In eDiscovery, document review accounts for 70 or more cents of every eDiscovery dollar spent. With data volumes soaring, legal teams need better techniques to control these costs. The twin tasks of reducing the number of documents and making the review process more efficient are ways to achieve this goal. While today’s technology can cull and organize potentially relevant data for review, this is only part of the equation: legal teams also need to protect against the disclosure of sensitive and privileged information. This is where the costs mount up quickly, and current approaches fall short.
Here’s why. The costs of inadvertently exposing privileged information makes it one of the most time-intensive part of the review process. But, shockingly, based on our analysis of over 10 billion review decisions across thousands of matters, more than 90 percent of documents identified as being “potentially privileged” through traditional approaches do not end up being withheld—creating substantial and unnecessary document review costs. Further, an alarming volume of privileged data still slips through the cracks—creating risk.
There’s good news for legal teams seeking to minimize the high financial stakes of privilege review: A new data-driven, contextual process can enhance privilege review by slashing review time and cost while improving precision and accuracy. Based on client and test cases, this approach routinely scores eight times better than traditional approaches, resulting in a significant decrease in total review costs and time.
The Old Privilege Screen
Legal teams typically screen for privileged data using keywords, which focus on the text of a document population. By design, keywords are overly broad and yield high hit rates with very low precision. Thus, human reviewers spend significant time reviewing the document set for privilege, despite a high percentage of false positives. Often, however, document properties beyond the actual text in a document determine whether or not it can be safely withheld for privilege, such as sender and recipient information. Keywords are notoriously deficient in targeting this type of context.
A new analytics-based approach to privilege review goes far beyond simple text searching. Instead of returning proportionately high volumes of documents matching keyword sets, algorithms identify documents using multiple factors that more closely conform to the principle definitions of privilege. The result is a more targeted and precise “potentially privileged” dataset.
This approach derives context from a combination of keywords, metadata, and human input to create a highly accurate privilege recognition framework. This is achieved through a combination of rules-based tagging and machine learning.
Rules-based tagging combines key document properties and privilege term analysis to increase search accuracy, such as known legal domains, internal and outside counsel attorney lists, and as other items that aid in the identification of sensitive content (e.g., personally identifiable information). Machine learning continually analyzes the key features and characteristics of the hosted data set, and applies ongoing analysis to classify privileged and non-privileged content. In order to better focus privilege review efforts, the process goes beyond the simple binary classification of “potentially privileged,” in which documents are either identified as potentially containing privileged information or not. This result set includes a privilege classifier score (similar to the TAR-based relevancy scores) that enables the targeting of specific subsets and yields a more focused and accurate review.
With this analytics-based classification process, human review becomes dramatically faster, since reviewers are working with a small fraction of the dataset that traditional privilege screens produce. Review teams can use the toolset for new matters, or in look-back projects to gain better insights into data and to save significant money and time. The process also improves results over time as the algorithms strengthen with more input, translating into a demonstrable return on investment realized through greater efficiency on future matters.
Cost and time savings are the obvious results of replacing standard search methods with an analytics-based approach to privilege review. Perhaps less obvious is the opportunity for legal teams to spend more time on strategy and preparation, which may result in more favorable outcomes.