Global energy company

Second, the law firm needed to prepare for interviews with key players, which necessitated incorporating new documents on an on-going basis into the review process. Furthermore, the review focus and scope changed each day based on the particular custodians’ interviews.

In the first phase of the project, the client and its outside counsel worked with OpenText to develop filters in OpenText™ Insight Predict to create simultaneous but separate review streams that accommodated rolling document collections into the process. The filters enabled the prioritization of review for documents associated with specific topics of interest, and new ones that emerged on a daily basis based on that day’s interviews with new custodians. Through date and custodian restrictions, the collection was narrowed to 112,000 documents for prioritized review.

The creation of separate review streams reduced the number of subject matter expert attorneys that needed to review the documents, along with associated legal fees. In the first stream, which included documents containing accounting subject matters, a special tax attorney reviewed prioritized documents in the system to identify documents of interest. These findings then led to the identification of additional custodians of interest, subsequent interviews and rolling document collections that needed to be immediately loaded into the system.

A separate finance review track was also created that included prioritized documents related to more general corporate finance subject matter. More than 50 percent of the documents prioritized by Insight Predict were deemed relevant.

In the second phase, outside counsel narrowed in on the two most important custodians involved in the allegations. Unlike the first phase, this phase required estimation sampling and high recall since the client wanted to set eyes on all documents that might be damaging to their case and did not want any “hot docs” to slip through the cracks. With a 30 percent estimated richness, the goal was to achieve greater than 80 percent recall.

Using Insight Predict, 10,000 out of 25,000 documents were ranked for review, and an additional 4,000 documents containing family members were reviewed outside of the Insight Predict system, using OpenText™ Insight’s review workflow. More than 10,000 documents were eliminated from the review by prioritizing the relevant documents to the beginning of the review. By the time the reviewers had reviewed about 10,000 documents, the batch richness had dropped to levels suggesting that most of the relevant material had been depleted. After reviewers completed the full-family review of relevant documents and coded a validation sample, the estimated recall of the project was calculated to be 89 percent. With the vast majority of the relevant material found, the client made the decision to stop the review, thus eliminating over 10,000 documents from human review.

In just 12 days, the law firm completed the entire review. With an 89 percent recall level after reviewing just 10,000 documents total using Insight Predict, the law firm was able to present findings to the company that guided its case strategy.