Peer to Peer Magazine

15 WWW.ILTANET.ORG BEST PRACTICES Three Ways Analytics Can Tell Your Client's Story Faster and Beer JONATHAN LAND Jonathan Land, of counsel, Quinn Emanuel Urquhart & Sullivan LLP, is the director of the firm's Litigation Support department. Contact him at jonathanland@quinnemanuel.com. SHEILA MACKAY Sheila Mackay, Vice President of Xerox Legal Business Services, is responsible for business development for some of XLS' largest clients and assists with a wide range of complex litigation and regulatory investigations. Contact her at sheila.mackay@xls.xerox.com. possible cost, first leverage an early data assessment tool that analyzes relationships between players in the case. » Aer deduplicating the collection, run a relationship analysis tool across the remaining documents to investigate who was emailing whom, and note the domains. This tool visually breaks down the who, what and when of complex email communications and shows paerns and spikes so you can determine which custodians, time frames and/or domains are vital to your review. Based on your analysis, you can learn that you should focus your review on four of the 20 custodians whose data you collected, meaning your initial document set now comprises fewer documents and a significant reduction in resources and manual review hours. » Then leverage a concept analysis tool to create groups of documents. With this insight, you can analyze concepts and make more informed decisions about potential keywords. Test keywords for effectiveness and apply the most appropriate terms for data reduction and/or locating key documents for early review. » The use of a concept-clustering tool can also identify clusters of blatantly nonresponsive documents for initial exclusion from the review. With these analytics, you can reduce those 750,000 documents to 200,000 for manual review, saving your client thousands of review hours. Scenario 3: Huge Volume, Short Timeline You've just finished processing the documents you collected from your client for discovery. Fortunately, opposing counsel agreed to limit the collection to necessary custodians and a reasonable time frame. However, aer deduplicating the data set, you're still le with 500,000 documents (mostly email messages and word processing documents). Manual review, at a rate of 50 documents per hour, would take 10,000 hours of aorney review time, which is 417 days of 24/7 review. This poses two problems. Your review deadline is just 120 days away. Your client's budget won't allow for such a costly project. For projects of significant volume that are light on non-searchable documents like photos, technology- assisted review (TAR) can be an efficient approach. Here, your team uses the TAR engine to generate a random sample of documents for responsiveness coding by senior aorneys with in-depth knowledge of the case. Aer your firm's senior aorneys code a sample of 2,500 documents, the TAR algorithm can compare their coding to each document's content, determining the criteria that make a document more likely to be relevant. TAR is run on the remaining document corpus, assigning probability scores reflecting responsiveness to each document in the collection, so you can prioritize the documents most likely to be responsive for review and ultimately discard obviously non-responsive documents. As a result, you can significantly reduce the corpus for manual review, allowing you to stay within the client's budget and timeline. Takeaways These sample scenarios demonstrate that the proper use of analytics tools can dramatically increase the efficiency of e-discovery projects while reducing costs. The key to is to understand how analytics speed up review based on the particular case and to choose the right tool (or set of tools) to identify and quickly review high-priority documents, followed by lower priority documents, and to eliminate the review of non-relevant documents. It is important to track and document all tools employed and steps taken when using analytics to reduce a data population so you are prepared to defend yourself if the document production is challenged. P2P 1 2

• Cover