Data Analysis/Mining.

As an inherently technical-based industry, the vehicle inspection business involves the collection and analysis of a tremendous amount of data. All motor vehicles that are tested must meet quantitative inspection standards in order to pass their test. This in turn leads to a number of critical data collection and analysis issues, including:

  • Proper identification of test vehicles so that they can be properly tested and the correct pass / fail standards can be applied (different types of vehicles are subject to different tests and standards).
  • Proper application of test procedures and standards to the identified vehicle.
  • Proper test decisions (pass or fail) depending on the test results for the vehicle.
  • Proper recording of the test results and transmittal to the appropriate systems (e.g., to the state vehicle registration database so that the owner is allowed to register a passing vehicle).
  • Proper calibration and auditing of test equipment performance to verify it continues to operate within allowable specifications.
  • Proper oversight and auditing of inspector performance to ensure they continue to correctly test vehicles.
  • Proper reporting of test result summaries and other program data to the oversight agency and the U.S. EPA.
  • Complex data mining techniques (“triggers” analysis) to identify possible inspection fraud; i.e., evasion of vehicle inspection requirements through substitution of improper vehicles (so-called “clean scanning”) and other deceitful means.

Given its position as the vehicle inspection quality leader, Gordon-Darby performs the above activities better than any of our competitors. Achieving this level of performance has required the company to develop and implement systems and processes that incorporate all of these functions. Part of these processes has been the development of analysis techniques, and subsequent ongoing analysis of many millions of vehicle test records and other data. While much of this effort is specific to the vehicle inspection industry, a number of elements have applicability to other business sectors.

Gordon-Darby personnel are expert at developing and applying advanced automated data QC and error checking routines. The purpose of these routines are to ensure all collected data comply with allowable specifications, are free from unwanted biases, and do not reflect anomalous equipment or inspector performance. Data QC is a standard required element of all analysis activities regardless of the business sector.
An important element of Gordon-Darby analysis efforts include trends analyses conducted on both test equipment calibration and other QC data, and vehicle test results. Regarding the former, statistical process control (SPC) and other types of analyses are routinely conducted on collected QC data to verify that the test equipment is operated within specification and vehicles are being properly tested. In addition, vehicle test data are analyzed to determine both the effectiveness of the inspection program over time in reducing emissions and trends in vehicle emissions independent of the program.
Our expertise in SPC analysis is applicable to a wide range of manufacturing processes, while additional Gordon-Darby trends analysis capabilities can be applied to any datasets involving information collected over an ongoing period of time or from a disparate group of individuals.
Gordon-Darby’s triggers analysis procedures largely involve comparing the performance of individual test systems or inspectors to the performance of each group as a whole. By identifying statistical outliers or other anomalous results in the data, the software algorithms can quickly hone in on patterns of questionable performance (human or equipment) and identify exceptions to normal/acceptable behavior. By automating these algorithms and the results they generate, it is also possible to sort through a tremendous amount of data using relatively few human resources, who can instead be focused on following up on identified anomalies in the trigger results.
These trigger principles and procedures have relevant applications in other business sectors such as homeland security and those involving general fraud prevention and equipment performance tracking.
In general, “trigger”-type enforcement reports are available on a daily basis and reviewed by management staff. If the reports indicate suspicious behavior, management conducts an investigation, including review of the surveillance videos when applicable. If warranted, the inspector may be targeted during a covert audit.
For fraud detection, Gordon-Darby produces a variety of reports showing vehicle test and other “trigger” data to identify instances of suspicious activity. Best-of-breed triggers algorithms, investigative queries and other industry-leading attributes are incorporated to provide our users with “instant access” to easy-to-understand results they can use to target possible problem performers (stations and inspectors) for follow-up enforcement actions.
We have also expanded our trigger analyses and reporting to privately owned inspection stations and inspectors in our decentralized programs; e.g., the New Hampshire OBD/Safety Test (NHOST) and Texas programs (GDI TIMS).
Gordon-Darby has developed triggers algorithms and associated Investigative Queries (IQs)™ that have been very relevant and helpful in identifying suspected inspection fraud and other poor performance on the part of decentralized inspection stations and inspectors. Our triggers/IQ applications combine a number of positive, industry-leading attributes that are designed to maximize the utility of the applications:

  • Robust computational methodologies based on U.S. EPA-funded best practices documentation
  • “Smart” applications that minimize the amount of thinking users have to do in generating triggers results
  • Point-and-click ease of use
  • Fully user-selectable analysis periods, including capability to perform before and after analysis of station/inspector performance relative to audits and other enforcement visits
  • Very fast return of comprehensive yet easy to understand graphical and tabular results
  • Linkage to vehicle registration records to quickly identify vehicles used in suspected clean scanning; i.e., using a known clean OBD vehicle in place of the supposed test vehicle in order to fraudulently pass the vehicle inspection
  • Easy download of results for further offline analysis
  • Adaptive detection techniques that evolve to catch cheaters who modify their actions to avoid detection

These investigative elements reflect and demonstrate the continued evolutionary development of the management tools we incorporate into our programs. Such an evolutionary process is possible due to Gordon-Darby’s long history of managing successful Inspection & Maintenance (I/M) programs as well as our long standing management and development staff with notable additions over this period.