Background
The US Federal Courts (the “Third Branch” of government) operates 13 appeals circuits and 94 districts throught the territory of the United States, along with a number of special pupose courts in Washington DC. There are over 400 court locations in the system, whose IT infrastructure is funded by the Courts’ Administrative Office (AO), based in Washington. The funds each court receives are determined by an allotment formula maintained by the AO. This is intended to fairly distribute a pool of funds among all of the courts using a model (called the “allotment formula”) that calculates a given court’s annual IT infrastructure expenses, taking both operation expenses and annualized capital needs into account.

By late 2012, The allotment formula had not been updated for nearly a decade and contained no provisions for funding technologies such as tablets and smartphones, WiFi, cloud services, and so-on, had been made. Furthermore, the existing formula did not take pricing trends, or changes to the staffing levels required to operate the court system into account, and as a result, the formula’s output was diverging rapidly from the actual infrastructure needs of the courts.

The AO needed to modernize its allotment formula and a competitive bidding process was begun to select a qualified external vendor. After reviewing the available options, the AO decided to stop the competitive selection process and awarded Quantellia the sole-source contract, based the company’s unique technology and qualifications.

The project consisted of the following phases:

1. Conduct a comprehensive electronic survey among all court locations to obtain a more accurate inventory of equipment and services acutally in use, a more accurate view of projected needs going forward, and an indication of where the current budgeting
   discrepancies lie.
2. Analysis of survey results, merged with actual spending records for the past 3 years to determine:
   • The right list of variables on which the formula should depend.
   • The new functional form of the allotment formula.
   • A measure of the formula’s accuracy, based on how close the
     formula’s reccommended court-by-court budget for 2013 is to actual spending in 2013*.
3. Development of the new allotment formula, and the preparation of the presenation and training materials to assist the organization in transitioning to the new tool.

*2013 results were not included in the analysis, and so provide an independent check of the formula’s accuracy.

Not just any survey/data acquisition tool
The first stage of the project, data acquisition, required that close to 400 courts be sent a link to an on-line survey tool. The survey tool had a number of unique requirements:

• Workflow. While the link to the survey was only sent to the administrative head of each court, most of the content would be filled in by delegating specific sections to the responsible department heads, who would delegate sub-sections to their reports, and so-on, with arbitrarily many levels of delegation. Delegates see only the sections and sub-sections delegated to them and, when they are finished, submit the completed sections back to the delegator. This continued until the first ricipient had all completed responses submitted back to them.
• Dynamic Content. The questions asked, response options offered, and the sections shown depended on the particulars of each court. So the survey had to dynamically create each questionnaire page depending on context as specified by a set of rules.
• Dynamic Data.
  World Modeler was not only used to gather data about existing assets, but allowed users to created new records for assets that had not been previously captured.
  

  Example Survey
  Pages… Click
  to enlarge

  In these cases, the rules engine had to construct an appropriate questionnaire based on the type of asset and the context.

Previous attempts to implement these requirements by generating and distributing PDF files had not delivered success. Quantellia’s World Modeler platform was able to statisfy the above needs, thanks to its underlying modeling engine, and data binding layer. The application was delivered to the US Courts on time and on budget, and deployed to about 1500 users. The data collected during the first six months of operation (along with only one help desk query), was subsequently analyzed as described below.

Data Analysis and Model Building
Following successful delivery and deployment of the IT infrastructure data acquisition application, Quantellia was retained to analyze the data it collected, along with actual spending records from the Courts’ financial records. By using linear regression, CART, and other techniques, a set of independent variables and coefficients was developed that correlated budgeting to actual expenditure with a discrepancy of less than 1%.

A set of classifiers was also developed that divided the entire court system into five groups, each of which had particular characteristics that affected their allotment formula. A distinct formula was developed for each of the groups, leading to the dramatic improvemements the accuracy of the final result.