Suppose most traffic accidents in town occur on Main Street. The crashes usually are caused by two cars colliding at right angles - one on Main and the other on Elm Street. What on earth would you do?
Put stop signs at the corner of Main and Elm, of course.
That conclusion is easy to reach if you have two things: good data describing local traffic incidents and a system for analyzing and reporting on that data.
Until recently the Texas Department of Transportation (TxDOT) had neither, except a manual data collection system dating back to 1976.
A $9.9 million software implementation is changing all that, making it possible for state planners to re-engineer existing roads and build safer, new ones based on verifiable statistical intelligence.
"From an engineering perspective, you might know you have 10 crashes, but if you don't know where those crashes happened that does not help you fix the problem," said Carol Rawson, deputy director of TxDOT's traffic operations division. "Now we can focus our efforts on the areas where we can save the most lives."
Delving Into Data
The new Crash Records Information System (CRIS) comes from MicroStrategy, a developer whose business intelligence (BI) platforms support federal and state entities, such as the U.S. Postal Service, Ohio Department of Education, Texas Department of Agriculture and the U.S. Department of Housing and Urban Development.
At some level, all data-analysis implementations, such as TxDOT's traffic system, share a common starting point, said Mark LaRow, vice president of products for MicroStrategy.
"With all BI systems, the core component is data," he said. "That is almost always the biggest hurdle - just having clean, consistent data in a database."
TxDOT has a high hill to climb in this regard. The state logs approximately 600,000 car crashes each year, and police write-ups aggregate a wide range of incident data, for example: date and time, number of parties involved, injuries, weather conditions, if it's alcohol-related and if seat belts were worn.
In the old manual-entry system, officers had to compress all this information into a maximum of 254 characters, with no standardized data fields. There wasn't methodology to make the data uniform.
Since CRIS was implemented in summer 2006, outsourced data entry workers have been manually inputting traffic data dating back to 2001. Fourteen technical support workers maintain CRIS, including a database professional and a few reporting experts.
At the same time the old data is being re-entered into CRIS, the system is accumulating fresh data based on written reports from police officers in the field. The new reports are scanned and turned into Tagged Image File Format (TIFF) images. While the data must be inputted manually, the fields are uniform, which creates a pool of information that lends itself to meaningful analyses.
Clamoring for Reports
The net result of all this data collection and management is reports generated by CRIS that are based on criteria selected by the system's operators. As the database matures, demand for the reports is increasing, which comes as no surprise to LaRow.
He points to a typical school district as an example. A report on student testing goes to the principal. Soon, the teachers want to see it, and then the parents. "It's not just five analysts who need this information, but the rank and file too. It's the analysts, plus everybody else," LaRow said.
"The prevailing feeling is that, if there is information out there, everybody should be able to get to it. So I show somebody a report, then they show it to someone else and it grows and grows," he said.
Rawson said it's one of the perils of a successful BI implementation and something
planners need to anticipate. TxDOT began by producing 10 reports, and demand grew exponentially.
It happens because good data can drive intelligent decisions, which helps save money, enable more efficient work, and in the case of accident reporting, ultimately save lives.
"Once the user starts to see it, you've opened the world. 'OK, we like it this way, but could you dice it this way?'" Rawson said. "So you have to be diligent in deciding the scope of the project, otherwise it will creep; it will take on a life of its own."
Rawson has watched that "creep" extend beyond TxDOT. She said anyone with a stake in civic planning has been knocking on her door in recent months. Metropolitan planning organizations, the state's insurance department and municipal police agencies share an interest in knowing the specifics about traffic accidents. "Everybody wants it," she said.
Currently it's not feasible for Rawson's office to analyze and report on accident records tailored to every interested party. However, the office provides raw data to external agencies that make specific information requests, but the agencies must find their own means for interpreting and reporting that data.
Stats Drive Spending
Talk long enough about analysis and reporting, and eventually it starts to seem abstract - but not to Rawson. She's cognizant of the connection between all those CRIS information fields and those two cars with crumpled fenders at the corner of Main and Elm. She'd like to help prevent that wreck, but she's an engineer who knows only one way to do it.
"Engineers are driven by data to tell you where they are having problems, to understand the factors that influence those problems," she said.
Since 1976, TxDOT has worked with only partial information, and the department isn't alone. Antiquated manual information systems aren't exactly a novelty in state government, but replacement costs can be prohibitive. LaRow estimates a competent BI platform can cost between $250 and $2,000 per user depending on the need.
Rawson calls it a case of money spent equivalent to money saved. She expects to enlist the help of CRIS in an effort to design a $100 million driver-safety program over the next several months - a program informed by solid, quantifiable statistics. "If you go to a statewide level, for example, you are looking for crashes where people aren't wearing their seat belts," she said. Now TxDOT can know that with certainty, "So I start to ask, 'What can I do to impact it so people will wear their seat belts?'"
Upcoming construction projects likewise will be influenced by CRIS reports. Project selection, maintenance efforts - even the placement of guardrails - could be guided by improved statistical intelligence.
Even before any tangible results have materialized, many in Texas government are finding this an alluring proposition, and it's keeping Rawson on her toes. "For people who have been hungry for data," she said, "we could be working 24 hours a day to make this available."
NEW ON THE PODCAST