The ability to reliably identify responsive documents early in the review process unlocks a number of new strategies to streamline e-discovery and control costs.

1. Tiered Review

The grouping of documents based upon probability of responsiveness allows the legal review task to be assigned to different levels of reviewers. For example, lawyers within the firm may be assigned documents that are highly likely to be responsive thereby preserving the knowledge gained from the review within the trial team. Documents that have a general probability of responsiveness can be assigned to contract review teams at lower billing rates and documents that have a probability of non-responsiveness can be assigned for general first-pass review to lower cost resources. Blending the billing rates based on the likely responsiveness of the documents significantly changes the economics of e-discovery.

Servient integrates Predictive Review with its full featured review manager. The assignment and management of documents to review teams based upon probable responsiveness is completely automated. Servient enables tiered review among multiple review teams with just a few mouse clicks.

2. Reduce the Review of Irrelevant Documents

Predictive Review serves as a much more accurate and defensible filter than base keyword searching. In essence, Predictive Review significantly improves the precision of e-discovery search without sacrificing recall. Predictive Review was designed to allow the legal team to review all documents identified as responsive during the learning iterations while limiting the review of non-responsive documents.

Predictive Review arms the legal team with scientific validation of the accuracy of the non-responsive classification through integrated statistical analysis. Servient allows the review phase of e-discovery to be completed efficiently by cutting out the unnecessary review of irrelevant documents.

3. Improved Review Quality

Predictive Review provides a new and powerful quality assurance tool that increases the consistency of review.

With Predictive Review, the review manager can compare the manual reviewers' rate of agreement with the automated document decisions. This comparison will quickly identify reviewers that are out of sync with the rest of the team. The technology also can identify discrepancies in review calls at the document content level. Armed with this information, the review manager can quickly identify misunderstandings regarding the factual basis of the matter and correct the review.

Increasing the consistency of the manual review not only improves the quality of the e-discovery process, it also improves the effectiveness of the machine learning process. With Predictive Review, the review manager can reduce the error rate associated with manual review.

4. Early Focus on Relevant Documents

Instead of manual review, which typically proceeds at a rate of reviewing 2 relevant documents out of ever 10 reviewed documents, Predictive Review instead focuses the review on predominately relevant documents at an early point in the process. By moving the responsive documents to the start of the review, the quality assurance and privilege verification phases of the project can start at a much earlier point, thereby reducing the rushed second level review and production preparation that often occur in today's litigation.

Also, the legal analysis of the important documents can move forward without the need to wait for the completion of the formal review. Servient has integrated the advanced machine learning into its full-featured e-discovery platform. Any search results returned in the review system can be filtered based upon the probable relevance of the documents. At an early stage in the case, attorneys can quickly access and review relevant documents without wading through the significant volume of irrelevant files that permeate most document sets.

5. Improved Data Culling

When applied against the entire litigation hold data set, Predictive Review locates more responsive documents in the collected data. Assuming traditional keyword filtering was applied before review, the learning model created as part of the Predictive Review process can be used to locate additional relevant documents from the collected data set. Predictive Review can identify documents that do not contain the selected keywords but have a high relevance probability.

In practice, Predictive Review enhances the recall rate of standard search technology. Predictive Review reduces the number of irrelevant files submitted for review, while increasing the number of relevant files identified in the e-discovery process.