I recently interviewed Melinda Kunjasich, Senior Director, Document Review Services at Epiq in London. I was interested to find out about the support which Epiq gives to clients using predictive coding and about how this fits into their analytics strategy in discovery review.
It is perhaps first worth looking at Epiq's webpage about predictive coding – terminology varies from provider to provider, and Epiq sets out clearly what they mean by predictive coding. Epiq’s predictive coding is:
a prescriptive linear review workflow that is a mix of statistics, categorization and reporting that can decrease the cost of a review by amplifying the efforts of a small number of reviewers on a large dataset. Predictive coding is particularly useful in regulatory productions when the dataset is enormous and the production deadline is tight.
Melinda Kunjasich said that clients get as little or as much support as they want from Epiq when they embark on a predictive coding exercise. It can be quite daunting, she said, if it is their first matter, and Epiq's support is then quite substantial both for process and review strategy, for optimising workflows, for sampling and all the other functions related to a predictive coding project.
While predictive coding is most generally thought of in the context of very large matters, it can have value in smaller ones. Other tools are used more frequently - almost all cases now use email threading and near-duplicate analysis which can save between 10% and 30% of the review costs.
I have known of predictive coding cases where the clients only turned to it because everything else has failed to bring the volumes down to a satisfactory level. More usually, Melinda Kunjasich says, clients come to Epiq at the beginning of the matter and get recommendations as to which of the suite of analytical tools are right for that matter.
This will depend on the size, complexity and other factors. The principles are the same whether the matter involves litigation, internal investigations or regulatory investigations.
The uses are varied and not limited to preparing the client's own disclosure sets: one client used predictive coding when it was necessary to review very quickly the disclosure given by the other side; any large exercise where it is important to get lawyers' eyes on key documents quickly can benefit from its use.