In this case, opposing counsel produced more than 800,000 documents (267GB) to D4’s client. Their client suspected that this was a “document dump,” i.e., a production purposely designed to make it harder to find relevant information. The collection produced was based on keyword searches that both parties had agreed upon, but the client suspected the opposing party had not reviewed the production for relevancy. Receiving counsel needed to find out which documents were most important to support their side of the dispute, but they were faced with a gargantuan task.
Rather than review the entire production, D4 and their client decided to use predictive coding to focus and narrow their review. This decision was rewarded. Predictive coding revealed that the team could defensibly focus their review on just 30% of the documents. Here’s why:
- More than half of the most relevant information was found in just 30% of the corpus (240,000 documents).
- The “bottom” 30% of the corpus contained information with such low relevance scores that it could be safely ignored.
- The team decided that the remaining 40% would not be reviewed unless they failed to find what they were looking for in the targeted 240,000 documents.
In the end, receiving counsel was able to fully avoid reviewing approximately 70% (560,000 documents) of what opposing counsel had submitted – with very little risk. Keep in mind that this was not a review conducted for production, but rather a review of documents produced by the other side. Thus, there was no risk on the client’s side for failing to produce a relevant document. Sufficient documents to support the case were found in the top 30%. As such, D4’s client saved approximately $1.4 million in review costs, while focusing their attention on the most important documents earlier in the process. The eventual outcome was that D4’s client won the case. This customer now uses predictive coding as a standard practice for incoming productions.