October 2, 2019
Clients frequently ask whether analytics, technology-assisted review (TAR), or an equivalent buzzword is a good fit for their matter or can assist in achieving a specific goal. The answer is a resounding yes! When strategically applied, data analytics tools and techniques developed by marketing analytics companies around the world can be brought to bear on almost any matter. The examples below represent just a few ways analytics can benefit your project.
High Level Overview of Data
One of the most frequent requests we receive is “please help me understand what I have,” be that either an opposing production or a recently completed collection. Fortunately, conceptual clustering is designed to do exactly that – with no user input. Clustering groups conceptually-similar documents together, e.g. all the marketing materials in one cluster, quarterly reports in another, and fantasy football stats in a third. Then, using several different user-friendly visualization methodologies you can quickly gain a high-level understanding of the types of information you have as well as tailor your review strategy accordingly.
Organization and Prioritization
Analytics can lend structure and organization to an otherwise unruly data set. Email threading, for instance, will identify and present email conversations in a logical thread view, much like Outlook or Gmail. This has the advantage of allowing the reviewer to interact with the entirety of a conversation rather than finding it piecemeal throughout the review universe. Similarly, conceptual clustering (referenced above) can group email threads or non-email documents regarding similar concepts into the same cluster so that a reviewer is not only reviewing the whole conversation arc, but is immediately presented with similar documents, directly increasing accuracy and efficiency. When there are multiple reviewers, this also promotes consistency.
Beyond organizing a data set, analytics tools can leverage a reviewer’s work to predict whether unreviewed documents are responsive or relevant to specified categories with an impressive level of accuracy. These document-level predictions can be used to prioritize the review of the most-likely to be interesting documents, leading to quicker identification of the most important material, so you can make informed decisions faster.
Quality Control (QC)
As you may have surmised by now, analytic tools are also powerful QC tools. Clusters with a high percentage of material identified as responsive may indicate that documents coded not responsive within that cluster merit extra scrutiny while a document coded as responsive but assigned a relatively low predicted rank may also warrant a second look. Similar methodologies are also immensely helpful when quality checking for privileged documents – particularly those that do not contain expected privilege search terms – or hot documents.
Reduction in Review Universe
Taking the document-level predictions referenced in the two sections above to their logical conclusions can lead to a reduced review universe based on the documents’ predicted responsiveness – and subject to rigorous statistical validation for defensibility – when the review universe is not rich in relevant material. This methodology forms the backbone of many modern predictive coding workflows and can lead to drastically reduced discovery costs as a result of reviewing only a fraction of the documents collected in a matter.
Assist in Discovery Strategy
Finally, analytic tools of all types can significantly aid in discovery strategy. For instance, an initial assessment of conceptual clusters can help a case team understand where to focus their efforts while a review of only those documents predicted most likely to be interesting can provide a low effort – high reward workflow for deposition preparation and general litigation strategy. For internal investigations, email threading visualizations and other communications analysis tools can quickly highlight unexpected connections between custodians or communications to third parties, which is particularly useful in investigations that don’t lend themselves easily to search terms.
The foregoing examples only scratch the surface of what analytics tools can help accomplish – in many ways the only limit on their capabilities is the proactivity and experience of the team consulting on their use.