A system for predicting health of an e-contract

Abstract

Large corporates and government institutions tend to have multi-year service engagements with IT-vendors who provide various kinds of infrastructure and management services including software, servers, IT-security, facility management etc. The engagements are formalized as legal contracts. Given the legal binding and the natural expectation to renew, there is a need to constantly monitor the performance against the stated objectives. Typically, a principle investigator or an account executive hold the responsibilities of deliverables and perform corrective actions as and when required. Given the complexity of the language and various variables involved, it is extremely difficult to proactively detect performance degradation. In this paper, we introduce the notion of contract health and describe Fitcon-a contract mining system that detects service level agreements from contracts, tracks the delivery performance against them and predicts the health of long term contracts. Fitcon is intelligent as it reads these natural language legal contract documents to set its monitoring goals. Fitcon tracks key parameters of the contract and predicts their performance. Comprehensive experiments on real contracts and corresponding past performance data are conducted to test the effectiveness of the system. We test our system across 11 large engagements and the analysis shows promising results. To the best of our knowledge, Fitcon is first such system that has been deployed into production for large scale contract health determination and prediction.