I still remember the meeting where our network director explained virtual server farms to our IT team. Even as a technologist, it was hard to wrap my head around, but now it’s the norm and makes perfect sense. I can’t imagine going back to the old way. I felt the same way when someone first tried to explain the seed sets and training process involved in technology assisted review (TAR). It sounded like a lot of work for the amount of payoff. I’m guessing that I’m not the only person who felt that way, because that was ages ago and still many people aren’t using it. For those people, I have good news. There is now a much better, easier to understand solution in continuous active learning (CAL) or what some refer to as TAR 2.0.
Why has technology assisted review taken so long to see wide adoption for processing eDiscovery data? To start, change is hard and many people are not going to do it until they have to. Guess what. The time is here. Data volumes continue to increase exponentially, so it is no longer physically or economically possible to process case or investigation data manually. As we pointed out in our recent email analytics post and guide, you have to have robust tools to cull out all of the junk before you can even begin processing the relevant data. Say it with me, “Technology is my friend.”
Unlike TAR where there is a great amount of work on the front end by subject matter experts to find perfect documents in order to create a model used to train the eDiscovery platform, so it can predict the remainder of the documents, CAL builds the model as reviewers are reviewing the documents. When it reaches the point that it can predict the rest, CAL notifies the reviewers and works its magic. There is a little setup on the front end, because you still have to create tags for the reviewers to use and review sets, but that’s no different from an old fashioned linear review of the data, so it really is a magical solution.
And, yes, I use the word magic intentionally. Some of what computers do for us really is magical, and we don’t need to understand every single nuance to appreciate the help. Don’t get me wrong though. I know that eDiscovery is serious business. Part of why TAR took so much time to be adopted is that it had to be tested and approved by the courts, which are very much in the dark ages of technology adoption. However, even the courts have come to accept TAR and its younger sibling CAL.
Fortunately, even magic can often be explained. If you’re using the right solution for CAL, like VenioOne, there are dashboards providing an in-depth visual analysis of your data richness and results that provide the necessary insights to help understand the magic throughout the entire process.
Maybe I’m being optimistic, but I feel like we have reached the tipping point where CAL will soon become the norm. It’s easier for the naysayers to understand and the road has been paved by TAR. Here’s how easy it is:
- Step 1: Project managers or subject matter experts create tags and review sets.
- Step 2: Reviewers review and tag the documents.
- Step 3: VenioOne uses the metadata and content to build a model.
- Step 4: CAL continuously refines the model dynamically as the review progresses and adjusts the documents presented to reviewers.
- Step 5: VenioOne notifies the reviewers that it is ready to predict the rest of the documents.
- Step 6: Project managers or subject matter experts do quality control.
If you compare the CAL process to TAR where a static model is built up front by potentially fewer people or even a single person, it’s the best of both worlds – human and computer combined. The review can start much more quickly with CAL doing the heavy lifting versus the subject matter experts. With CAL, you also have additional checks and balances built in, because CAL’s model is dynamic versus TAR’s static model. CAL can continue to test and refine the model as the reviewer’s are reviewing documents. Now do you see why I’m so optimistic? Say it with me again this time with enthusiasm, “Technology is my friend.”
For more information, check out our CAL starter guide and a comparison of Linear vs. TAR vs. CAL review workflows.