Predictive Coding


Predictive review is the future of litigation discovery technology!

Also known as “computer assisted review,” predictive review is technology that helps save significant amounts of legal fees for document review in eDiscovery and other document-intensive matters.  This is because the majority of documents are coded by computers, not humans, for privilege and relevance.  This is not the case of machines replacing humans, however.  Humans still play a critical role in predictive coding by coding a small sample of documents and helping the computer “learn” about the case.

Highly effective for document-intensive matters, predictive coding is similar to traditional document review methods with a few important exceptions.  With a traditional document review, senior attorneys and project managers train a team of human reviewers on the review objectives at the beginning of a project.  Then, once the review team completes its work, the senior attorneys and project managers perform a quality check.

The difference between predictive coding and a “traditional” review as described above is that computers perform the majority of the document review team’s work.  The senior level attorneys and project managers most familiar with the case still play an integral role in predictive coding.  How?  Rather than training a human team of document reviewers, they train the computer to perform the review.  This training is accomplished by having the senior attorney or project manager review and code a sample set of documents.  The computer then analyzes those decisions and applies that same logic with great accuracy to the entire document collection.

It is the process of manually reviewing a small sample of documents for relevance and privilege.  Computers then assist the review by “learning” about the key issues in the case based on the decisions made during the initial manual review.  The remaining document collection is analyzed by the computer and automatically makes relevance and privilege decisions.  Upon completion of the analysis, the same quality check is performed by the senior attorney or project manager to ensure accuracy and completeness.

The key difference between predictive coding and a traditional document review project is that the computer performs the majority of the coding independently after it has received human input.  Predictive coding is not only a more accurate way to conduct document reviews but it is also far less expensive than paying human reviewers to code the majority of the documents.