Case Study: Predictive Coding Reduces Time and Cost of Regulatory Review Matter

A high-profile investigatory matter involving discovery requests from a state regulatory agency and a state attorney general.

Client Challenge

Our client needed to conduct a high-quality, defensible investigation while protecting sensitive corporate information. Their regulatory agency relationship was at risk because:

• Their vendor was unable to deliver reliable, secure, high-volume document review and meet the expectations of the regulatory agency and attorney general.
• Existing keyword-based culling processes were ineffective within a collection from 100+ custodians/1.2 TB of raw data.
• There was unfamiliarity with technology that could expedite or enhance processing, culling, and review.

Consilio Response

With client and outside counsel collaboration, we delivered review and consulting services powered by analytics capabilities. These specifically included:

• Designed and applied a custom deduplication process after a portion of the collection passed through prior vendor’s systems in order to defensibly eliminate duplicative content not identified by out-of-the-box deduplication tools.
• Acted as a data clearinghouse, coordinating with seven separate outside counsel firms to eliminate duplicative collection, review and production across nearly one hundred separate productions from multiple regulators.
• Leveraged Consilio proprietary scripts and processes to identify likely irrelevant email senders in order to eliminate non-responsive documents without the need for human review.
• Focused manual review of the most important and likely relevant data, delivered by review and project management teams experienced with efficiently handling detailed technical document collections under high pressure circumstances.
• Proactively coordinated application of multiple sets of review criteria, enabling downstream leveraging of reviewed documents for anticipated subsequent litigation without the need for additional human review.
• Utilized email threading to minimize review of duplicative content and supported outside counsel in defending the use of this technology when questioned by regulators.
• Predictive Culling™—applying decisions made during the high-priority custodian review to the remaining 1.6 million documents using an SVM algorithm, eliminating the need to “train” the system through repeated sampling.
• Assisted in presentation and negotiation of protocols with the opposing parties resulting in an agreement for a responsiveness threshold.
• Developed a collaborative custom solution for identifying potentially privileged documents and reducing the number of documents requiring manual privilege review.

Results Achieved

• Saved ~$2 million while meeting tight deadlines with high-quality, defensible review efforts.
• Classified 1.4 million documents in less than two weeks (with the first production in five days).
• Reduced the number of documents requiring manual privilege screening before delivery to the regulatory agency by 75% (from 1.6 million to 376,000).
• Reduced the number files passed on to human review by 40%, eliminating costs for unnecessary review of irrelevant documents.

2M

SAVED
while meeting tight deadlines with high-quality review

1.4M

DOCUMENTS
classified in less than two weeks

75%

LESS DOCUMENTS
requiring manual privilege screening

40%

LESS FILES
passed on to human review