Migrated from eDJGroupInc.com. Author: Greg Buckles. Published: 2010-03-23 10:45:32Format, images and links may no longer function correctly.
The first two journal entries on autocoding definitely resonated with the market and generated a lot of on/offline responses. One of those responses was a call and subsequent briefing from my old colleagues at FTI Technology on their latest offering, FTI Acuity. [EDITOR’S NOTICE– Greg worked at Attenex prior to the FTI acquisition.] A quick visit to the FTI site leads one to believe that Acuity is just a packaged services-hosting offering with a fixed per-GB or per-Document pricing model that they term Integrated Document Review. DiscoverReady was one of the first Attenex partners to offer fixed-fee per document review pricing back in 2005. The FTI materials barely mention the new predictive coding and quality control functionality that has been added to the Ringtail-Patterns platform.
In Automated Review Part 1 & Part 2 we saw offerings that leveraged human review decisions to extrapolate either defensible relevance criteria (Equivio’s Relevance) or for a first pass relevance decisions (DiscoverReady iDecision). The new Acuity predictive coding walks the middle ground by placing a coded reference set of 500-1000 items back in the active review batch and then adding a suggested mark on clustered items within DocMapper. The reference items/dots are designated by a cross and proposed marks by a partial color. FTI says that the combination of known decisions and automated suggestions provides dynamic training, increased review speed and consistency. Review management best practice has always included providing reviews with exemplar documents during training with daily updates from the QA group. This was a manual process and we all knew that it was shutting the door after the cat had escaped most of the time. The predictive marking and clustering is still done of a batch basis. It does not dynamically update across review batches, but it can be done during the QA phase to catch any conflicting decisions.
FTI Acuity User Interface
FTI has always been careful not to endorse using the Document Mapper cluster spirals to bulk mark items. The new reference marking features enable to review manager to control the displayed marks and the confidence threshold levels of the item suggestion, cluster and discordant flagging. Their model now assumes some bulk marking of clusters since you can now see how closely they align to prior decisions and whether there are any discordant items (conflicts between the review marks and suggestions) within the cluster. Cluster level recommendations and marking are likely to be done at the discretion of counsel, but I can see immediate ECA scenarios where you have a known set of relevant items from a key custodian. This reference set would immediately enable you to see the primary common concepts and to even extract some rough relevance metrics across sampled custodians.
Early anecdotal reports from FTI have Acuity reviewers coding at 200-300% of prior Document Mapper review rates. A certain percentage of the documents cannot be clustered or are not similar enough to the reference documents to have suggested marks. This matches up with DiscoverReady’s experience of 10-20% ‘un-stackable’ documents. Automated coding will always require a certain level of initial training, second pass decisions and wading through of oddball docs.
Any new technology or methodology goes through an educational sale phase that is usually focused on the small percentage of early adopters. H5 has survived this period and we seem to be moving into a growth market for automated or predictive review offerings. The harsh economic realities of 2009 have put pressure on everyone to cut the main cost of civil discovery, manual review of ESI. Corporate counsel might have been unwilling to fight over more efficient review platforms in the past, because that left law firms fully responsible for waiver of privilege and the impossible goal of a perfect review. The pressure to cut costs and the dawning realization that technology is more efficient and consistent than manual review have pushed Biglaw to either surrender control of the review or get out in front of the curve. The days of the old-school linear review platforms are numbered. We are seeing the traditional platforms add analytics like the CT Summation-Equivio offering just to stay competitive.
It is interesting that except for Equivio’s Relevance, all of these automated/predictive review packages are hosted-only offerings. One gets the feeling that the providers would like to keep the technology close and encourage corporations to utilize their contract reviewers. The FTI Acuity offering comes packaged with consulting services to manage the lifecycle and specifically calls out statistical validation by their experts. Great software has flopped when released early into the wild without a skilled support network. So these managed services offerings may simply be an incubator strategy to ease new products into the market while being able to control backlash from insufficient customer resources, education or ‘not as designed’ usage cases. We shall see how Biglaw responds to this potential encroachment of their dominant revenue source, associate document review hours.