Traditionally, predictive technologies, like TAR, were only beneficial on large matters with lots of data. DISCO Review uses continuous active learning (CAL) and optimized workflows to provide predictions at the beginning of the review, no matter the size.
“This was my second managed review with DISCO, and I will definitely go back for a third. Facing a tight deadline, DISCO was able to quickly assemble a team to assist us with an expedited discovery project. Thanks to the speed of the reviewers and the quality of their work, we met our deadline with time to spare and stayed under budget.”
Adams and Reese LLP was representing a client in a large commercial dispute involving expedited discovery, and had one week to make a production from thousands of documents.
■ 6,415 documents to review in seven days
■ Review team of three DISCO power users delivered speeds of 86 docs/hr and quality results
■ Project completed ahead of schedule and under budget
The review team used DISCO AI to prioritize the documents. While the ultimate percentage of responsive documents was only 9.2%, DISCO AI sent reviewers batches that contained 27.5% responsive documents on the first day — nearly 3x the number they would have found if DISCO AI had not been used for prioritization.
Based on the reviewers’ coding decisions, DISCO AI builds a model for responsive documents and sends the documents most likely to be responsive to the reviewers first. As the reviewers accept or reject the predictions, the model keeps learning and getting better.
Rather than creating a set of static batches at the beginning of the review, DISCO’s “just-in-time” batching will create a batch only when a reviewer requests one. This technique ensures that each batch contains the documents most likely to be responsive based on the most up-to-date DISCO AI model, effectively making each batch “smarter” than the last.
As more responsive documents were found over the course of the review, the percentage of relevant documents in each batch dropped. On the final day, the reviewers found only 0.44% responsive documents in their batches.
With the help of DISCO AI, the review team found 86% of the responsive documents after only three days, and after reviewing only 48% of the document population.
The DISCO team worked closely with Adams and Reese LLP to ensure the reviewers were well-trained and didn’t sacrifice accuracy for speed.
The DISCO review team maintained high quality review standards even at its fastest pace, with a privilege overturn rate of 1% and responsiveness overturn rate of
When reviewers receive batches that contain similar documents whether responsive or non-responsive they don’t have to switch context between each document and are able to make coding decisions more quickly.
Using the DISCO platform — with 1⁄3 second document load times and lawyer-focused user interface design — the reviewers were never limited by the speed of their technology, and quickly developed a rapid pace. The team achieved a speed of 55 docs/hr on the first day. On the final day, when the team received batches with only 0.44% responsive documents, the team achieved a peak speed of 164 docs/hr.
Over the course of the review, the review team averaged 86 documents per hour. Given the industry average of 55 documents per hour, DISCO Review offered a 1.56x increase in efficiency.
The DISCO team finished the review ahead of schedule and under budget. The small team of only three reviewers powered through thousands of documents to meet the seven-day production deadline, while delivering a high quality result for Adams and Reese’s client. Even on a small matter, DISCO AI efficiently prioritized documents for a cost-effective review.Download this Case Study as a PDF