It took 37 years before Kevlar – a bullet-proof, fire-resistant material first used for tires – was applied to making home shelters strong enough to resist tornadoes. It took decades before advances in reinforced fiberglass technology made for the Apollo space project were applied to the making of tennis rackets.
In retrospect, these crossover applications of technology may seem inevitable – but they are not. According to Ian C. MacMillan, director of Wharton’s Sol C. Snider Entrepreneurial Center, crossover applications depend largely on serendipity. “You have to hang around for 15 years for someone to make the connections,” says MacMillan. In many cases, the serendipity never happens – and technologies die on the vine before they achieve their full commercial potential.
To overcome those problems, MacMillan, along with Wharton professor of operations and information management Steven Kimbrough and John Ranieri, vice president of the bio-based materials business at Du Pont, have developed a patented process that will help companies analyze databases of information about technologies and suggest new markets where they might be commercialized. “We have a serendipity generator,” notes MacMillan. “Serendipity happens every now and then, but this process reiterates the connection.”
Why is serendipity such a rare commodity? “The fundamental problem is that technologists know nothing about markets, and markets know nothing about the technologies,” says MacMillan. “It is like a black hole. It is very hard to see behind your experience space.”
For example, the initial developers of Kevlar weren’t thinking about a market for protecting people against tornadoes; it wasn’t their job to develop products for storm protection. Only rarely are companies large enough to have the resources and knowledge to explore new crossover markets using technologies they developed in-house. One such company is Du Pont, which developed Kevlar in 1965 and extended its use to several new areas – including bulletproof vests – and into a new residential storm shelter that it unveiled in March 2003.
But even companies like Du Pont don’t have the resources to probe into every possible application of their current roster of technologies. “Exploration is very expensive,” says Ranieri, and consultants and researchers are so costly that “you start with what you know and you are not going to be surprised [by the results].” Ranieri, who met MacMillan when he participated in the Wharton Advanced Management Program a year ago, later teamed up with him and Kimbrough to develop this new process.
Silos of Specialized Information
Although an unprecedented amount of information about technology is now available online, Ranieri notes that “everything is set up to look for exactly what you are looking for” rather than to assist in the process of finding crossover, innovative applications. In addition, information “is stored in silos” that are hard for non-specialists to penetrate. Until now, there has been no way to search for attributes like “lighter, faster or quicker” with technology categories, he says.
The Wharton team’s new process aims to meet this challenge by using a methodology that “combines computer research techniques with human research techniques,” says MacMillan. Kimbrough likens the new process to the methodology used by Google, the popular search engine. Although Google is automated, it exploits information painstakingly collected by thousands of individuals (at no cost to Google) and loaded onto their web sites. Kimbrough explains that Google’s page-ranking algorithm “exploits tons of work [done by] people who put Java links on their web sites; it exploits their manual labor.” Like Google’s algorithm, this new process “is not an intelligent parser and it is not artificial intelligence,” Kimbrough says. “Like Google, we exploit the public information that is available.”
Another key problem with Google, today’s most powerful search engine, is that it can only identify characteristics that match up perfectly with key words that people use to search for information. It can’t infer any attributes that aren’t precisely spelled out and loaded onto a web site. Moreover, it can’t make subtle, creative connections between descriptions and attributes, and it can’t find information that is hidden in the text.
“There is a lot of information in texts [on technology] about what stuff can be used for,” notes Kimbrough. “A document is like a bit of DNA. There are discernible patterns that can be very subtle.” Unlike traditional products, the Wharton team’s new process searches through documents and makes connections between highly technical descriptions of properties – often familiar only to narrow “silos” of technologies – and broader terms that could suggest market applications to those who work in other areas. As Ranieri describes it, “We found a clever way to make a link between attributes and markets.”
Kimbrough likens this to the process that savvy poker players use when interpreting the inadvertent signals given out by their opponents. Just as good poker players read subtle facial expressions and bidding styles, this new process searches through characteristic patterns in the text to find hidden meanings. The goal, Kimbrough says, is to reveal information that the authors might not have intended to give out “or didn’t necessarily know.”
Although it’s too early for the developers to discuss technical details, Kimbrough acknowledges that this new process requires a significant amount of human input. “In part, we use human beings to create databases of attributes that can be matched up.”
Turning a Negative into a Positive
To illustrate how this can work, Kimbrough cites the case of thalidomide, a product that made headlines decades ago when it was prescribed – with disastrous consequences – to pregnant women for treatment of morning sickness. Recently, some cancer researchers have become fascinated by a characteristic of thalidomide that helped make it dangerous – its attribute as an “anti-angiogenesis” drug. More simply put, the drug prevents the creation of blood vessels – obviously a very negative attribute when administered to pregnant women. However, as Kimbrough notes, one way “to stop cancer is to stop the creation of blood vessels.” So now, with anti-angiogenesis drugs possibly in demand, this negative attribute of thalidomide could make it a valuable tool in fighting cancer.
What if a process had been available decades ago to sift through databases of information about thalidomide and analyze its attributes in such a way that technologists and marketers in other product areas – such as cancer research – might have seen its potential as a crossover application? The general issue, notes Ranieri, is that in the absence of new tools that find hidden connections, no one can fathom all the possible crossover applications. But “if you can map those attributes [and match them] to needs, you change the whole game.”
The Wharton team’s new process also tackles another key obstacle to crossover applications: Different groups of technologists use different sets of terms to describe attributes of their products. The terminology used by aeronautical engineers to describe fiberglass, for example, is not necessarily the same as the terminology used by developers of tennis rackets to describe those properties of fiberglass that are important to their products. Explains Kimbrough: “The attributes are not likely to be the same attributes” and “so [our] process must figure out a match.” Thus, while consumers might think of “flexibility” as a property of fiberglass tennis rackets, technical documents about fiberglass may include a vast range of technical terms but not a simple English word such as “flexibility.”
Testing Real-world Results
To demonstrate how this process gets results, MacMillan’s group approached a Fortune 10 company and performed two sets of successful tests. In the first set, MacMillan studied what the company had developed in the past and showed how this new process could have generated crossover uses of old technologies more rapidly. “We told them, ‘Show us the products you developed over 20 years and we will show you how we are doing the job quicker,’” he says.
Having achieved that goal, MacMillan’s team did a second series of tests that demonstrated how the process could generate new ideas for crossover application of current technologies, including some “that they may not have thought of.”
Now that their patent has been accepted, the team is preparing to convert their prototype methodology into a full-fledged commercial process. MacMillan describes the approach this way: “Companies bring us their patent, and we take it and tell them 50 places where the technology underpinning the patent might use it.” If all goes well, companies could eventually be licensing the use of this commercial process and be running it on their own.
Or the new process could be made available online to those willing to pay for it. The hope is that even companies with significant in-house resources will use this tool as a short-cut for stimulating creative new ideas without spending a lot of money. “Everyone will use this in a different way,” says Du Pont’s Ranieri. “Everyone has different needs.”
The Wharton team acknowledges that this process is not a magic bullet nor is it intended to be a foolproof generator of sure-fire hits. “We are not saying that all of these ideas will work,” says Kimbrough. “We are saying that we will get a lot of good contenders that a human being could go through.” Moreover, “there is no guarantee that the product [that results from the best ideas] will be a success.” Even the most creative, innovative product ideas still must be manufactured at a competitive price, be marketed effectively, meet necessary regulatory requirements and so forth.
For all that, Kimbrough says, “this process can be a useful step in generating ideas for innovative applications and crossovers of technologies to new areas.” Adds Ranieri: “It is for big companies like Du Pont and also for small companies. Anyone who has an entrepreneurial drive will love it.”