.

It was a highly unscientific sampling, of course, but as I later asked more people about Marimba, confusion emerged as a common theme. To set the record straight, here's what the company is and what it does: Marimba is a 156-person business in Mountain View, Calif., that sold $17 million of software last year; its software helps systems managers deliver, over a network, regularly updated applications to computing devices like your office PC. That's an important function--and promises to be increasingly so as more and more software is distributed and maintained via networks. Either people don't know that or they don't care--because, the thinking goes, whatever Marimba is doing can't possibly live up to the company's early hype. "Big hat. No cattle," mutters one venture capitalist.

Yet there isn't a soul in Silicon Valley who doesn't know who Kim Polese is. A 37-year-old Berkeley native who was a product manager at Sun Microsystems before co-founding Marimba with three Sun colleagues, Polese (she says it rhymes with "Scorsese"--as in Martin) is one of Silicon Valley's highest-profile entrepreneurs. When she was named one of Time magazine's 25 most influential Americans in 1997, in the company of folks like Madeleine Albright and George Soros, Marimba had fewer than 30 employees and barely any revenues. At other points in the past three years, Polese has been toasted as a "superstar of the Internet age," "the diva of push," and "the Madonna of Silicon Valley," a geek sex symbol who's more famous than her company.

That is something Jacqueline Ross, Marimba's new vice president of marketing, and Lutchansky Communications, Marimba's third PR firm in three years, say they are determined to change. "This is going to be 'The Marimba Show' from now on and not 'The Kim Polese Show,'" Robin Lutchansky tells me after Polese's speech, which presumably then is one of the last showings of "The Kim Polese Show." Such a shift is particularly pressing, since there's an IPO on the horizon: Marimba has drafted a prospectus announcing its intention to make a run at the public markets, and as FORTUNE went to press it was scheduled to land on the SEC's doorstep any day. And though the folks in Silicon Valley have taken a lot of strange stuff to the public markets in recent years, no one has yet figured out how to take a person public.

Here's the problem. If Polese's talk at the Marriott had been titled "Marimba's Application and Distribution Management Solution for the Corporate Enterprise," it's unlikely that even 20 people would have showed up. Marimba's product is no hot portal, no online auction site; it's not even a piece of software the average Joe can pop into a computer and use. It does have a jazzy name, Castanet; but Castanet is a technology tool. IT managers at large companies buy it, and they use it behind the scenes in network servers. By many accounts, the software works well. But to be honest, it's pretty boring stuff. Technology analysts call it "plumbing."

Now, there's plumbing--Microsoft Windows is plumbing, as are Cisco Systems products--and then there's plumbing. Marimba's product is in the latter category: ask analysts today about the upcoming IPO, and they talk about how its software has a place in what's currently a niche market. That's not language that's going to propel a wildfire IPO. That's not the kind of language Kim Polese hoped to be associated with. Her goals for the company have always been a bit grander, as anyone will tell you who remembers Marimba's run as the hot Silicon Valley startup.

Polese resists the idea that she orchestrated a campaign to make Marimba a glamour company. "People think that there was some grand PR scheme, and that we're the ultimate overhyped marketing company, but the irony is that it was all incoming. We didn't go out and seek it," she says. Nor did she, Polese says, set out to become a cyber-celebrity. But that's exactly what happened. Marimba was associated with two of the biggest waves of hype ever to hit Silicon Valley--the Java programming language and push technology--and now Marimba is trying to jump on at least two others, but we'll get to that later. In the past two years, Polese--who gets credit for Sun's early marketing success with Java--has done all sorts of things that put her squarely in the limelight. She has spoken at about 50 conferences, posed with a group of women for Anne Klein fashion ads, and sat for hundreds of press interviews, including, of course, this one. And here's the thing: If you were an entrepreneur in Silicon Valley, where hype is crucial for drawing public awareness to technology breakthroughs and essential currency for building new companies; if your company were one of the first funded by the high-profile Java fund of the preeminent venture-capital firm Kleiner Perkins Caufield & Byers; if you truly believed your product could change the world; and if, on top of it all, you were easy on the eyes in a world of pasty-faced white men, you too would have done anything you could to get the word out--you might, in fact, have done the very same things that Kim Polese did.

That's the way business seems to work in Silicon Valley. It's just that in Marimba's case, all the hype became too much of a good thing.

Despite Polese's insistence that she never sought the limelight, the first steps she took after starting Marimba in January 1996 didn't exactly discourage attention. She and her co-founders--engineers Jonathan Payne, Sami Shaio, and Arthur van Hoff--left the Sun team that developed Java just as Java was beginning to gain national attention. So everyone--journalists, programmers, and venture capitalists, that is--wanted to know what they would do next.

To make the most of that initial attention, Polese hired Silicon Valley PR firm Niehaus Ryan Wong to, as she puts it, "manage the interest." She and Bill Ryan, a Niehaus Ryan principal, sat down in Marimba's cramped, one-room Palo Alto office and talked about ways the company could sustain its buzz until it was ready to launch a product. The trick would be to do this without divulging any details of what the company was up to, partly for competitive secrecy and partly because the four founders really weren't sure what they were going to do.

The solution Ryan and Polese agreed upon is exactly the opposite of what Lutchansky and Ross are now trying to do--they decided to shift attention away from the software and focus instead on the founders. "The idea was to use the notoriety of Kim and the other founders--to put the attention on them as people, not on what the company was doing. We as an agency have always pushed the idea of the CEO as an important part of the company's brand," says Ryan. "We believe there's a lot to gain by making the CEO famous."

The attention on Marimba's three male engineer founders lasted only long enough for Wired magazine, then the arbiter of all things technologically hip, to run a full-page photo of the founding foursome in its April 1996 issue. After that the spotlight shifted to Marimba's female CEO, and it never left.

In May the Silicon Valley business magazine Red Herring trailed Polese around the JavaOne conference in San Francisco. Having agreed not to reveal anything about Marimba's business plan or product, reporter Alex Gove instead wrote adoringly about Polese. "A lifelong dancer who chose her company's name because it suggested movement and life, Ms. Polese slips gracefully through the crowds of nerdy male programmers who flock around her." The accompanying photo revealed an ebullient Polese, hands frozen above her head in a Latin dance move. Later Polese would try in vain to get the press to stop writing about her dancing hobby, but at the time it was great to get such big play in a well-read Valley monthly. Front-page stories like the one that ran in a June 1996 issue of Information Week, where a cover photo of Polese heralded an article titled THE JAVA IMPRESARIO IS TANTALIZING NET USERS WITH HER MYSTERIOUS NEW COMPANY, were just what Polese and Ryan wanted. "There wasn't a journalist in the Valley who wouldn't have sold his mother's soul to be able to break this story," recalls Ryan, proudly.

That's hyperbole, of course, but when Polese finally delivered details of Marimba's product in early October, everyone from Network World to Newsweek wrote about it glowingly. At a launch party at the San Francisco restaurant 42 Degrees, 200 people were invited, but 500 showed up. Everybody who was anybody was there. Soon after the party, FORTUNE wrote that "the buzz around Marimba is deafening. Marimba is creating a new standard for the way information is distributed on the Internet."

Despite all that buzz, Castanet was still a nascent technology in search of a market. It was designed to transport software from any server on a network to any computer attached to that network. This seemed a service that might apply to many industries. Marimba's first business plan, written in June 1996 and used to secure $4 million in venture funding from Kleiner Perkins, laid out no fewer than seven categories of potential customers--publishing, entertainment, retail, finance, enterprise, customer support (now called e-commerce), and directory services (better known today as portals).

In meetings before the launch, several analysts had advised Polese that the so-called enterprise market--FORTUNE 500 companies, government departments, and the like--was where Marimba would find its most lucrative customers. Stan Dolberg and Eric Brown at Forrester Research and Hugh Delany, then at the Gartner Group, figured that Castanet would appeal strongly to big companies. They could use it to update all that software sitting on employees' desktop PCs, and also to distribute software to suppliers and customers who needed access to their corporate network. Those are just the kinds of things companies do with Castanet today. As a veteran of Sun, whose only customers were enterprises, Polese agreed that Castanet could find a place in this market. But back then, the enterprise market wasn't sexy. "We knew we weren't going to have anyone banging down our doors," says Rick Bruner, the former account manager at Niehaus Ryan, "if we went out talking about solving the 'software updating' problem." But there was something sexy out there. It was called "push."

No one seems able to pinpoint exactly when it happened, but by the fall of 1996, media, entertainment, and technology companies were eyeing a seemingly revolutionary model for delivering digital stuff over the Internet. Instead of having to go to a Web page and pull down files, users could subscribe to services that would automatically send, or "push," information to them. Pioneered by a now humbled company called PointCast, push was breathlessly declared in a Wired cover story in March 1997 to be "the radical future of media beyond the Web." When Polese and her team were developing Castanet in 1996, they used TV metaphors to describe how it worked--Java-based "channels" created by Marimba's customers would broadcast content such as news and interactive games, or applications like a financial portfolio or an expense report, to a "tuner" residing on desktops of people who subscribed to those channels. Though Polese says she didn't formulate this model to fit into push, everyone immediately recognized Castanet as a push technology. Once that happened, the phones at Marimba rang off the hook with people wanting to set up meetings to figure out how their company could do push. There were executives from Japanese consumer electronics companies flying in. There were meetings with IBM and Oracle, and with Netscape and Macromedia. One morning a vanload of Swedish executives from telephone company Telia showed up practically unannounced. Marimba engineer Carl Haynes recalls an endless stream of people in fancy suits wending their way through Marimba's offices to meet with Polese. "One time I hung out with Lou Dobbs," says Haynes, remembering his encounter with CNNfn's president and host of CNN's MoneyLine.

Dobbs was there to negotiate a deal. Many reporters dropped by for a more plebeian reason: Kim Polese gave them the rare opportunity to attach the photo of a young, attractive woman to a relatively complex story about high tech. When Red Herring did a follow-up story on Marimba in December 1996, it ran it on the cover with a glamorous close-up of Polese, her face softly lit and airbrushed, eyes beaming up at the camera. "Kim is a glamour figure in the industry," explains Red Herring editor Jason Pontin. "She's a young woman in an industry dominated by middle-aged white men, and she was pioneering this incredible technology." What seemed good for Marimba was great for Red Herring: The Polese issue was a smash, selling a higher percentage of newsstand copies than any other issue.

Michael Krantz was the Time writer who lobbied for Polese's inclusion in the magazine's March 1997 list of the 25 most influential Americans. Says Krantz: "I thought the ideas of streaming software that she put her chips down on were really cool, and I still do. But does the fact that she's a very attractive woman help raise her profile, as opposed to being another dweeby guy who doesn't comb his hair? Absolutely."

At first Polese's celebrity seemed to help Marimba. At the Web Design & Development Conference in San Francisco in February 1997, Marimba was able to announce no fewer than 50 "strategic partnerships." Hordes of journalists crowded into a room at the Moscone Center to hear Polese explain how customers like CNNfn, Columbia TriStar Interactive, and a troika of Intel, Macromedia, and PBS Online were going to develop channels of exciting new content. "This is the next wave on the Internet, allowing everybody to get past the frustration of clicking and waiting to download something," she told the standing-room-only crowd. There were going to be so many Castanet push channels that you would need a directory to keep them organized--and Excite had signed up to provide that.

But then something happened: Push fell apart. Corporate IT managers noticed that their networks choked on all the traffic that push technology, particularly PointCast, was shoving around. They quickly redirected development dollars away from push and toward making their Websites better. And that's when all the hype began to bite back. In spite of the fact that Marimba's technology was more elegant than PointCast's, many of its deals, which weren't really hard and fast to begin with, evaporated.

The PBS/Intel/Macromedia project for an online store of PBS videos, for instance, never moved beyond what Cindy Johanson, head of PBS Online, now refers to as a "prototype demonstration product." MGM Interactive wound up dumping Castanet for another technology for its streaming video project. Says Ken Locker, then head of MGM Interactive: "They [Marimba] thought it [Castanet] was going to reinvent the Internet and overcome the issue of bandwidth. It looked great on paper, but it was just a little ahead of its time." One company that supplies software to Aetna tested Castanet, but ultimately the insurer built its health-insurance enrollment service into its Website, rather than "pushing" forms and other software to employees' desktops. A deal with Netscape tanked. And Excite's channel directory? Yup, you guessed it: kaput before it really got started.

In fact the only aspect of push mania that lingers today is Marimba's association with the ill-fated trend. Co-founder van Hoff says he still gets asked about push. "If I see one more IF PUSH COMES TO SHOVE headline, I'm going to throw up," he steams. It has convinced him that Marimba should have been a little quieter. "We were in the media too much talking about the vision for the company, and that's just not something that's sustainable. You want press when you have some customer successes and large-scale implementations," he says.

Polese disagrees. "The No. 1 thing you want when you're starting a company is momentum, so you can attract investors, customers, employees, and partners," she says. John Doerr, the kingmaking Kleiner Perkins partner who has worked with Polese, shares that philosophy. Speaking to FORTUNE last fall, Doerr said: "Early on you want to get into the minds of key influencers and build lots of enthusiasm for what you're doing. I think that's critical."

Still, Doerr knows hype can't guarantee success. While Kleiner is known for successes like Sun and Amazon.com, it has also backed ventures like Go, Slate, and Dynabook, forgotten startups that garnered enormous hype before they tumbled.

By the spring of 1997, with push and Marimba's attendant consumer-channel strategy all but dead, the company started to focus exclusively on enterprise markets. Steve Williams, Marimba's vice president of sales, changed Castanet's pricing to suit FORTUNE 500 businesses and landed major deals with companies like Ingram Micro and FedEx.

Bereft of push, Polese had a more difficult time slotting the company into a category the public could easily understand. In summer 1997, David Cope, whom Polese had brought in to be Marimba's first vice president of marketing, hired a new PR firm, Thomas Associates, which he felt had more expertise with enterprise clients. (Niehaus Ryan resigned from the account before it was fired.) Polese and Cope came up with an acronym to replace the buzzword they'd lost, but for one reason or another ADM (application distribution and management) never quite developed the tingle of "push."

If potential FORTUNE 500 customers couldn't get a handle on just what Marimba was, they sure knew Kim Polese. Recalls a former Marimba executive: "IT guys would say, 'I've heard of you and Kim, and she's got pretty red hair, but what do you do?' It happened every single time we went to new accounts." Polese did make speeches and customer visits in which she explained that Marimba was now in the business of facilitating the delivery of software within large enterprises, but these were overshadowed by her continued presence in the public limelight. She had a starring role in Silicon Valley's budding political scene, thanks to patron John Doerr, who invited her to visit the White House as a member of TechNet, a lobbying group he'd helped create. That led to regular meetings with Vice President Al Gore and a collaboration with him on a public-school project called Education Dashboard, which Polese highlighted in several speeches. She was a mainstay at Silicon Valley fundraisers for Bill Clinton. She was even part of a small group of tech execs who met with former Russian prime minister Viktor Chernomyrdin at Cisco.

Then last fall she appeared in the Anne Klein fashion ad. Polese says she agreed to pose for the ad, which appeared in magazines like Vogue and Harper's Bazaar, because the other women pictured--Ann Richards, Esther Dyson, and Reba McEntire, among others--lent a "credibility factor." Explains Polese: "Here's a group of strong, classy women, and I think that reflects well on Marimba." Polese also maintained her high profile with constant public speaking, something she believes is an important part of her job as CEO. Working with Raymond Nasr, a Novell PR exec who writes speeches for CEO Eric Schmidt and moonlights as Polese's co-writer, she often stays up late honing rewrites.

Speeches keep Polese in the limelight, but are they good for Marimba? Another former exec recalls with frustration a presentation at a Harvard conference where Polese shared the stage with heavy hitters like IBM's Lou Gerstner, Sun's Scott McNealy, and Oracle's Larry Ellison. The others all talked in some detail about how their products addressed the future of technology. Polese's presentation was more a visionary disquisition on the new economy and the Internet's place in public schools. "It's beautiful that she's educating the world about this, but we wanted some pull-through," says the former exec. "Let's get everyone to understand what we do as an organization and put some money in the bank."

Polese doesn't think the objection is justified: "I don't flog the product because I have respect for the audience. They came to hear a speech about a vision of where the Internet's going. I put what we're doing in the last ten minutes, where I'll talk about software as a service and about application hosting."

In spring of 1998, marketing VP Cope left Marimba, quitting before he was fired. Since then, seven other members of the marketing team have left or been fired. PR firm No. 2, Thomas Associates, parted ways with Marimba last August.

Which brings us to marketing round No. 3 and Marimba's new vice president, Jacqueline Ross. You know, the one who's trying to shift everybody's attention to "the Marimba Show." Ross has led a search for a new catch phrase to supplement ADM; she and her team came up with ISM, for Internet services management (at least the word Internet's in there). She retooled Marimba's Website, replacing an oversized photo of Polese with smaller images of each key executive. The Website now tells people that Marimba is "Enabling the Internet," and Ross posted a new marketing statement that manages to employ the words "e-business" and "e-commerce" no fewer than 15 times. "This is what we were doing all along--we just didn't have the words to describe it," says Polese, referring to Internet services management.

If this sounds like new spin for the same product, it is. Marimba has an IPO looming in the next few months, and it must position itself for that big event. Rob Enderle, an analyst at the Giga Information Group, a research firm in Cambridge, Mass., says: "They're showing good growth year over year, and undoubtedly there is a market for their product. But it doesn't look like a big one, and it doesn't look like it's growing. Right now they're a small, interesting company, but they're not the potential world beater they were when they first came out." (This is, of course, the same Rob Enderle who was quoted in 1996 telling Information Week that Marimba's product "has the potential to revolutionize the software industry.")

Polese hopes investors will be impressed by big customers like Bay Networks, Seagate, and Bear Stearns, which use Castanet to send updated applications to their outside business partners--a form of e-commerce. And she is trying to convince people that Castanet will have a key role in the emerging and much talked about trend of application hosting, where it could serve as the mechanism that sends out software from a remote server to a corporate customer's desktop PC. Polese understands that some people will pigeonhole Marimba as what Nina Lytton, an analyst at Open Systems Advisors in Boston, calls the "pink poodle of the plumbing market." But Polese herself hasn't given up on a grander dream. She wants people to believe that Marimba is, as she puts it, "an important and visionary company." She's still a believer.