Medical device company

But there was a problem. The defendants’ production was a mess. Many documents were poorly scanned and lacked metadata. The OCR text for the scanned documents was riddled with errors. Thousands of emails had been so completely redacted that even the address and subject line showed only “redacted.” Given the condition of the data and garbled OCR, keyword searching was ineffective and inconclusive. Reviewing just the documents that hit on highly focused searches, only 5% were potential deposition exhibits and only 51% where either relevant or hot. The attorneys were certain they were missing important documents, even as their time to prepare was running short.

Prioritizing hot documents for review

With depositions looming, the attorneys turned to OpenText™ Insight Predict for help zeroing in on hot documents. Using technology assisted review (TAR) the firm succeeded in prioritizing a significantly greater number of hot and relevant documents than they had using keyword searching alone.

Working with OpenText, the team started by having a lead attorney QC the documents already tagged as hot. Then the attorney reviewed a few hundred more targeted hits and some further samples to identify additional hot documents.

Using those documents as seeds, Insight Predict ranked the entire population for hot documents, concentrating them at the top of the list. The top thousand unreviewed documents were then pulled for the attorneys to evaluate. In this way, the proportions of hot and relevant documents were greatly enhanced. Through keyword searching, only 5% of documents found were hot and 46% were relevant. But through TAR, 27% of the top-ranked documents were hot and 65% were relevant. 

The above graph shows the breakdown of that top slice of roughly 1,000 documents out of the 77,000 documents ranked. The second bar shows the 258 documents judged by the reviewing attorneys to be hot. Nearly all the rest of the documents—the first bar of 616— were judged to be at least relevant.

With over 92% relevance and over a quarter of the documents actually deemed “hot,” the attorneys now had a rich, small set of documents to work through. Insight Predict rankings allowed them to quickly and efficiently find everything they needed.

Gaining good results from difficult data

Because Insight Predict allows many seeds from judgmental sampling, documents already coded helped to achieve results far better than would be expected from the challenging document set. The above table compares the review ratios for the unranked search hits and for the high-ranked documents ranked by Insight Predict. Thanks to TAR, the plaintiffs’ attorneys worked with a set rich with hot and relevant documents, enabling them to prepare thoroughly for depositions by reviewing the hot documents that pertained to the deponents and issues in the case.

Cutting costs and increasing productivity

TAR is one the best ways to find hot deposition documents in the opposing side’s production. It even helps overcome problems of missing metadata and mangled text and continues to improve as the system learns more about the case. TAR saves time and money, helps legal teams prepare sooner and enables them to focus on what is important.