Search for Tomorrow

November 19, 2007

By Alison Overholt

It was an advertiser’s worst nightmare. Last summer, the New York Post ran a breathless story about a gruesome murder in which the victim’s body was hacked to pieces, the parts stashed in an old suitcase. Opposite the online version of the story ran an advertisement cheerily touting the benefits of . . . luggage.

The offending ad was served up using Google’s search-marketing technology called AdWords. For the uninitiated, search-engine marketing lets advertisers bid on keywords or phrases. Top bidders then have their ads appear alongside search results whenever a user types in that phrase. Or in this case, the ads run alongside editorial content containing the keywords. Suitcase in the article. Suitcase ad. We have a match.

It was an excruciating goof, emblematic of the tricky juncture where the search-engine marketing industry finds itself today: The dream is to transform the Internet into a sales tool that finally delivers on the promise to generate eminently qualified, targeted, and trackable customer leads that convert quickly into big sales. Marketers that embraced the first generation of these tools have already achieved phenomenal success by targeting ads to consumers seeking products and services on search engines. Now the toolmakers, predominantly Google and Yahoo, want to serve up ads to the rest of the Web — delivering relevant messages not just when buyers come to a search engine already hunting for something, but any time, and any place.

There’s big money at stake in nailing the solution. In an advertising environment that has steadily weakened over the past three years, search marketing has breathed new life into online advertising by showing how powerful it can be when an advertiser catches a shopper’s attention at that perfect moment when she is ready to buy.

Advertisers rewarded the nascent industry by doubling its revenues in 2003 to the tune of $1.9 billion (a figure that is expected to jump again in 2004 to $2.8 billion dollars, says Forrester Research), or nearly one-third of all online advertising spending. And those hefty totals ring up in small increments: Expedia, for example, is the top bidder on Yahoo’s service for “Miami vacation,” paying 83 cents a lead.

Three techniques are emerging that could push online ad revenues even higher: contextual ads, behavioral ads, and local ads. But none of them are slam dunks. In pushing the envelope to make related text ads as ubiquitous as the 30-second TV spot, the search engines and the marketers that use them tap dance along a very fine line between what is helpful and what is obnoxious, what is exciting and what is simply in very poor taste.

Keeping Things in Context

“We advertise on TV and radio,” says Steve Hartmann, the director of online marketing for eHarmony, an online dating service. “But we discovered that a lot of people only vaguely remember our name. Maybe they’d just catch the word ‘harmony’ and that we were a dating service. When they typed that in at a search site, that’s when they’d find us.” Hartmann discovered that close to 70% of eHarmony’s new customers from online advertising channels arrive through search-based ads.

Seeking to go to the next level, Hartmann bought contextual ads from both Google and Overture (owned by Yahoo). He did so because the ability to place eHarmony ads where serious-minded singles spend time on the Web — say reading an article on CNN’s Web site about a scientific discovery on the brain chemistry of love — sounded ideal. Google and Overture dominate search marketing and offer contextual ads through partners such as AOL, MSN,, and the Web sites of The New York Times and The Washington Post.

But contextual ads don’t seem to target consumers as effectively as pure search ads. “We’re definitely not seeing the traffic from newspaper sites that we see with search engines,” Hartmann says.

There are many reasons for this lower success rate. Many publishers are leery of these ads for fear of blurring the lines between editorial and advertising, so their reach is limited. The bugs in the system also remain, ergo the very targeted but terribly unfortunate luggage ad. These incidents expose the flaw in the logic that purchasing a given keyword can guarantee relevancy to the material next to the ad.

Overture has responded by instituting an editorial review process. “We needed human influence to deal with those words that are ambiguous in meaning,” says David Karnstedt, senior vice president and general manager of direct business at Overture. Google, meanwhile, believes its technology can fix any editorial problems.

Perhaps more problematic to contextual marketing’s prospects is the very nature of the Internet experience. When someone types the name of a product or service into a search engine, chances are he wants to find it and buy it. When that same person surfs a news or content site, he may just be catching up on the day’s events. “They’re not in shopping mode, they’re in browsing mode,” says Danny Sullivan, an analyst who runs And there’s not any fine-tuning that can be done to fix that.

Fresh Ads for Good Behavior

Because of the inherent flaws in contextual marketing online, the stakes are even higher for the online advertising industry’s next big play: behavioral marketing. This technique promises to serve up ads based on a Web surfer’s habits and mind-set. “You’re targeting the person, not the content,” says Forrester analyst Charlene Li. It’s far more ambitious — and more advertiser-friendly — than contextual marketing. “You could never target intent before, in any medium,” says Li, capturing what’s exciting about the new method. “You just put your message out there around content that seemed likely to attract the right people and hoped it worked.”

To deliver on that opportunity is a daunting technological task. It requires analyzing the surfing habits of millions of users in order to define segments based on what users are reading, how often they read it, and what products they search for. One of the first entrants in the market is Kanoodle, a small New York-based search-marketing firm. It has joined with online advertising network 24/7 Real Media to launch BehaviorTarget, a behavioral search service. “We’ve created a taxonomy of 486 topics that we’ll roll out slowly as we reach critical mass with each audience segment,” explains David Hill, president of media solutions at 24/7 Real Media. “If you visit sports sites several times a month and fashion sites several times a month, you might fall within our ‘active women’ behavioral segment.” Unless you’re a guy who likes sports and fashion, or works in one of those industries.

Though a potential gold mine for advertisers, data collection on this scale — and at this level of detail — is, for many consumers, a little scary. It raises a host of privacy issues as well as questions about who retains the rights to or ownership of particular kinds of information. Lance Podell, president of Kanoodle, is quick to defend his company’s concern for privacy. “I’m tracking a cookie, not you, and you will be able to opt out at any time. That cookie doesn’t know your address or your Social Security number, it just knows your behavior. IP addresses aren’t being collected,” he says.

Lycos is one of the first companies to sign up for BehaviorTarget. Says Steve Gross, Lycos’s vice president of marketing, “We need different and creative ways of making money off of our user base. We have a unique user base in the tens of millions, but our ability to understand that base was limited.” Consider a group of users who have been defined as car enthusiasts because they visit automotive sites such as and Gross is eager to sell ads aimed at that group and deliver marketing messages to them even when they’re visiting a technology blog. “Television advertisers would love to understand composition this well, to identify audiences and what they want, beyond just age and sex, which is how they sell ads now,” says Hill.

So although it may seem simplistic to get lumped into a bucket because of a few Web-surfing habits, it is better information, as Gross notes, than most advertisers have to work with. Yet even its boosters acknowledge that behavioral marketing is a work in progress. “There are things we’re still working out,” admits Hill. “For example, somebody goes to a series of auto sites, but how many times must they visit before they become part of that segment? Our plan is to use our software to watch their behavior over the course of time, then do branding studies to hopefully get the mix and the formula right.” If that works, and if the privacy guardians don’t howl too loudly, this may be the method with the best chance of taking off.

Think Local, Act Local

If the search engines have learned anything from their original keyword-advertising system, it’s that when people type a product name into the search window, they want to buy that product now. Chances are, though, that users see ads from online retailers who can ship it within the week, not the minute. How can you feed immediate needs, such as where’s the closest party-goods supply store that’s open late? Hit ’em where they live.

In March 2004, Google launched a beta site called that lets users search by zip code or local address. Thirty percent of all advertising is purchased in local markets, and the Yellow Pages ad market is estimated at $14 billion per year — a stash that Google would love to break into. So far, the company has focused on importing Yellow Pages data into its local engine and getting the location algorithms right, without serving up additional ads alongside search results. The big question will be whether the advertisers arrive when local goes live.

Seth Berkowitz, vice president of business development at the car-buying information site, says no thanks. In theory, he’s a likely customer, since he sells customer leads to local auto dealers. But he’s focusing on buying location-specific keywords through traditional search-marketing products. “I love it when people type in ‘Los Angeles Honda Dealers’ because 20% of those people execute on those ads,” he says.

“To make local search work,” says Jupiter research analyst Gary Stein, “you have to get all the plumbers and dentists on there, then you have to get people to shift over to thinking about using Google to find plumbers.” Don’t expect local search to be a significant moneymaker for anyone until national retailers hop on board to let you know where the closest Home Depot is and whether they have that sander in stock you’re itching to buy.

It’s obviously too early to pass judgment on any of these emerging online marketing techniques, although each has significant flaws. Remember that, the precursor to Overture, was laughed at in its early days for selling simple keywords on a search engine. No matter which, if any, succeed, what’s clear is that these models for segmenting, targeting, and reaching customers are forever changing the way companies think about advertising as a sales tool. When you can select your target customer by geography, actual (not projected) buying patterns, and browsing behaviors — and track the return on investment of each ad by following a customer from the time she is targeted to the time she makes the purchase — it’s hard to go back to fuzzy math and schmoozy ad salespeople. The technocrats will have the last laugh, even if it’s at the marketers’ expense.


Why Twitter will change the way business communicates

October 30, 2007

By: Robert Scoble

Hard to believe that only 10 or 15 years ago we interacted with coworkers and colleagues with memos and phone calls. Email and instant messaging changed all that. Now there’s a new communications revolution coming. These services mix contacts, instant messaging, blogging, and texting, and they’re poised to make email feel as antiquated as the mimeograph.

The best known of the new services is Twitter. Since its debut last spring, it has been one of the fastest-growing apps in the history of the Internet. The best way to describe it is as a microblog service in which you tell people what you’re doing or thinking at any given moment. The hook is that you’re limited to 140 characters. “It’s strangely addictive,” says NBC videographer Jim Long. “Evidently, people are interested in what I’m doing, and I genuinely care about what they’re doing.”

Twitter’s basic idea has proven so popular that others have copied its premise and added features. Jaiku lets me include blog posts, my link blog, and more along with my mini posts. Pownce users can send files to one another, as well as calendar events. At Facebook, I can add such information as my favorite music and the syndicated Web feeds I’ve shared in Google Reader.

All this adds up to a new way to share information about yourself. Although the content of the messages can vary wildly from voyeuristically interesting to terribly dull, a frequent stream of updates can strengthen your brand. My 4,000-plus Twitter “followers” can get my blasts online or via text message, and each one is also its own Web page, which means that Google can see it and let people search for it. When you’re traveling frequently and working from coffeehouses or the backseat of a cab, these services are great to keep in touch with coworkers back at the office and with customers nearby. “I post where I travel and arrange user meetups,” says Betsy Weber, an evangelist with software firm TechSmith.

The professional intimacy these services create–hey, if you know someone’s whereabouts and musical tastes, you’re halfway home–can also win you clients. “People won’t do business with you until they like you or have a sense of trust,” says Cathryn Hrudicka, a consultant who uses Facebook, Jaiku, and Twitter. She has already gotten referrals from people she has met online because she has shown she’ll be available when clients need her.

Sales and marketing are lagging in seeing the potential here. When I used all these services to tell the world that my wife and I were expecting a child in September, I anticipated hearing from the world’s largest consumer-products companies begging me to try their latest diapers, food, car seats, and financial instruments. What came back? Nothing. Where was Procter & Gamble? Given what it and other companies spend acquiring new customers, there’s an untapped gold mine in Twitter and Facebook because we’re volunteering so much information about what we’re doing right now, whether it’s working on a project or eating a chicken-salad sandwich. Learning how to tap it correctly–both to sell to me directly and in seeing major trends in the millions of daily public posts–will be the next major challenge for these companies.

If we revisit this conversation again in three years, I suspect that we’ll have found all sorts of little uses for these services, and they’ll simply become what email is today: something we must do just to participate in the heartbeat of business.


eBay’s Chaos Theory

October 22, 2007

By: Chuck Salter

It was early in 2006, and Matt Carey, the new CTO of eBay, was attending his first focus group about the online shopping site. It was a memorable experience, to say the least. “It’s hard to use,” complained a longtime customer. She had been collecting antique glass on eBay for years. But lately, the treasure hunt was more frustrating than fun. “I get lost,” she said. “I can’t get back to my search results. I have to go all the way out and start over.”

“This is not good,” Carey thought to himself. This particular buyer was, as he puts it, a “dyed-in-the-wool, right-down-the-center customer.” What she was describing is known by the pejorative “pogo sticking.” To Carey, who had just moved to eBay after 20 years at Wal-Mart, it was the equivalent of “having customers not able to shop in your store because they can’t find the aisles.”

It is not news that eBay has lost the magic that made it an Internet darling a few years back. After peaking at $59 a share in late 2004, the company’s stock plunged to $23 two years later. CEO Meg Whitman may boast about the company’s latest stats–record number of users, revenue, and items listed for sale–but the fact is that the rate of growth at the company is slowing. EBay has tried to jolt itself by investing as much as $4 billion in Skype (which has yet to pay off) and $1.5 billion in PayPal (which has been far more successful). Yet 70% of revenue still comes from the core marketplace business. And as Carey recognized, the weakness there has become impossible to ignore.

How troubling is the slowdown? Despite the double-digit increase in listings and gross merchandise sales that the company reported last year, both of these key indicators have steadily decelerated over the past three years. In 2006, gross merchandise sales grew by less than 20%, the smallest rate ever. More troubling still, the number of active users–those who bid, bought, or listed at least once in the previous year–rose by only 14%, the slowest rate since 2001.

EBay is responding with a whole new strategic gamble–one some company insiders say is its most ambitious ever. The mastermind is John Donahoe, 47, whom Whitman brought aboard three years ago and installed as president of eBay Marketplaces (and as her heir apparent). His bold stroke–what he calls “our number-one strategic priority”–is recasting the site to focus primarily on buyers, not sellers.

As obvious as this realignment might seem, it is a sea change for an outfit that long regarded sellers as its main customers; some 1.4 million vendors rely on the operation for their primary or secondary income.

Donahoe’s key partner is Carey, 42, who is charged with making the buying experience efficient and fun again. Improving one of the Web’s most heavily trafficked sites without disturbing its global–and vocal–sellers’ network and its millions of loyal buyers is a challenge that Carey compares to a “four-wall expansion” at Wal-Mart: turning a standard store into a supercenter without disrupting day-to-day operations.

The good news is there are signs of progress. Wall Street has noticed–the stock has gone up by about $10 a share since its low a year ago. Still, shares remain about 40% below their high, and the ultimate outcome of this effort to revive eBay’s growth is in doubt.

“This is definitely an inflection point,” says Robert Peck, an analyst with Bear Stearns. As Jeff King, eBay’s senior director of product search (what eBay calls “finding”), puts it: “This is our biggest bet.”

Twelve years after a pony-tailed programmer named Pierre Omidyar built an unfussy auction Web site one Labor Day weekend, it’s easy to forget how swiftly and thoroughly eBay changed the online-shopping game. Within four years, customers had listed 130 million items and sold nearly $3 billion worth, giving rise to a new type of entrepreneur, the at-home eBay retailer. The site’s charm lay in the fact that the merchandise was utterly unpredictable, and in the way that auctions introduced an element of competition. The initial hodgepodge of obscure collectibles and discontinued items at bargain prices was joined by hard-to-find new products and pricey cars and jewelry. Part flea market, part Mall of America–eBay chalked up $52.5 billion in total sales last year, more than the sales of Amazon, Apple, and Nike combined. There’s still nothing else like it in size and breadth.

From the beginning, the strategy was to amass an unrivaled array of goods that would attract buyers. It worked well. In fact, as Donahoe now admits, it worked too well. The site became bloated and unwieldy. At any given point, it features about 100 million items for sale, with nearly 7 million new listings every day. “EBay’s abundance was one of its attractions,” Donahoe says. “But if you type in ‘BlackBerry’ and get 23,000 search results, it’s not that helpful.” (His offhand math is not far off the mark: A search in September produced 3,911 phones and PDAs, and 17,771 accessories.)

Donahoe is sitting in the employee cafeteria at eBay North, one of two corporate campuses in the San Jose area, in early September. It’s just after 8 a.m. The campus is coming to life, the parking lot starting to fill. Donahoe is already in midday form after his 6 a.m. Pilates class in the company gym. In a sense, he’s trying to do for eBay Marketplaces what Pilates does for his lanky 6-foot-5 frame: improve its flexibility. “This is not a one-time project,” he says of the drive to revamp the buyer experience. “We’ll make big changes over the next couple of years and keep iterating and innovating.”

Whitman and Donahoe worked together in the 1980s in the San Francisco office of Bain & Co.; Donahoe stayed and eventually became Bain’s worldwide managing director. In many ways, he says, eBay has been going through a natural evolution, from a wildly successful startup to a public company with global reach to, well, a maturing business. “Early on it created a market,” he says. “Now we have competition on all sides.”

Today, the company’s homegrown vendors can sell through their own Web sites, as well as channels such as and Shoppers have even more online options. A bargain is only a Google search away, and brick-and-mortar retailers have worked hard to upgrade the shopping experience on their sites with virtual assistants, gift registries, product videos, customer reviews, and liberal return policies.

Once an e-commerce innovator, eBay fell behind. “We were shackled by our own success,” says Eric Billingsley, who runs the engineering side of the finding operation. “When the company was growing 80% or 120% year over year, the mind-set was, ‘If it’s not broken, don’t fix it.’?”

“The buying experience hasn’t changed dramatically since 1999, compared with the rest of the Internet,” says Scot Wingo, president and CEO of ChannelAdvisor, which makes software to automate everything from auctions to shipping for sellers on eBay and other sites. “The highway is now crowded, and others are going faster.”

Historically, eBay made sellers the priority for a very good reason: They generate revenue. Sellers are the ones who pay eBay fees for listing an item, posting a photo, even processing a payment through PayPal, which eBay bought in 2002. But Donahoe realized that eBay had to stimulate shopping, and to do that, the company needed technology designed around the buyers’ needs. In late 2005, Donahoe began looking to hire a new CTO. Given the site’s size and complexity, there weren’t many candidates with the appropriate experience.

Then he met Matt Carey. Carey didn’t know much about eBay–in fact, he had never used the site–but he had helped build and oversee the technical infrastructure behind the world’s largest retailer, one of the most data-centric businesses on the planet. In two decades at Wal-Mart, he had experienced firsthand both unprecedented growth and the challenges of maturation. A half-hour into the interview, Donahoe excused himself and called a colleague: “We have to have this guy.”

Carey inherited a catastrophe. Shortly after he arrived in San Jose in December 2005, the site’s core listings, largely auctions, and those for its 600,000 individual stores were combined for the first time–a blunder no one now takes credit for. Previously when you typed in, say, “Sony PlayStation,” the search engine combed through only the core listings. To see the other merchandise, you had to surf over to the eBay Stores site and do a separate search or browse the stores. The goal of combining the entries was to show a broader mix of inventory on a single search; the effect was to give more exposure to the store products. The new setup was rolled out with no customer testing.

Within weeks, nearly every measure of eBay’s business was down. Bids. Return visits. The conversion rate, or percentage of listings sold. Average sales price.

In hindsight, it’s hard to understand why no one at eBay foresaw what would happen. Because eBay charges less for store listings than core auction listings, once they all appeared in a single search, many sellers shifted their inventory to save on fees. Suddenly, store merchandise, which tends to be pricier, was crowding out the auctions–and the bargains. Auctions bottomed out at just 17% of total listings, yet they still accounted for 91% of sales.

The misstep triggered headlines, a falling stock price, and pointed questions from analysts. Whitman explained repeatedly that the marketplace was out of balance. In March 2006, eBay rolled back the program. Finally, in August the company used its only real lever: It raised fees for store listings.

That fiasco became the catalyst for overhauling the buyer experience. Carey asked for a detailed report: When were shoppers abandoning the site? How much were they scrolling through the new search results? He discovered that there was no mechanism to create such a report. It took “many, many, many hours and days and weeks,” he says, to unravel exactly what customers were doing. It turned out that eBay collected all sorts of data about transactions–“It knew that business like the back of its hand,” Carey says–but little related to shopping. “I said, ‘We got gaps in the data. We got holes,'” he recalls. And his mission was to plug them.

Carey grew up in Okmulgee, Oklahoma, where his father operated the family’s furniture-and-appliance store. It was located in a four-story building, the tallest in town. As a boy, he dusted furniture in the showroom and rode the elevator for fun. As a teenager, he delivered air-conditioners, sold bedroom suites, repaired TVs. Meanwhile, his mother was working for IBM in information systems. “When we were young, she used to take me and my brother to the data center, and we’d sleep on the floor in her office while she wrote programs on punch cards,” Carey says.

After graduating from Oklahoma State University, he went to work for Wal-Mart as a programmer trainee, combining his retail and tech know-how. On his first day, he wrote a program automating a sales report for Sam Walton about the Sam’s Club stores–all 12 of them. The IT department was small enough, with only 300 or so employees, that he met Wal-Mart’s CIO early on. “I think I’m going to want to do your job one day,” Carey told him.

The CIO invited the 24-year-old to work alongside him for six months and learn the ropes. Carey eventually had a hand in developing virtually all of Wal-Mart’s major systems, from software that analyzed every inch of shelf space to programs that identified inefficiencies in the company’s global supply chain. “The lesson there was, it’s all in the data,” he says. “If you start with the lowest level of detail, you can answer any question about the business.”

He’d watched from Bentonville, Arkansas, over the years as colleagues left for tech companies like Amazon and Dell, and when eBay came calling, he was intrigued. Still, leaving the only employer he’d ever had was terrifying. “You’ve got no idea how hard that was,” he drawls. “No idea.”

Before moving to San Jose, Carey put the family’s dining room set up for auction on eBay. It sold within days. “A retired couple in Hot Springs drove in with a truck and picked it up,” he says. “I thought, Wow, that’s $1,000, man! This is totally powerful.”

He got his first taste of the eBay culture on day one. Everyone works in cubicles, but executives get individual conference rooms, decorated in a theme their colleagues pick out: Blondie for Whitman. Dennis the Menace for Donahoe. And Elmer Fudd for Carey, an avid hunter. Seeing his conference room for the first time–with two double-barrel toy shotguns mounted on the wall, plus a couple of comic-book covers–he remembers thinking, “Okaaay. Am I in the wrong room?”

Carey set about creating what he calls a “culture of analytics,” particularly around buyers and product development. More experimenting, more testing, more data. “I want to eliminate feelings and get down to true math,” he says. In just 10 months, his team built a faster and more flexible technology platform. His developers also began testing applications on small randomly selected samples of the eBay population (typically 1% or 2%).

In the old eBay, one former engineer had so many failed launches that he had earned the unfortunate nickname the Rollback King. Now, if a new feature doesn’t improve buyer engagement–a new metric, in which return visits, bidding, buying, and other activities are weighted–it doesn’t graduate from trials to reach a broader audience. “In a Darwinian sense,” says Billingsley, one of eBay’s top developers, “to be a survivor, something has to keep producing.”

The evolution of the eBay search engine is continuing, driven by the need to boost browsing and sales. One step is to give shoppers more relevant information, more rapidly. Until recently, the search engine relied on sellers’ product descriptions. When you typed in the name of a product or brand, the software looked for those words in the sellers’ 55-word listings. The results were then ranked according to the closing date of the auctions. If you entered “John Deere,” you could get a listing for a John Deere tractor or a set of John Deere sheets. By eBay’s definition, both were equally relevant.

Playing catch-up with other consumer-oriented sites, the company is now applying the “wisdom of crowds” to create a new feature called “best match.” Every click on the site is measured; the outcome of every one of the 2,600 searches per second is tracked to determine what leads shoppers to bid or to buy. If you submit “John Deere” today, you’ll see the John Deere products that most previous shoppers purchased. “We get flack that we’re trying to control search, but we’re letting the buyers vote with their clicks and say what’s relevant,” says search-meister King. “It’s a big, big, big change for us.”

Narrowing even the most relevant search by price or brand or size has been a particular problem for eBay. Unlike other retail sites that sell a set inventory, eBay has to index and classify a constantly changing universe of whatever people are selling. So where or has you pick from predetermined price options, for example, one new eBay feature lets you set your own price range. The site also steers buyers to those sellers with the most positive feedback.

EBay is launching a “snapshot view” in certain categories in time for the holidays; instead of the usual prominent text and thumbnail images, a larger image pops up as you scroll over the picture of a sweater or a vase. It’s the sort of functionality online shoppers have come to expect. “If they’re shopping for clothes,” says King, “they’re comparing us to Nordstrom now.”

What about serendipity–that item you weren’t looking for but are delighted to discover? EBay staffers talk about serendipity all the time. So at the bottom of the list of matches are a few outliers. “If we got rid of cheetah iPod covers, we’d lose a little of eBay,” King says.

What does all this mean for the sellers? Chris Hinze, who turned to eBay when asthma made him abandon his auto-mechanic business, is enthusiastic. Working out of his home in Portland, Connecticut, the 46-year-old refurbishes fixtures bought wholesale into what he calls “power showerheads” with dramatically more water flow. He’s an eBay PowerSeller, meaning his sales amount to at least $1,000 a month and buyers give him high feedback scores.

Hinze attended eBay Live, the annual gathering of thousands of sellers, for the first time this past summer. After one session, he approached King and mentioned that searches for “shower heads” and “showerheads” produced significantly different results. Back in San Jose, King had his team add the terms to their “stemming” project, which combines related words in the finding system. The result: a flood of customers for Hinze’s Superpowershower. He sold three months’ worth of merchandise in three weeks. “Crazy, huh?” he says.

It’s a good example of the power of eBay’s algorithms, both to steer shoppers toward what they’re looking for and to boost a small business 3,000 miles away. In essence, that was Omidyar’s original vision: linking strangers through a virtual transaction that served both parties well. An honest, efficient marketplace, he called it.

But tinkering with the search engine creates new winners and losers; some sellers bubble up, others disappear. No matter what, somebody’s unhappy, suspicious of favoritism, accusing eBay of tilting its playing field. Even minor tweaks can disrupt business for sellers who rely on automated software to manage hundreds or thousands of auctions. It’s all there in the often vitriolic discussion boards on the site.

The biggest question facing eBay today is whether the totality of the changes that Donahoe and Carey are implementing can do for eBay and its millions of sellers what the “showerhead”/”shower head” fix has done for Hinze.

Therein lies eBay’s central conundrum. “We don’t pretend to have all the answers,” says Donahoe. “We’re doing things that will upset some people. But we’re not just listening to the average noise. We’re sharply focused on what our buyers want and need.” Ultimately, the new strategy is a risk, but it’s one that eBay can’t afford not to take. Faced with the classic growth-company problem, it’s betting that it can regain momentum by becoming more like mainstream retailers while still offering stuff you can’t find anywhere else (Michael Vick’s purported handwritten notes for his televised apology in August: $10,200).

The buyers will decide if eBay made the right move. If they shop the site more regularly and purchase more Nintendo Wii consoles and Coach bags and iPhones and Elmer Fudd comics and antique glass, the sellers will applaud the changes. At eBay, there’s little doubt what’s at stake. “If we don’t change, we get marginalized,” says Carey. “We can’t let that happen.”

Employees, who seem to take pride in running a global democratic marketplace, profess a greater sense of mission. “We haven’t even released an eighth of what we’ve done,” says Billingsley. “That’s what excites me. It hasn’t even begun.” Customized pages are in the works. More social-commerce features. An eBay to Go widget with your favorite auction listings to post on your Web site or your MySpace page, complete with a clock to remind you to bid before it’s too late. It all sounds good.

But is it enough? Even eBay’s revamped search engine can’t find the answer.

Source: Fast Company