Using Review Technology to Boost Quality & Efficiency
In a recent case from the Northern District of California, producing parties intent on saving costs refused to hire a third party, instead relying on five attorneys to conduct the entire review of "every bit of that giant mass of information" with no search terms to narrow the data. Seeing "no end in sight," the court noted the need for a new method and ordered the parties to split the cost, as offered by the requesting party, of retaining a third-party service provider to assist with review and production.2
This case highlights the formidable challenges of discovery: managing costs, meeting deadlines, utilizing available means to produce all responsive documents, and resolving disputes in the spirit of cooperation and good faith. Importantly, these challenges can be met with the capabilities of intelligent review technology (IRT), which enables document review teams to conduct a repeatable, defensible and efficient process. IRT augments the human-intensive aspects of the document review process, frees attorneys to work on case strategy, improves the quality of review, and results in faster and cheaper discovery.
The foundation of IRT, workflow, provides the technical framework upon which the other aspects of IRT function. Prioritization is then used to analyze reviewer categorization decisions, elevating documents most likely to be relevant to the case to allow the most relevant documents to be reviewed first. Categorization advances the document review project by analyzing human categorization decisions to recommend categories for documents not yet reviewed by a person. IRT ultimately integrates the irreplaceable input from a human team with smart technology for maximum accuracy and efficiency.
Key features attorneys should look for in intelligent review technologies are:
- Workflow automation, which minimizes human work and inconsistencies in the staging, distribution, routing, assessment and quality control of the review process.
- Supervised learning, which automatically produces statistical models to prioritize potentially responsive data by learning from manually reviewed documents.
- Statistical quality control, especially sampling, which monitors the progress and effectiveness of prioritization and review, and supports defensible decisions to cease review.