(FORTUNE Magazine) – Stephen Hilbert, chairman, president, and CEO of Conseco, was thoroughly warming to the plain-spoken fellow across the table, Lawrence Coss, chief executive of the mobile-home mortgage lender Green Tree Financial. It was September 1997, and in the industrial-chic glare of Manhattan's Royalton Hotel, the two made an odd pair. Handsome and flamboyant, Hilbert sports Saville Row-style double-breasted suits and silk pocket squares; pals around with celebrities ranging from Donald Trump to Gene Hackman to A.J. Foyt; and is the leading patron of arts and culture in Indianapolis (where the symphony hall, botanical garden, and sports arena are named for him or his company). Coss, on the other hand, is a balding, gravel-voiced, sober-suited soul who shuns the limelight as much as Hilbert adores it.

Hilbert recognized, however, that the pair had plenty in common as businessmen. Both came from blue-collar backgrounds and built FORTUNE 500 companies out of sheer doggedness and salesmanship. Both had produced fabulous returns for shareholders. And both had been royally rewarded for it: The two were regularly near the top of CEO pay charts. Most important to Hilbert, they served virtually identical customers. Coss' borrowers were the same low- or middle-income factory workers, waitresses, and farmers to whom Hilbert sold life and health insurance. "Even then at that first lunch," says Hilbert, "I realized that the beauty of Green Tree, pure and simple, is that their market is our market."

Hilbert was on the prowl for an acquisition, and Green Tree, with its highly compatible clientele, looked like an unusually good fit. Indeed, the more he talked with Coss, the clearer it became that the fit extended beyond their customer base. Both ran scrappy entrepreneurial shops and delighted in outfoxing lumbering rivals with gilded names. The two swapped stories about their early days as salesmen, when Hilbert hawked encyclopedias door-to-door and Coss flogged used cars. They even share a hobby: Coss raises quarter horses at his retreat in South Dakota, while Hilbert owns thoroughbreds, one of which ran in the Kentucky Derby this year. "I realized not only that their market was our market," says Hilbert, "but that their culture was our culture." Seven months later that statement was literally true, as Conseco agreed to buy Green Tree for $7.6 billion.

What impressed the market about the acquisition, however, was not the cultural harmony that so captivated Hilbert. It was the price. To pay $7.6 billion for Green Tree--a company with a net worth of less than $1 billion--was astounding even in the midst of a takeover mania in financial services. Investors reacted by pounding Conseco stock on the news of the deal, and they haven't let up since. At $57.75 before the deal was announced, Conseco now hovers around $21, which amounts to a loss of almost $7 billion for shareholders.

Hilbert vigorously denies that he overpaid. The general decline in financial stocks and typical Wall Street myopia have driven his stock price down, he says, but one day the world will call this "one of the best if not the best deal ever done in financial services."

In one sense, of course, a terrific deal was struck, but not the one Hilbert means. The sale of Green Tree may well have been the deal of the decade. And the person who pulled it off was not the flashy CEO of Conseco but his self-effacing lunch companion, Larry Coss.

What must make the failure of the Green Tree deal particularly galling to Steve Hilbert is that making acquisitions is his specialty. In the insurance business he is known as a brilliant consolidator, probably the best in the industry.

The son of a telephone operator and a maintenance man in rural Indiana, Hilbert was a washout in college who found his true calling while selling encyclopedias door to door. At age 19, in his first year pressing doorbells, he earned $19,000, the equivalent of $100,000 today. After a stint in the Army, he swapped encyclopedias for insurance policies and quickly rose to be the head of life insurance sales for Aetna in Indiana.

It was there that he observed how big insurers' bloated costs and old-fashioned products left them vulnerable to a quicker, more committed rival. He realized that a company that consolidated lots of small carriers and streamlined their operations into a flexible, low-cost competitor could gain scale like the big guys without taking on their cost structure. Such an outfit could steal their lunch. That's especially so, Hilbert figured, if the new insurer focused on a market that was badly served by the giants from New York and Hartford: blue-collar households of factory workers and farmers, "where the wife packs the husband's lunch, just like my mother did."

To start acquiring, Hilbert needed seed capital, but he didn't try Wall Street. "Even the Indianapolis investment bankers wouldn't give me a cup of coffee," he recalls. So, in 1979, Hilbert turned to his future customers. Backed by a $10,000 loan from his father, Hilbert conducted a kind of barnstorming IPO, crisscrossing all of Indiana's 92 counties in a dusty, leased Cadillac DeVille to sell shares in his new insurance company to everyone from farmers to gas-station owners. "Even then, Steve had an uncanny ability to get people to believe," says his lieutenant at the time, Ron Hunter, now CEO of another Indianapolis company. By 1982, after almost three years on the road, Hilbert had raised $4 million from no fewer than 2,500 investors. Now he was ready to start building.

From the start, Hilbert's favorite acquisition targets were companies that sell life, major medical, and long-term-care insurance to Middle America. Once the target is in hand, Conseco moves in to shut down its headquarters and fold investing, customer service, systems--everything but the sales force--into Conseco's sprawling home in the Indianapolis suburb of Carmel. The accent is on speed. When Conseco closed its purchase of Great American Reserve in 1990, for example, convoys of trucks packed up computer tapes, PCs, and files at Great American's headquarters in Dallas on the Friday night following the closing. When Great American reopened for business on Monday, customers who were used to talking to a Great American rep in Dallas actually found themselves talking to a rep in Conseco's customer service office. Most never knew the difference.

All told, Conseco has absorbed no fewer than 44 insurers during the past 17 years. By folding his acquisitions' operations into his superefficient, consolidated back office, Hilbert figures that he has been able to pare staff by at least 40% at the companies he acquires. The strategy has served Hilbert's shareholders amazingly well. From 1987 to 1997, Conseco finished with a higher return than any other company in the FORTUNE 500, gaining a remarkable 52% a year.

To be sure, Hilbert has shared liberally in the bounty. Throughout most of the '90s his pay packet has included a $250,000 salary, plus a bonus equal to 3% of Conseco's annual net profits, along with the usual CEO's bargeload of options. From 1992 to 1996, he pocketed $277 million and was America's highest-paid CEO over that period.

Hilbert's income funds a sumptuous lifestyle. In Indianapolis, his annual 500-guest Indianapolis 500 bash is legendary. But for sheer presentation, nothing quite matches his home. To get to it, a guest drives down a quarter-mile-long allee of trees, branches fully lit, before arriving at a 23,000- square-foot, Norman-style chateau. Inside the mansion's towering cupola, visitors can admire murals depicting the life of Alexander the Great, whom Hilbert identifies as one of his favorite historical figures. ("He certainly did a lot by a very young age," he marvels.) The bedrooms are decorated with ornate landscapes and curvaceous nudes painted by the French rococo masters Watteau, Boucher, and Fragonard. "I used to collect impressionists," confides Hilbert, who recently auctioned away a Monet for around $5 million. "But if you want to know the truth, you get a much better value in some of the great 18th-century French artists."

The art of basketball is also suitably honored (this is Indiana, after all) in a separate building that houses Hilbert's personal basketball court. The venue is a full-sized replica of Indiana University's Assembly Hall (sans bleachers, of course). The reproduction is accurate down to the electronic scoreboard, the victory banners hanging from the rafters, and the official Hoosier uniforms that Hilbert and his buddies don for their games.

Hilbert shares all this with his sixth wife, Tomisue, 29. Part of the Hilbert legend is that he and Tomisue met six years ago, when she popped topless out of a cake at his stepson's bachelor party. Hilbert denies that particular tale but allows that Tomisue was working as an exotic dancer when they met. Her career choice didn't bother Hilbert: "She was a single mom working to support her child," he says. "I respect that she worked and didn't go on welfare." Now retired, Tomisue takes lessons on the saxophone, an instrument she played in the high school band. She and Hilbert also travel to races with some of their 20 thoroughbreds. One of them, Stephen Got Even, finished fourth this year in the Preakness and 14th in the Kentucky Derby. "Winning," says Hilbert, "would be like watching a child win the Olympics."

In the months after the Royalton lunch, Hilbert frequently returned to his idea of buying Green Tree. It was the kind of big, brassy deal he had been looking for, one that could transform Conseco from a successful insurer to a much broader company--a financial services Wal-Mart for his Middle American market. Green Tree seemed like the ideal partner. There was the opportunity for cross-selling: Green Tree borrowers could be efficiently sold Conseco insurance products, and vice versa. Better yet, there was Green Tree's 30% profit growth rate, which could spice the modest pace of Conseco's well-run but mature insurance business. "I knew this was the deal to make us No. 1 in Middle America," says Hilbert.

That meant first coming to an agreement with Coss, an extraordinarily shrewd businessman who grew up in hardscrabble poverty in South Dakota and had honed his selling skills on used-car lots. "The guy was the greatest closer I've ever seen," says Kenneth Roberts, a longtime colleague, now the head of a small Green Tree rival, who hired Coss as a used-car salesman many years before. "He would tell car buyers, 'You have to give me a deposit tonight if you want this low price.' Or, 'This is the last model and another buyer wants it, so it's now or never.'" While Hilbert is enthusiastic and impulsive, Coss is reflective and cautious, a master thinker who conducted strategy meetings while marching colleagues through the walkway-connected labyrinth of stores and offices around Green Tree's headquarters in downtown St. Paul. "He seems like a simple guy from South Dakota, so people underestimate him," says a former Green Tree employee. "But he's incredibly street-smart."

In any event, a Conseco-Green Tree merger was out of the question as long as Coss' stock kept flourishing. That made Green Tree simply too expensive for Hilbert to consider. Then, in late 1997, Green Tree shares cratered. The main problem: Coss' rapid-growth strategy for the previous ten years had just backfired.

The reason Coss had been able to grow much faster than competitors while deploying a paper-thin capital base was that he securitized the loans he was making to customers. That is, he assembled the active loans into big packages of so-called asset-backed notes and offered them as publicly traded securities. That allowed Coss to raise the cash he needed to lend without having to sell tons of stock or long-term debt. He could simply use the proceeds from a securities sale to book more mortgages.

As a big securitizer, Green Tree used an accounting method called "gain on sale," which requires booking the profits on each package of new loans not year after year, as the cash gradually rolls in, but up-front, as soon as the lender sells the securities. (In September, Green Tree switched to a more conservative measure based on actual cash collected.) In good times, Green Tree would issue $1 billion in asset-backed notes over a quarter and immediately chalk up 7% of that, or $70 million, in earnings--even though very little of the cash payments from customers had yet materialized.

Though it's an entirely legitimate method, gain-on-sale requires a chain of tricky assumptions. A securitized loan may last for 15 or 30 years, yet gain-on-sale accounting requires Green Tree to forecast exactly how much cash it would produce over its entire lifetime. That required dicey predictions on the number of clients who'd default and prepay, as well as on the future course of interest rates.

Neither Green Tree nor its competitors foresaw how a sharp decline in rates would pummel the supposedly profitable loans already on the books. In the past, mobile-home buyers had been loyal, undemanding customers who, once they had a mortgage, seldom shopped for a better deal. But when rates started falling in 1995, new players saw a chance to wrest business away. Big institutions like Nationsbank, as well as a flock of upstarts, started calling Green Tree's customers, offering to refinance their 13% home loans at, say, 11% and add more cash in the bargain. Coss faced what he'd never seen before: a flood of prepayments in the mobile-home market. The accounting implications were clear: The packages were going to bring home far less in real earnings than Green Tree had already booked in paper profits.

Coss had no choice but to admit that his assumptions were too rosy and that he would need to book a big loss. In November of 1997, he announced that Green Tree would take an estimated $150 million write-down for the fourth quarter. Coss personally assured edgy analysts and investors that this write-down would be the last.

He was wrong. In early 1998, Coss confessed that the prepayment problem wasn't over and would cost another $40 million. But the shocker was an additional $200 million restatement of 1996 earnings for faulty accounting that misjudged prepayments. "I only want to go through the step of giving that kind of bad news once in my life," says Coss. "It really hurt my credibility to do it twice."

The second wave of losses threatened Coss with a lender's worst nightmare, a paralyzing inability to borrow to make its loans. Besides securitizations, Green Tree needed lots of short-term borrowings to fund the loans until it gathered them into big enough packages to sell off. Green Tree relied heavily on commercial paper, regularly issuing almost $1 billion in 30-day and 45-day notes to "warehouse" its mortgages. But in the wake of the write-downs, Moody's and other agencies lowered Green Tree's commercial paper rating and warned of future downgrades. That scared off the biggest buyers of Green Tree notes, the money market funds. Suddenly, Green Tree could no longer issue new commercial paper at a reasonable price to pay off the debt coming due. To make things worse, the banks that Green Tree had engaged as backup to the money market funds drastically cut their commitment. With Green Tree technically in default, the banks renegotiated the size of the backup loan, cutting it from $1.5 billion to $750 million--and charging a higher interest rate.

As a last resort, Green Tree turned to the investment bank Lehman Bros. In February 1998, Lehman agreed to lend Green Tree an additional $500 million--but at the price of sharply diluting Green Tree shareholders. In addition to charging a steep variable interest rate on the two-year loan, Lehman demanded warrants to purchase 2.7 million Green Tree shares, equal to 2% of the total outstanding, at the then-prevailing price of $22.50. (In the merger 60 days later, the warrants would yield Lehman a $43 million profit.)

The loan from Lehman staved off a crisis, but it left Coss convinced that he needed far more capital. "We were always vulnerable," he admits now. "We grew like a rocket without a strong financial base." He decided that to keep growing safely and profitably, his creation required a big infusion of equity.

To get the equity, Coss had two options, neither of them especially attractive. He could try to sell Green Tree to a well-capitalized acquirer, assuming one could be found. Or he could attempt to keep Green Tree independent by raising a modest amount of capital to tide him over until the stock rebounded, if it did. With Wall Street cheering once again, Coss reckoned, he could float a big stock offering that would put Green Tree back on solid financial ground.

With that in mind, Coss began to explore his choices. He held discussions with Gary Wendt, the head of GE Capital, and Jack Grundhofer, CEO of U.S. Bancorp, about possible equity investments or an outright purchase. He also made a tentative plan with Lehman to float a debt and equity offering in April, as the first step in a plan to stay independent.

Hilbert, meanwhile, realized that his chance to snatch Green Tree had arrived. In early March he called Coss and proposed a meeting. If Coss was anxious to make a deal, he certainly didn't show it, turning Hilbert down not once but twice. Then, as the Conseco chief stewed, Coss went on a weeklong vacation in Hawaii. ("I had a ball," says the usually phlegmatic Coss.) As his sole concession to Hilbert's overtures, he agreed to have their two CFOs meet at Green Tree's headquarters and talk about the company's financial health. The day Coss returned from his vacation, March 30, Hilbert called again. This time Coss agreed to meet, in Green Tree's paneled conference room overlooking downtown St. Paul.

Going into the meeting, Coss told FORTUNE, he sincerely thought that staying independent was his best bet. He regarded the company's financial troubles and low stock price as merely temporary. "We were in the penalty box, but we were going to get out," he says. He wasn't really interested in GE Capital or U.S. Bancorp because they weren't a good fit with his entrepreneurial culture. He ultimately agreed to the merger with Conseco, he says, not because price was paramount but because Conseco offered the best fit.

If indeed that's what he thought, he left a distinctly different impression. As Hilbert remembers the meeting, Coss immediately said that he wanted a price with a "5" in front of it--in other words, at least $50 a share. Coss also told Hilbert in so many words that it was now or never: Hilbert had to meet his price within the week, because Coss planned to start a road show with Lehman in early April to promote the proposed bond and stock offering. Once he had raised the capital, he explained to Hilbert, Green Tree wouldn't need a partner. Throughout the negotiations he made it clear that there were other suitors circling. "He made us aware of the potential buyers and used them to his advantage," recalls Conseco CFO Rolland Dick.

Anyone who knows Coss says what was going on was an extremely high-stakes game of poker--in which Coss' best card was his own chutzpah. Consider the "5" handle. The stock was then selling in the high 20s, so Coss was asking for at least a 78% premium for a company that had been on the ropes only three months before. In fact, the stock had never closed above $50, even in its glory days. As for the other suitors, it's not clear that Coss had any other serious buyers. The best candidate, apparently, was U.S. Bancorp of Minneapolis. CEO Grundhofer told FORTUNE that Coss called him about selling the company. "We had a moderate interest," says Grundhofer, "but we were tough on valuation. We just wouldn't have been willing to pay the kind of numbers that Conseco paid." Donald Howard, a Green Tree director and former CFO of Salomon Bros., confirms that account: "The Green Tree board considered only one bid." And that was Hilbert's.

Nor was the Lehman offering really an attractive option. Yes, it would have raised more capital, but at fire-sale prices. The interest rate on the bonds would have been around 10%--junk bond territory. And the stock would have sold not at the $50-plus that Coss was demanding from Hilbert but at its market level, in the high 20s.

Even so, Hilbert worried that the deal--and his chance of creating a financial Wal-Mart--could easily slip away. To this day, he remains convinced he was competing with serious buyers: "If we hadn't bought Green Tree, someone else would have paid $53. I know it." He adds: "I wouldn't be surprised if GE Capital was camped right outside his door. Larry certainly let us know he was being pursued. And I knew that this was a coveted organization."

Anxious to clinch the deal before any other buyer appeared, Hilbert rushed to meet Coss' deadline--set one week after the CEOs had first discussed a union, an unusually short period for the two chiefs to evaluate the economic potential of combining companies and determine a reasonable price. Like any smart used-car salesman, Coss allowed no time for Hilbert's enthusiasm to cool. On April 6, Coss and Hilbert gathered with their attorneys and investment bankers to hammer out a final number. Hilbert proposed a figure in the 40s. Coss countered that he would not sell for less than $53 a share. If Conseco wouldn't pay, he would walk away. Hilbert agreed. The next day, Conseco announced an agreement to buy Green Tree for Conseco stock equivalent to $53 per Green Tree share, or $7.6 billion.

For the onetime used-car salesman it was the closing of a lifetime. The price Coss finagled was higher than Green Tree stock had ever fetched--and a staggering 83% premium over its close of $29 the previous trading day. "Larry is the best poker player I've ever seen," says a former colleague of Coss'. "Smart as he is, Hilbert walks into a room with Lawrence M. Coss, and Hilbert loses."

While Hilbert raved about a combined "juggernaut with a $20 billion market cap," investors stampeded. The day of the announcement, Conseco shares tumbled 15%, not least because Conseco announced that Green Tree's forecasts for prepayments, defaults, and the general economic outlook were still far too optimistic, and more big write-downs were coming. The hit eventually came to $549 million, not including a $148 million restructuring charge. With Green Tree's equity running dangerously low, Conseco had to pump in $1.1 billion to bulk up its new acquisition.

Just when it looked as if things couldn't get worse, Russia's default and the Long-Term Capital debacle shook Wall Street in the fall of 1998. As investors rushed for safe havens, the market for asset-backed securities slumped. A slew of smaller rivals tumbled into bankruptcy, and larger companies took huge write-offs. First Union's $2.2 billion acquisition of the Money Store, for example, has proven a disaster, and bank officials have acknowledged that they badly overpaid. Ironically, Conseco's adviser, Merrill Lynch, had cited the huge premium First Union paid for the Money Store as a justification for the price Conseco proposed for Green Tree.

For his part, Coss left Conseco at the end of 1998 but remains a director. His brilliant brinkmanship made many of his shareholders rich (at least those who bailed out of Conseco soon after the merger). However, he hasn't sold any of his stock, and has even bought more. But with more than $150 million in Conseco stock, he's still a lot better off than he would be had Green Tree remained independent and collapsed like so many of its competitors.

Hilbert, meanwhile, strongly disagrees that he overpaid for Green Tree, and to prove otherwise he is growing its loan bookings at a dizzying 30% a year. (The figure for solid, conservatively managed competitors like Household International and The Associates is in the high teens.)

Unfortunately, it won't work. Green Tree (the name is changing to Conseco Credit) is still an excellent business, but it can't be the locomotive that pulls Conseco out of its hole. Though the evidence so far is inconclusive, the market fears that Conseco is sacrificing credit quality to pump up sales--and that a big chunk of the fat earnings Green Tree is reporting will never show up in cash. Says Patrick Finnegan, an analyst with Moody's: "Past examples of companies that grow that fast show that they sacrifice credit quality."

The second problem is also basic: Conseco is too weighed down with debt to aim for the skies. Hilbert borrowed the $1.1 billion he used to bail out Green Tree by issuing expensive preferred stock. That helped push to 46% Conseco's ratio of debt and preferred stock to capital, according to Moody's; financial stalwarts boast ratios closer to 30%. Despite the already heavy load, Conseco has to keep jacking up its borrowing to fund Green Tree's growth. "Conseco just doesn't have a big enough cash flow both to grow Green Tree this rapidly and to pay its interest if it wants to maintain its credit rating," says Finnegan. He adds that Conseco needs far more equity capital to weather another meltdown in the securitizations markets.

Which brings us to the final--and most ironic--parallel between the careers of Steve Hilbert and Larry Coss. Right now, Hilbert finds himself in a version of the dilemma that Coss faced before Conseco bailed him out. On the one hand, he can radically slow Green Tree's growth and either sell assets or float equity to lighten his debt load. That is probably his best option, and Hilbert is slowly moving in that direction: He's already agreed to explore selling a riverboat casino and other properties, which could raise about $500 million. If Hilbert embraces a comprehensive plan to raise equity, Conseco could perform very well from here, but it will require Hilbert to float shares at what he considers a ridiculously depressed price. Even then, it's highly unlikely that shareholders will ever get a decent return on the $7.6 billion that he poured into Green Tree.

Hilbert has one other alternative. He can sell the company to a suitor with deeper pockets. At the stock's current price of around $21, Conseco could be attractive to an ambitious financial services firm. Hilbert says he simply isn't interested, even at $30 a share, a 43% premium to the going price. "I don't know if I'd say no to $30 before or after I stopped laughing," he says. One assumes this proud company founder is sincere. Then again, maybe he's bluffing. If so, he's learned from the best.

REPORTER ASSOCIATE Patty de Llosa