Inspired by his father’s unsolved murder, Austinite David Wheeler wrote a killer computer program that helps cops catch crooks.
On May 27, 1981, millionaire Roger Wheeler was gunned down gangland-style while walking to his car at a country club in Tulsa. Wheeler, 55, was the president and CEO of Telex Corporation, a computer hardware company, and had interests in a magnesium plant in West Texas, a jai alai operation in Florida, and a host of other businesses. The murder made national headlines, but no one was charged in the case, and it remains open to this day—to the great dismay of Wheeler’s son David, who at the time was working for his father as a computer consultant. From the moment he heard the shocking news, David became consumed by the crime. He kept in constant touch with the Tulsa police and appeared on the first episode of TV’s Unsolved Mysteries.
By all accounts 46-year-old David is still consumed—“He’s a man with a mission,” says Tulsa police detective Mike Huff, who headed the initial investigation and is still in charge of the case—though these days the scholarly looking redhead seems almost reluctant to tell the story of the murder. When pressed, he’ll offer details that are fresh to him even today; he can describe the wet clothes his wife pulled from the washing machine and tossed into a suitcase when they rushed to catch a plane to Tulsa after learning that his father had been shot. But Wheeler opens up when asked about his brainchild, InfoGlide, an Austin software business that helps local and federal authorities fight crime and businesses combat fraud. The company was inspired by Wheeler’s desire to see his father’s killers behind bars—“I guess you could say we made lemonade out of lemons,” he says—though investigators in the case have never used his software, since it works best on recent crimes. Still, his ultimate goal is far more ambitious than nabbing a single perpetrator. “I want to catch more crooks than anyone else,” he says flatly.
Catching crooks by computer is the latest trend in law enforcement: Austin’s police department, for example, has a computer-savvy high-tech unit devoted to crimes plaguing high-tech businesses. Traditionally police culture has favored what former San Jose police chief Joe McNamara calls the “NYPD Blue mystique”: crime fighting with a combination of intuition and good interview skills. “It’s much sexier for a chief to talk about putting more police on the street than getting new technology,” says McNamara, who created a high-tech division in his own department. But with many detectives in many cities overwhelmed by caseloads and reams of data, computers have become a logical place to turn.
That’s where InfoGlide comes in. Since founding the company in 1991, Wheeler has developed two basic programs now in use around the country. One is the Detective Toolkit, a software program designed to enable investigators to identify serial offenders, from armed robbers to rapists to killers. The Toolkit is already being used by the Tulsa Police Department as an integral part of its investigations and by military intelligence investigators, who want help in tracking terrorists; it’s also being tested by the U.S. Justice Department’s Regional Organized Crime Information Center in Tennessee, which is creating a special unsolved homicide and missing-persons database. And InfoGlide markets another software package called Fraud Investigator, designed to detect and prevent insurance fraud; it’s currently being tested by several major insurance companies.
InfoGlide is located in the Austin Technology Incubator, a launching pad for new high-tech firms sponsored by the University of Texas. “This place is my heaven,” says Wheeler, his outspread arm indicating his suite of offices in the rather clinical setting of the MCC building, whose west wing the incubator occupies. Wheeler’s work is supported by a venture capital firm and by a technical and management team, including Canadian computer programmer Paul Leury, who developed a program for the Royal Canadian Mounted Police to help catch violent offenders. Last year the system developed by Leury, called ViCLAS (Violent Crime Linkage Analysis System), was credited with identifying the killer in a notorious ten-year-old case involving a promising young track star found brutally raped and murdered in Toronto. ViCLAS pointed investigators to a set of similar crimes that had occurred more recently across the continent in Vancouver; a DNA test on a sample taken from the suspect confirmed the link to the earlier crime.
Leury, who is 29, says he found in Wheeler a kindred spirit—and a programmer who thought along the same lines. When he was still working with the Mounties, he says, he had coveted InfoGlide’s special data-analysis capability. All of InfoGlide’s products are based on a new type of search technology that Wheeler developed and patented. Searches for matching information, such as those on the Internet, use a limited number of fields and require more or less perfect spelling to get results, but Wheeler’s system actually analyzes as it searches, allowing for human error. His system, he explains, looks for a “degree of similarity” rather than perfect matches in searching for connections between crimes or criminals—an important distinction in police work, when information can be vague or inconsistent, and witnesses may give conflicting descriptions. If an investigator enters information on a case into his computer and then asks the Toolkit for links to other cases, the results of the search will supply an ordered list of crimes likely committed by the same individual—with the closest matches ranked on top. The system, Wheeler says, “mimics the detective’s decision-making process.” Best of all, the results of a search come in a few minutes instead of a few hours.
Developing the basic search system took a year, but building InfoGlide took seven years, because of all the delays, frustrations, and setbacks of any high-tech start-up. “It was a long road to get here,” Wheeler says. And apparently it was a road with plenty of detours and blind alleys. “I put my family’s financial security in jeopardy,” admits Wheeler, who has three sons. At one point, he recalls, he had to sell the family car to keep going. But he persisted, driven by memories of his father.
Following the murder in 1981, Wheeler’s life was thrown out of kilter. The crime strained relations between him, his mother, and his siblings, because of his insistence that his father’s interest in the jai alai business got him killed. Wheeler says his father told him shortly before his death that he had begun to rid the operation of mismanagement. (He still believes in a link to jai alai—and so does Huff.)
Wheeler started a computer consulting business in the early eighties, then sold it and entered the graduate program in physics at UT-Austin. He dropped out in 1990, partly because there were so few job openings in physics and partly because he was still focused on his father’s death. He called Huff and proposed developing a computer program to deal with managing the data in cases like his father’s, which by that point had filled four filing cabinets. Soon Wheeler began to envision a broad program to help police officers in their investigations. Tulsa had been plagued by a serial rapist, known as the Southside Stalker, who was ultimately shot to death by the police during a robbery attempt, and he could see for himself the difficulties the police faced in trying to analyze mountains of information.
Working partly in Austin and partly in Tulsa, Wheeler concentrated his efforts on developing a program to track serial offenders, following investigators to crime scenes and watching as they interviewed witnesses. “It took me a while to realize the breadth of the problem,” he says. “The way the system was designed was almost hopeless—there was a major flaw in their database technology, and it was helping crooks get away.” The flaw, he says, was that “they were using accounting software to track crimes.” Like those of most police departments, Tulsa’s basic computer system had been adapted from other uses rather than specifically designed for the peculiarities of police work. In other words, they were using software that depended on “perfect” data—and yet the information that was going into the system was far from perfect.
Wheeler realized he had to design his basic search system for “the worst kind of data you can get.” It had to allow for mistakes and discrepancies, including misspellings by officers writing a report. “When you have a crime,” says Wheeler, “what are the chances of even getting accurate reporting about height and weight? With the old system, you’d get a list of prime suspects who didn’t do it.” One of the main advantages he had going for him was the nature of the offenders themselves. Most repeat offenders, he learned, develop certain patterns in their crimes. They may prefer, say, a certain geographical area, a certain kind of victim, or a certain time of day. This includes sex offenders: Even though they tend to escalate their crimes, there are still elements of each incident that can be linked to earlier ones. “These crimes are fantasy based,” Leury says, “and the fantasy can never be satisfied.”
The first version of Wheeler’s Toolkit was designed for Tulsa’s armed robbery division, where it soon proved as useful as he had hoped. In one of its first cases, it immediately linked a robbery at a Pizza Hut to another at a Whataburger—precisely what the police had independently determined. In another case it linked a series of robberies carried out initially by two partners, one black and one white, one using a sawed-off shotgun and one a .45 pistol. The white robber later committed crimes on his own, but the Toolkit linked him, via the .45 and a few other details, to the previous crimes. After the suspect was apprehended, Wheeler says, the investigators were able to taunt him with the dates of the previous robberies. “He thought they had been watching him all along,” he says, “and he confessed and gave up the name of his accomplice.” In yet another series of robberies, a small-business owner had been using a few henchmen to steal jewels from wealthy women, so each of the perpetrators had a different physical description. But the other details were linked by the Toolkit back to the business owner, who, it turned out, knew each of the victims.
Huff says Tulsa’s robbery rate has dropped since the department began using the Toolkit. The software was also tested by the department’s sex crimes unit, which has credited it with identifying a rapist after only two assaults in the Brookside area of Tulsa. And the Toolkit has since been adapted in Tulsa to other kinds of crime, including domestic violence. “The best thing about it,” Wheeler says, “is that it is able to prevent more crimes.” Police departments in other cities, however, are proving a hard sell. “They’re resistant to changes in the way they do business,” says Wheeler. “And I think they’ve been burned by a lot of things that don’t work. They tend to want more practical things, like case management—just keeping track of cases—rather than case solving. We’re working on giving them that capability too.”
Interestingly, one case InfoGlide’s technology hasn’t been able to solve is Roger Wheeler’s murder. “It has never been used on my father’s case,” says Wheeler. Huff explains that he would need data from other law enforcement agencies that he hasn’t had access to. Regardless, he thinks the case will still be solved someday. “We think one of the killers may be a serial offender,” he says. “That would be ironic.”