Unschooled in the ways of commerce, scientist Michael Kuperstein finds himself the owner of a technology with unlimited applications. Now all he needs to do is choose the right niche

Between the business plan and the marketplace lies the reality, as Michael Kuperstein has discovered since opening Neurogen Laboratories Inc. in January 1988. When idea meets customer, things change. Fate intercedes. Markets open, close, and shift. All this surprised Kuperstein, who never intended to turn his grant-funded research operation into a business in the first place, let alone worry what his key markets would be. Those intentions didn't last long.

Kuperstein, then 34, with a Ph.D. in brain science from the Massachusetts Institute of Technology, had motives more academic than mercenary in starting Neurogen Labs. He couldn't find a good tenured job at a university in the Boston area, so he set up Neurogen in Brookline, Mass., as a private institute to further his research on neural networks -- computer solutions based on how the brain works. Conventional computers are programmed with expertise that lets them perform specific tasks. Neural network computers, meanwhile, are programmed to learn from experience and figure out the correct course of action -- much as humans use their brains.

Soon after Kuperstein got the lab up and running in early 1988, the angst started to surface. "My first thought was, 'Oh, no, what do I do now? How am I going to survive?' " When would the grant money dry up? As Kuperstein got deeper into the research, a way out presented itself. "Gee, there's a lot more here than I first thought. The opportunity to commercialize what I was doing was tremendous." It was time to think not just about science but about business. In August 1988 Neurogen grew a mercantile arm, Neurogen Inc. Kuperstein suddenly found himself an entrepreneur -- albeit one with a problem. Owner of a proprietary technology with seemingly limitless applications, he had to choose which to pursue. Where was Neurogen's market?

Kuperstein had to find a customer.

Neural networks have applications in four basic fields, two of which -- robotic control and pattern recognition -- were options available to Neurogen. In robotic control, neural networks guide robots, making them more versatile in a factory setting because the robot can actually learn from its environment and modify its behavior. In pattern recognition, neural networks use computer imaging technology to identify anything from targets to tumors.

For Kuperstein, this is where the road first forked. If Neurogen were to survive as a business, he had to make a choice and travel down one path or the other.

Kuperstein chose pattern recognition. The choice might seem odd. The year before, Kuperstein had developed the world's first neural network robot. He had garnered a lot of praise and press for doing so. But by the time Neurogen was up and running, the robot wasn't. It was still a prototype. Getting the robot into production would take more time and money than Neurogen had. Even though his heart lay with the robot, Kuperstein knew he could get to market much faster with a pattern-recognition product.

But what exactly would that product be?

Kuperstein had no idea. Instinctively, he set out to copy another computer imaging company's product. He would create an electronic stylus with which stock traders could record transactions. The nation's stock exchanges would be his market. But in the midst of his research a serendipitous, eureka sort of moment interceded, presenting Kuperstein with an application that others had yet to crack -- the recognition of handwriting by a computer.

Within two weeks after that revelation Kuperstein had completely refocused Neurogen's mission and written a program that could recognize four characters. He was on his way to devising a computer program that could read handwritten numbers. Kuperstein knew this could be done in a relatively short period of time with a limited amount of money. Kuperstein stopped thinking about styluses and the stock market.

By April 1989 Kuperstein had a production prototype of a product called Inscript. He also had added two more employees. Arthur Gingrande Jr. became Neurogen's vice-president of marketing and sales. Jerry Fisher became vice-president of product development. Neurogen Inc. was formed with Kuperstein, Gingrande, and Fisher as the company's founders. Betweeen June and February the three of them raised $200,000 privately to keep the company going.

Inscript is a deceptively simple-looking device, a collection of off-the-shelf semiconductors mounted on a PC board along with Neurogen's proprietary software. This simple array, however, masks the considerable brain work programmed into the software. Added on to any IBM-PC/AT based scanning technology, Inscript can read virtually any handwritten digit at the rate of 15 characters per second. It can do so as long as that digit is "segmented," or not touching another digit. It can also read dashes, periods, and commas.

Now, how could the company begin to find its niche? Where were Inscript's customers? The problem for Neurogen was they were everywhere. The company -- unknowingly at first -- had stepped into a megamarket, the forms-processing industry.

Americans write 50 billion checks a year. They charge everything in sight, each time attaching their signature to a telltale set of numbers. They fill out countless forms -- medical, insurance, taxes. They annually send 170 billion pieces of mail to one another, each bearing a simple five-or nine-digit zip code. This torrent of paper needs to be processed by someone.

The companies this unenviable task has fallen to have come to rely on an army of key-entry operators, drudge workers engaged in a dreary task. For them the tedium is unrelenting. For their employers the cost of all this labor is equally onerous. They would like nothing better than to cut the human being out of the paper loop.

Neurogen's product had manifest appeal. For some 25 years giant technology companies had been laboring to solve the riddle of handwritten character recognition. Now, Neurogen, a comparative gnat, had seemed to do so with Inscript -- and at a reasonable price. The PC board, the software that drives it, and the necessary support from Neurogen cost a client $60,000. Volume users need more than one board, but they also get volume discounts. As a general rule, buying Inscript entails a onetime, nonrecurring expense that, Kuperstein claims, a user can earn back in four to nine months, depending on the volume of data processed.

It fell to Arthur Gingrande to open up the market in the spring of 1989. "We figured that if I showed this to enough end users we'd find out where the demand was," he recalls. "We wanted to let the market tell us what we ought to sell. Then we'd react. It was a case of dynamic positioning."

The company had no precise list of who would want Inscript and in what order. "There was no market research available on this," says Gingrande. It did, however, foresee seven large markets into which it could sell: banking, insurance, sales ordering, credit card companies, mutual funds, the U.S. Postal Service, and the Internal Revenue Service. Getting a foothold in these big industries seemed impossible, given that Neurogen was little more than three guys in a second-story walk-up office.

As Neurogen set about getting into the market, it imagined that the banking industry would offer the "the quickest entry point," for two reasons. First, banks had the necessary volume. They process checks by the carload. Volume translates into huge labor costs. Neurogen could clearly save banks a lot of money.

Second, the banks by dint of their size and needs also had another critical characteristic. They had invested in the necessary "enabling" computer technology that Inscript needed in order to integrate into the banks' systems.

Yet for all that recommended the banks to Neurogen, they lacked a certain intangible -- the willingness to take a risk. Recalls Gingrande: "They would say, 'IBM is already taking care of this for us,' or 'Banctec already has the problem well in hand.' I'm not sure they really understood what we had."

Undeterred, Neurogen shifted its focus. It pitched those banks' vendors, the systems integrators and value-added resellers that sold computer solutions to the banks. "They were all interested in our technology," says Kuperstein, "primarily to see how far along we were. And by the nature of the questions they asked us we could deduce how far along they were." The answer, Kuperstein believed, was not far.

Neurogen then went back to the banks to put pressure on them through their vendors. "We told them we could deliver what they wanted today," says Gingrande. This time the banks, with their vendors' endorsement of Neurogen, took notice. Their interest was further piqued by a very favorable review of Inscript in a leading industry newsletter that had appeared in the interim.

But then a technical problem intervened, one that Neurogen had yet to allocate much of its scarce resources toward solving. When people write a dollar amount on a check they often crowd the numbers. They fear that someone else might insert an additional digit. On checks, many handwritten numbers touch. In the jargon of the pattern-recognition business, they are "unsegmented." No machine has yet been able to pull two overlapping numbers apart and identify each of them fast enough to satisfy customers. "We were really naïve," recalls Gingrande. This rude discovery confronted Neurogen with a hard choice. It could plow its limited resources into solving the overlapping character problem, and thereby crack the primary and hugely promising check processing market. But doing so meant neglecting other, more accessible markets. The alternative was to develop other markets and put banking on the back burner until the company had more resources at hand.

That's what Neurogen did.

"It took us a long time to find out a simple fact," says Arthur Gingrande. "The attitude a person has about a form determines how legibly he will write on it." The banking experience got the founders at Neurogen to debating a now obvious question: if check writers were paranoid, then who, when it came to writing numbers on a form, was not?

The answer came at an electronic imaging trade show in October 1989 when Neurogen and a large sales-order entry company -- which was looking for a faster way to process its forms -- bumped into each other.

When Kuperstein, Gingrande, and Fisher studied the sales-order forms, they liked what they saw. The forms had a separate box for each character. The numbers had to be segmented. They realized that salespeople had the opposite motivation of check writers. They need to write clearly so the order gets entered correctly and they get paid a commission.

In the order-entry company, Neurogen had a live contact. A development contract between the two companies resulted. Neurogen, in return for a fee, would try to customize Inscript so it would integrate with the larger company's technology. In the course of the work, a potentially lucrative secondary market opened up for Neurogen.

Neurogen's partner in the form-reading effort, it turned out, also printed and corrected standardized tests. With Neurogen's help, it could redesign the answer form -- and potentially remake the standardized test industry. It made little sense to have rows of five small spaces arrayed across the page, with the student marking just one. If a system that could read a handwritten number was available, it would be simpler to have the student write one of five numbers in just one space. This would save paper. It would also diminish every student's nightmare: recording a page full of wrong answers by improperly aligning questions and answers.

Banks had seemed a hot market and had not immediately panned out. Sales order-form processing was down the list and standardized tests scarcely conceived of in Neurogen's first business plan; yet those markets suddenly seemed full of promise. The U.S. Postal Service, meanwhile, fell somewhere in between.

Neurogen had always considered the post office a potential bonanza. One hundred seventy billion pieces of mail go through the system each year, each with a five- or nine-digit zip code. Letter writers tend to be more like salespeople than check writers. Write the zip code neatly -- no numbers touching -- and the letter will get there. Technically, this market seemed within reach.

But as promising as the post office seemed from a volume standpoint, it was positively daunting as a bureaucracy. It was an institution wrapped in red tape. Dealing with the post office meant a long ramp-up time on any project, something a small company like Neurogen could ill afford. On top of that, the post office has three different sections that see to its various procurement needs.

Neurogen decided it couldn't afford to penetrate the post office on its own. Its strategy would be to find an enabling partner -- a bigger, savvier company that already serviced the post office. Neurogen could ride the coattails of such a systems integrator.

Through a chain of contacts, Neurogen found Scan-Code Inc., a systems integrator that sold systems to companies that put bar codes representing nine-digit zip codes on first-class letters. Two years ago Scan-Code didn't exist. The company had formed quickly after Congress passed legislation privatizing parts of the postal system. Under the new law, the government paid private companies 4¢ for five-digit zip codes and 5¢ for nine-digit zips it pre-sorted.

Introduced by Scan-Code, Neurogen got a two-hour meeting with a postal rate commissioner and an assistant postmaster general and got its chance to show off the technology. The post office gave Inscript 2,000 handwritten zip codes to read. Inscript rejected 30% for being mechanically unreadable and in need of human handling. Of the remaining 70% it read 99.5% correctly. The newer, faster Inscript is capable of reading 80% of five-digit codes, and of that 80%, 99% would be read correctly. Scan-Code and Neurogen now have a development contract to work on mail sorting solutions.

Another market presented itself when a year ago a major credit card company announced an effort to upgrade its technology in order to read handwritten characters. After a year spent probing the bureaucracy of this large company, meeting with one vice-president after another, and demonstrating its technology, Neurogen landed a development contract.

In the process, Neurogen kicked over another possible market. It assumed that the technology available for reading typewritten characters -- optical character recognition (OCR) -- was mature and adequate. That was not really true. New typefaces, photocopies, laser printing -- all resist recognition by machine. So do the imprints made by the embossed numbers on credit cards after they have broken down with use. Neurogen's technology can read this "degraded" type. This submarket, called OCR cleanup, is one that Neurogen had totally ignored at the outset. Now it accounts for about one-quarter of its business.

Neurogen now has seven employees. It added two software engineers in March and hired a chief executive officer, Robert Tabor, in April. Tabor's job includes honing the company's mission and revising the business plan. He will try to determine which leads thus far are promising, which are dead ends.

In rethinking what its business should be, Neurogen has recently targeted six major market niches it wants to fill. These are, in order of their current accessibility: forms processing, OCR cleanup, credit-card processing, the mail, remittance processing, and check processing.

Next, the company identified the enabling technological developments that would have to occur before it could penetrate each niche. These developments were then put on a time line, so Neurogen could guess what niches would open up in which order.

This process accomplished two things. It made it easier for the company to allocate its scarce resources in solving various technical problems, and it helped determine how much capital Neurogen would need in order to meet the timetable.

By this fall the company expects to have raised a round of financing from venture capitalists totaling between $1 million and $3 million. "We need a cushion," says Kuperstein. "Right now we're in a position of strength. We can raise money on our terms, not theirs."

In the meantime, his company will begin generating sales revenue. Neurogen, which made its first sale and installation last February, projects sales of roughly $800,000 in 1990. The company thus far has prospects for 23 separate installations. In 1991 Neurogen expects sales to grow to $6 million. Kuperstein bases this figure on having 5 to 10 volume customers placing orders of anywhere from $500,000 to $2 million each. That projection works out to a gross of between $2.5 million and $20 million. Kuperstein reasons that $6 million is appropriately conservative.

Those sales, a number of them to Fortune 500 companies, will create a critical mass and give his company legitimacy, believes Kuperstein. Other large customers will then surface. Sales in 1992 will jump to $20 million, $40 million the year after, and $75 million the year after that. At $75 million, Neurogen expects to have between 30 and 40 customers placing orders of between $1 million and $2 million.

Kuperstein will not disclose the company's profit margin, other than to say that at the outset it will be "very large." He bases this on the belief that Inscript is a unique product for which there is considerable demand. "Our pricing policy is based on the labor savings we will provide." Currently, the cost of each document processed by a human in a large company ranges between 0.5¢ and 1.5¢. When you're a big bank or credit-card company, that adds up fast. Kuperstein has based Neurogen's financial projections on the experiences of two hypergrowth companies, Microsoft and Sun Microsystems. The cost of sales in software product companies such as Microsoft is quite low, so gross profit typically exceeds 50%.

At $800,000 in sales, Kuperstein allows, "we're breaking even." He adds that in this first year of sales, Neurogen's major cost -- engineering and manufacturing -- will amount to only between 12% and 15% of gross sales. As time goes on, competition will doubtless intensify and margins will come down.

Hiring a CEO will allow Michael Kuperstein to assume a role, along with the chairmanship, that he's more comfortable with, that of chief technologist. Much of his time will be spent doing research -- in order to assure his company a future. Kuperstein envisions the research side of the business, Neurogen Labs, growing to about 10 researchers and forging a symbiotic relationship with Neurogen Inc. Neurogen Labs has already received seven grants totaling $400,000 from government agencies to further its work on neural network robots. Kuperstein sees licensing its inventions to Neurogen Inc., which will then reciprocate with royalty payments -- which, in turn, will contribute to the funding of further research.

Kuperstein doesn't worry about falling behind on the technology -- so vital to keeping the company from being a one-product phenomenon. He is an editor of the two best-known journals on neural network technology. "I know the field. I'm at the pulse of the community."

"Michael's famous," says Arthur Gingrande. "He's one of the top people in his field. We have drawers full of résumés from people all over the country who want to work with him."

Kuperstein projects an assurance that borders on arrogance. While he knows of perhaps 30 to 40 small companies in neural network technology, he is not worried. He says, in fact, they are off the mark. Almost all, he claims, are providing "tools" -- neural network technology. Neurogen, by contrast, says Kuperstein, offers "solutions" -- the harnessing of neural network technology to resolve specific problems.

As for the goliaths, Kuperstein isn't worried about them, either. He plans to get far enough ahead of them that they will be moved to forge strategic alliances with Neurogen, not squash it. At the moment, Neurogen seems to have that edge. At a recent imaging trade show -- the largest one of the year -- Neurogen shared a booth with Citicorp. Both Kuperstein and Gingrande found it mildly embarrassing to find people flocking to their display booth, oftentimes leaving representatives of Citicorp with little to do but twiddle their thumbs and stare into space.

That kind of attention leaves Michael Kuperstein, an academic with no prior business experience, brimming with confidence -- something he doesn't lack to begin with. "When people see our demonstration they're blown away," he says. He believes he has something the world wants. Does that mean someday soon he sees himself presiding over a company in a large, gleaming building with hundreds of employees?

Kuperstein answers that one without a moment's hesitation: "Yup."

Stay tuned.

-- Research assistance was provided by Leslie Brokaw.



Neurogen Inc., Brookline, Mass.


Develop and market a technology that enables computers to read handwritten numbers


Sales of $800,000 in 1990, $20 million in 1992. Breakeven in first year, aftertax profit of 11% in 1992.


Establishing a presence in the product's large, bureaucratic markets, among them banking, credit-card companies, and the U.S. Postal Service; lack of market research
Michael Kuperstein

Last job: Academic research

Equity: More than 50%

Salary: $45,600

Personal funds invested: More than $50,000

Goals: Take Neurogen public in about five years. Reach $130 million in sales after five years.

Why started the company: "To escape the political and practical constraints of academia and find a more creative outlet for my talents. Commercializing research is the only way the tremendous potential of this technology will be realized and people will fully benefit."

Neurogen Inc. Projected Operating Statement
($ thousands)* 1990 1992

SALES $800 $20,000

Cost of sales 280 7,000

Research and development 160 2,400

Selling and general & admin. 200 3,000

Customer support 160 4,000



Taxes 0 1,400


% profit after taxes 0 11%

* Inc. estimates based on Neurogen forecasts and market information



Managing partner, Advent International, a Boston venture capital firm managing $1.3 billion worldwide, with early-stage investments in high-tech companies

Looking at this deal is like evaluating a balance sheet: there's always some good news and some bad news.

Kuperstein seems to be approaching the organization of the company the right way. He's hired key operating people, and he's putting together a nucleus of development people that can attack more markets with this product line. He's smart to recognize that he is a technologist and hire a CEO to run the company.

But these guys are clearly chasing too many markets. What's happened -- and this is not atypical -- is they're being driven by the technology. They've got to get this down to one or two, maybe three markets, not six. They're trying to be too big too soon.

They've got to utilize good market research, and if they don't have it they should get it or hire expert consultants. A well-conceived product marketing plan would save them a lot of time and capital. It appears that they've been fooling around for a couple years looking for a market -- a classic problem.

If I were investing, I would have the company concentrate on selling through the integrators and the value-added resellers; it's a good approach for this company. The product seems like it would be difficult to sell on a direct basis. The integrators and resellers would enable Neurogen to provide a focused product to the market, as opposed to technology looking for a market.

The banks, the credit card guys, the check processors, and the post office don't want to deal with a mom-and-pop company. They want to have somebody install a system, keep it running, and ultimately bring in a new generation of product. A small company can't provide this security blanket.

Going from zero to $6 million to $20 million to $75 million is wildly optimistic. Also, basing projections on Microsoft's success is not realistic. Microsoft had significant revenue and profits because it attacked a very fast-growing, broad-based market -- it had IBM in its camp. These guys don't.

I think the company will succeed, but it's going to take it longer and require more capital than Kuperstein thinks.


Vice-president for new-business development, HNC Inc., San Diego, a $10-million manufacturer of application products that use neural networks

Most of the applications they've targeted are operations oriented, and those people tend to be hard-nosed and conservative when it comes down to buying anything. That's bad, because I think this is a young company selling a lot of sizzle while other people are selling steak -- solving customer problems in very difficult operations areas. Making broad claims without being able to back them up with something that can be used immediately at the operations level might cause Neurogen a lot of trouble.

I think they've underestimated the character segmentation problem. It's not an area you can plug an engineer into for a couple weeks and get a solution. IBM has been working on it for about 25 years, using a lot of resources, and hasn't solved it.

If they can solve the segmentation problem as well as the character-recognition problem, then they can package that into a product that is very high value added. But as things stand Neurogen is offering only a piece of the puzzle. In making his growth forecasts, Kuperstein likens his company to Microsoft and Sun Microsystems. Well, they were offering the lion's share of the total solution. Neurogen right now isn't.


Senior vice-president, 3i Ventures, a venture capital firm in Newport Beach, Calif.

Kuperstein's attitude toward venture capital, and how appealing the company will be to VCs, seems slightly cavalier. I suppose it depends on how proprietary the technology really is, but technology is moving very rapidly in this world, and they could get leapfrogged by a competitor. And there's a market acceptance risk: they're attacking large, high-volume markets with cost-sensitive applications, and that's a sell that will take some time. So I don't think it's as close to a slam dunk as the founders seem to think.

Bottom line, I think the company's managers have a dilemma. I got the sense they would like to bootstrap the operation -- not raise too much capital, not bring on too many people, not build up the infrastructure, not give up a very large percentage of their stock. If that's the case, then they really need to focus their attention on one market, one application, and carve out a very strong competitive position in that small niche area, at least initially.

I think each of the six specific markets is probably perfectly legitimate. If anything, I would worry that they will not be easy to track. People may be quick to buy one system to test it and see how it works, but the real test for the company will be whether it can sustain itself until customers start to make multiple purchases.


Managing partner, Price Waterhouse's Entrepreneurial Services Center, Cambridge, Mass., a provider of business advisory services to start-up and early-stage companies

The good news for this company is in the recent events, the hiring of a CEO so Kuperstein can stay on the technical side of the business. The CEO needs to be a real business planner and a good tactician. These guys need to recognize that they can't just go out and develop this technology and see who wants to buy it.

After all, they are not only trying to go out and compete. They are on a crusade as well. They have to convince huge customers, who are ingrained in their ways, that what they're doing makes sense. The best way to do it would be to convince an intermediary, a systems integrator, to work with them. They'd have to give up margin, but it's worth giving up some margin to get the business an intermediary can bring.

But they're still doing market R&D. They've let two years slip by, and they still don't know what their market is. Making the discovery about banks and the problem with characters touching when they did is such a fundamental mistake. And the niches they've identified are so huge. It's a misnomer to call these markets niches. They need to focus. Instead of trying to do six different things, they need to stop and say, "This is the one place where we think we can make a dent in the market" and go for it. If they don't, the lack of funds, given what they're trying to do, will kill this company.