A close-up study of how a company brought in its own expert to upgrade its system, and what that was like.
Bringing in an expert to overhaul your systems can take you to the next level -- if you survive the experience
Challenge: To provide customized high-tech debt-collection services to a rapidly growing set of clients representing more than a million debtors
Solution: Bring in a full-time computer expert to replace an aging, cranky computer system with a networked system and customized relational database-management software to track and send out letters to debtors
Resources: A 486 Novell server to hold the company database; a 90-megahertz Pentium-based database server; four Pentium machines; a bulletin board for clients to dial into to upload account information or download reports* * *
It is 3 a.m. on a Sunday in August 1994, and in the recesses of an office park in Calabasas, Calif., Jose Gonzalez absorbs the bad news: his database has puked. Again. "Man, I can't believe it," fumes the weary programmer. Sometime around 10 o'clock, his log tells him, the system mysteriously rebooted itself, regurgitating the 200 million bytes of data that Gonzalez had spent the weekend feeding it. "All that work is lost," he mutters. "We'll have to start all over again." It will take two days to reload the data. So instead of heading home to bed, Gonzalez looks to watch dawn break from a stark, overlit office, hunched over a Pentium machine, listening to the fluorescents buzz and his keyboard clack, trying like some Sisyphus of software to push a binary rock once more up the hill.
The episode is just one more reversal in the race to install a new information system at Collectech, the high-speed, $5-million collection agency that employs Gonzalez. And while an act of God might help, Collectech's managers would settle for a clean data dump right now. The company's conversion from an outmoded central processing system to a custom-built client-server network hinges on it. More than a million records must be imported from the old system to the new before weeks of testing and debugging will make CTech Express, as its creators have christened it, safely habitable. Depending on whose schedule you consult, the conversion is already six months, nine months, a year late. "The pressure is on," says Gonzalez. "The board is desperate to see it happen. And if we're not up by September 1, I figure we'll be having one serious discussion."
If that discussion takes place, Marwan Kashou will be leading it. The impatient president and cofounder of Collectech, Kashou is a dark dynamo of a man accustomed to making deals and sales, and even wrenching change, happen fast. A new challenge can get this former weightlifter "pumped out of my mind," as he likes to put it. But this metamorphosis -- an ambitious technological overhaul of the company's operations -- is different. "We're taking a fairly mundane business and transforming it into a high-tech deal," he says.
Collectech got by with outdated computer systems through its first five years of manic growth. But those days are now over. "This company cannot grow at anywhere near the same rate without the right information systems," sighs Kashou. CTech Express, he insists, is the key to nearly doubling sales to $9 million this year. In fact, Kashou has more or less staked the company's future on the success of the new system. And so, indirectly, he has staked the company's future on Gonzalez. It is an arrangement that has required all of Kashou's faith. It has required a lot of Gonzalez's sweat. And it has required some $200,000. "This system is going to work," says Kashou. "It has to."* * *
Marwan Kashou is not a natural-born computer geek, though he does own a pocket protector (a gift from employees), and he can get characteristically "pumped" talking about the future of E-mail or electronic data interchange. No, whatever this 34-year-old son of a Palestinian grocer knows about computers, he picked up on his own in the past few years. He holds no engineering degrees, harbors no hacker past, in fact, had never so much as laid his hands on a keyboard before 1988, when he learned his first great technological truth: if you are not a student of technology, you are its victim.
Collectech, as its name implies, was intended from the beginning to offer a sophisticated alternative to the bounty hunters ordinarily retained to prompt debtors to pay. Rather than demand a 30% to 40% cut of every dollar collected, Collectech charged a few dollars per debtor up front and promised earlier, friendlier intervention: letters that intervene before third-party collection agencies have to be called in.
Automation was more than a technique for controlling costs: it was a license to exist, the only way Collectech could achieve the volume needed for profits. The company, which rarely employs phone collectors, processes more than 75,000 delinquent accounts every month, about six times what a conventional collection agency paying 40 collectors might turn. Through automated letter writing, it can turn a profit pursuing accounts that other collectors shun. Waste disposal businesses, paging service firms, security companies -- these and firms in dozens of other industries had long carried hordes of nonpaying customers whose low-balance accounts aged so rapidly that the debts had always simply been written off as uncollectible.
Collectech's very first letters were produced a few hundred at a time on a system consisting of a Wang minicomputer that ran some no-frills word-processing and database software, a printer, a burster to cut and trim the letters, and a web of custom COBOL code to keep it all working together. The system cost $122,000. It was also already five years out of date the day it was installed -- a day that happened to arrive some four months later than had been originally estimated, thanks to difficulties in getting the software up to speed. That delay shut down sales and precipitated the business's inaugural cash crisis. It was a rite of passage for Kashou the salesman, who learned that if he had no code, he had no product, and if he had no product, he had no sales. As the Collectech sales jocks put it, "No heat, no eat."
If the original system was antiquated, it was the best the company could do on a start-up budget. Once it was finally running, it sufficed. After all, it had only to fire off letters and keep track of them.
But problems quickly began to proliferate. Data-corrupting disk-drive crashes drove home the importance of hardware reliability. Having to cajole a contract programmer to add features and fix bugs initiated Kashou into the mysteries of software and those who wrote it. He was getting a crash course in computers, like it or not, on the front lines of his business.
The company's sales soared, from $170,000 in its first year to a profitable million dollars in revenues in year three. But the rapid growth only exacerbated problems with an inflexible and not always reliable system. Even when the disk wasn't crashing, it lumbered through queries like a weary nag, whirring endlessly before finding the answer to a request. Because the database behind it was little but one huge file, finding information meant navigating an essentially unmarked forest of bits and bytes. Searches took forever. And every report had to be custom designed by the contract programmer. Nothing could be modified or repaired without him.
When the reliability problems became too frequent and too frightening to bear, Kashou persuaded his partners to invest in transferring the software to a PC network. But the contract programmer never did get the old software to run quite right on the new hardware. Screens froze, forcing users to continually reboot. Memory errors corrupted files, scuttling the day's work. A simple data-entry error could stupefy the system and derail the daily batch processing of letters and invoices. Having to restart the system could shut operations down for an entire day. The company was accumulating strikes with important accounts.
The company's second system also lacked badly needed features, such as accounting capabilities. Service reps still had to track commissions and accounts receivable manually. The software couldn't accommodate credit terms and other specially negotiated service and payment plans that salespeople were now regularly extending to customers.
To cope, the back-office troops created a web of spreadsheets and manual tracking methods. The beleaguered controller reconciled revenues manually each week and juggled no fewer than a dozen spreadsheets to come up with financial statements each month. Meanwhile, the company was bringing in new clients at the rate of two a day. By 1991 there were nearly a million debtor records in the cranky system. And the problems were getting worse.
But Kashou kept putting off a system overhaul. It wasn't until 1992, when sales reached $3.2 million, and profits, after generous bonuses, had neared half a million that executive vice-president Brad Jadwin stepped before the company board to plead for an entirely new system. "The snafus were starting to affect us in the marketplace," says Jadwin. "We were risking client relationships."
The board agreed to bite the bullet on earnings for a year. This time, vowed Kashou, the fix wouldn't be a Band-Aid. "If we were going to take this company to the next level," he says, "we were going to have to rebuild the system from scratch."
The company searched for an off-the-shelf software package, but it couldn't find one that met its needs. The only solution was to bring in a programmer, but this time Kashou was loath to deal with one on a contract basis. "I was uncomfortable that our lives were depending on guys who weren't ours," he says. It was time, he decided, to add an experienced programmer to the payroll.
Kashou knew enough to realize that a new system would rest on the database software he selected, so he took pains to query the programmers he interviewed for their recommendations. Most said he'd have to spend at least $30,000 on a database package just to get started. Then one candidate explained how the company could get by fine with a $5,000 database package to start and then painlessly upgrade it to more powerful versions when the need arose. Kashou hired him, and Jose Gonzalez joined the company. "It was like signing Superman," says Kashou.* * *
Like a lot of software impresarios, the 38-year-old Jose Gonzalez exudes a Peter Pan quality: his dark bangs, his gangly, six-foot-two-inch frame, and an ironic smile all belie 22 years of workaholic code crunching and high-tech company building. "I am a classic nerd," he says, his words inflected with just a trace of his native Cuba. The first thing Gonzalez did at Collectech was to conduct a four-hour seminar on the future of computing. "I wanted to see people's reactions," he says. "Did their eyes glaze over when I talked about this stuff? Would they resist the ideas?"
Kashou remembers Gonzalez's speech clearly. "He was talking about having everything integrated -- the contact management software the sales force used, Microsoft Word, Excel, all running under Windows," he recalls. "This is genius, I thought. It was blowing me away. I couldn't sleep at night thinking about the possibilities."
As a first step, Gonzalez installed Microsoft Mail electronic-mail software. The field reps, who had long been deprived of information and fast communications, took to it so quickly that by the second week Gonzalez had to dedicate a second PC as an E-mail server. Soon the entire company had embraced E-mail, employing it to deliver customer service, communicate sales and income reports, conduct meetings, and even hold elections for employee of the year. Before long, Collectech's E-mail-happy staff had an average of a half-billion characters sitting in the electronic post office at any given moment.
Armed with laptops, addicted to E-mail, Collectech's employees were suddenly getting off on technology. Top managers bragged about their software collections. Nobody had to tell them how to format a disk. Clearly, evangelizing wouldn't have to be Gonzalez's primary mission. If he were going to build the sort of database that would take Collectech to the next step, understanding the business would have to be the top priority.
Gonzalez spent months putting all the company's processes under a microscope. What he saw hardly thrilled him. "Not even Marwan understood how much was being done manually," he says -- nearly all the invoicing and all the financial statements. "People here were compensating for the limitations of the system," Gonzalez noted. "It wasn't really a system." It was more like a dragon to be slain.
Before he could start building a new system, Gonzalez needed to understand each and every task the company performed, both on-line and off. He spent days with the back-office people, who were struggling to prop up the operation. "I sat, I listened, I watched, I walked around, I learned," he recalls.
And he drove his increasingly impatient boss to the edge of distraction. "He wanted to know every bit of information the company tracked and why," says Kashou. "He kept asking me these questions about what we wanted to know, what our customers wanted to know, what questions we'd be asking tomorrow. I kept saying, 'What are you talking about? Let's just build the thing.' "
Meanwhile, Gonzalez was slowly coming to the realization that Collectech was, more than anything else, a selling machine. "Everybody here sells," says Gonzalez. "I'm selling you right now." So Gonzalez started hanging out with the sales force, especially those who pushed the limits of the current system with their custom demands. Harlan Hornbruch in the Seattle office was particularly hard to please; he had earned the title "Mr. Option" for form-fitting services to his clients, and the back office lived in fear of his special requests. Gonzalez made Hornbruch's needs his model.
Collectech's marketers had a grab bag of wants and needs for Gonzalez. Could they call up a graphical cost comparison between Collectech's services and those of competitors during a sales call? Could they whip up and print out a customized letter to the prospect on the spot? Could they enter orders electronically? Could they look up client data when- and wherever they needed it? Could they offer customers an easy way to swap information on-line with the company? Sure, said Gonzalez, they could have it all, and more. The sales reps swooned with anticipation.
Kashou became rhapsodic, too. It occurred to him that with gigabytes of account history at its fingertips, Collectech could become an expert in various client industry niches. "With the right systems in place, we'd become this gold mine of data for customers who wanted to know how they stacked up," he explains. "We'd be in the information business in a big way." He envisioned whole new levels of customer interaction and devotion based on E-mail and electronic data interchange. "Forget taking them to a ball game -- we could bond electronically all day," he says. "We'd have these electronic hooks deep into customers. I thought, man, you do that and you're not just another vendor."
But building those capabilities, it turned out, would take more than the four months originally allotted to the project. Gonzalez had already spent that much and more simply studying the guts of the business. When he finally started showing Kashou prototypes of the system and its features, Kashou figured Gonzalez must be only weeks away. That's when Gonzalez explained how "prototyping" worked: you quickly throw something up on the screen, you show it around to get feedback, and then you redo it and get more feedback. And then you repeat the process 3, or 10, or 50 times, until you've got what you want. Gonzalez wasn't weeks away from finishing, he was months away -- if they were really lucky. More likely, it would take a good year to finish writing the heart of the code. That was perhaps Kashou's most painful lesson: demos were sometimes distant dreams, and warp speed was something that happened on TV.
December came and went. January, February, March passed and still no system. Summer turned to fall, and while much of the code had been written and the new hardware installed -- including a 486 Novell server and a 90-megahertz Pentium-based database server -- Gonzalez still faced the laborious task of testing the system to find and fix the usual myriad of bugs. After that would come the job of moving the data from the old system to the new one. More than a million records would have to be pulled through, along with 2,000 customized sales letters that were being distilled into a few dozen templates. The old system would not go gently.
Still, on the horizon, across the miles of code and hours of work, dangled a powerful payoff: the torquing of Collectech's sales engine to ever-higher revs. "Then we'd be selling, really selling," promised Gonzalez. It was the one thing Kashou needed to hear to help him keep faith.* * *
It is late September, and Marwan Kashou is pacing the worn carpet of a one-room MIS department, preparing to address the troops. CTech Express is late. Again. Testing has already been suspended twice. The bug reports keep coming: the system pulls up the wrong invoices when posting payments; it chokes when printing complex letters; one server fails completely. Maybe there are memory problems. Maybe it's the hardware. No one knows for sure yet. But whatever its source, this, Gonzalez admits, is one tarantula of a bug. "Printing letters -- it's only our business," he shrugs.
Testing and debugging the new system competes for Gonzalez's time with the demands of propping up the old system, which requires periodic triage. Gonzalez has temporarily abandoned hopes of adorning the CTech Express with fancy features to focus on simply getting a modified model of it running. Deadlines, revised and rerevised, have been missed three times already. "It's a little scary," he says. "We're looking at the numbers for next year, and all those projections for sales and production are predicated on this new system being up and running. We know what we have to do. It's just that we have less than three months to do it."
To complicate matters, the quarter has gone soft. Revenues will fall short of projections. Costs outpace estimates. An exasperated Kashou, needing to inspire Gonzalez and his team even as he admonishes them, looks for the right words.
"We need to get extreme," he tells them. "This is like a K-2 expedition. We need all the horsepower we have, and guys, we need it now." Kashou informs Gonzalez that all the resources of the company are now being dedicated to the system conversion. He'll sign whatever check Gonzalez tells him he has to.
Two more people are added to Gonzalez's team. And Kashou hires a software consultancy to finish the development work on the program designed to link customers electronically. No, he's not losing faith, he insists: "I'm just getting real." But his growing fears that Gonzalez might be leading the company off a technological cliff brings Kashou to seek second and third opinions. "I've got to continue to drive this process," he concludes. "It doesn't matter if I don't know what happens outside in sales for a few weeks."
Yet Kashou worries that the company is losing its credibility with its sales force, which has been eagerly selling the benefits of the new system to customers. Now customers have begun to clamor for the ease of electronic transfer held out by CTech Express.
Ten days later, Kashou gets to test-drive the new system. He's elated with what he sees. "An invoice never looked so good to me in my life," he says. An employee who also tests the system E-mails Kashou: "This is better than meeting Christie Brinkley." Kashou fires back: "CONCUR!!!"* * *
One October morning, a regional sales representative telephones Kashou to announce that the company has just snared the largest account of its life with a huge telecommunications company.
Sales will swell with the new, $2-million deal. But so will production volume. The account will introduce another million records into the system, enough to overwhelm the low-end database serving as the starter engine for Collectech's new system. "We've outgrown it before we ever got it up and running," laments Kashou.
Gonzalez and Kashou agreed they couldn't afford another premature brush with obsolescence: "We had to get a product that was proven and that we wouldn't blow out of in less than a year," explains Kashou.
Data conversion and testing have to begin all over again. But another transformation, one that had been slowly taking place over the past year, is consummated by the crisis: the programmer is becoming a pragmatist, the CEO a systems expert. Kashou, once the hard-nosed minimalist when it came to automation, is letting a long-term vision of his system guide his investment decisions. Gonzalez, once a purist who preferred programming elegance over expediency, is now looking for a sure bet that won't batter the company's bottom line. The two, both driven and stubborn, are finding that the common ground between them suddenly seems more vast than their differences.
The two quickly agree on a plan. They decide to bet on the newest release of a Sybase database, which promises 32 terabytes of storage, or about 16,000 times that of its predecessor. The cost: $20,000, or nearly as much as Kashou had been advised to spend on the entire system.
Kashou is now seeing a chiropractor to ease increasingly intense muscle spasms in his back. His expectations that he might ever finish running this race, much less win it, are dashed. The objective now: to simply keep going. "It's never really over, I guess," he says.* * *
It's January 1995, and CTech Express is up and running. Does that mean the system meets all the company's needs? Kashou knows better than to ask that question now. As with a programmer's loop, in which a sequence of software instructions repeats itself over and over, Collectech seems destined to continually retrace its steps toward the perfect system.
Despite the 18 months of sweat and setbacks, despite the $200,000 price tag, Kashou and Gonzalez refuse to profess discouragement with their now-mutual fate. "You have to be prepared to take your hits and keep moving," says Gonzalez. Adds Kashou: "Nothing comes cheap or easy or on time. But it takes pressure to make diamonds, right?" Or to put it another way: no heat, no eat.
POWER TO SPARE
If CTech Express had a fender, Jose Gonzalez would polish it with pride. "This thing can beat the heck out of any minicomputer," he says with a satisfied smile. His hard-won system, a multiserver-network that was more than a year in the making, takes up less space than a coat closet. But these days some pretty heavy-duty power can come in small packages.
The boxes housing the main processors that crunch the company's data stand in tight formation along the floor underneath two adjoining desks. The Novell server holding the company database is a 486 machine with 32 megabytes of main memory and 12 gigabytes of on-line disk storage -- about 20 CD-ROMs worth. That's where the database resides. Connected to it by a 100-megabit line is a database server, a $3,000, 90-megahertz Pentium-based machine, dubbed Alice, that routes information in and out of the database. Further turbocharging the system are four $2,300 Pentium machines, which divide and share computing tasks, such as whipping up and printing out a letter. Marketers can phone in from anywhere using modem-equipped 486 color laptops to download client data from the system.
Not everything in CTech Express is state-of-the-art. Three communication servers, one 286 and two 386s -- old warhorses by the Pentium standards of today -- support a client bulletin board. More than 300 clients now dial into the board, using off-the-shelf communications programs such as PC Anywhere, to upload account information or to download reports. Perched above the servers are a few worn keyboards and a bank of six mostly beat-up black-and-white monitors. "These monitors have gone through hell. They're old junk," says a bemused Gonzalez. "But that's all we need to do the job."