At first glance, Formula One racing and ediscovery seem to have little in common. One is adrenaline-pumping and high octane — while the other is fueled by coffee and its only speed records are average words per minute to type in a Boolean search.
But, looks can be deceiving. Ediscovery practitioners face extremely high stakes over a course (comprising millions of documents) as daunting as any race track, and both must rely on the best technology and the right experts to emerge victoriously.
Establishing a tight race strategy, having the most supercharged race car, and an expert team all contribute to taking home the top prize. This multipart series unpacks how ediscovery practitioners can get across the finish line first by finding the right balance of expertise, advanced technology, and optimized workflows beginning with finding the shortest route to winning.
How does DISCO do it?
While the racecar driver is facing down 190 miles of asphalt and the legal practitioner is neck-deep in data, the objective in both cases is to find the quickest route to completion without making any mistakes. DISCO shaves miles off the course to help win the ediscovery race by reducing the volume of data practitioners have to engage with prior to and throughout the document review process. Effectively giving legal practitioners the fastest car on a shorter road has resulted in 60% quicker time to evidence and dramatic cost savings, guaranteed.
What’s the secret sauce? DISCO AI
What is under the hood in the DISCO ediscovery supercar and why is it so much better than legacy technology-assisted review (TAR) tools? DISCO’s AI models are driven by abstract semantic representations (the math underlying language) rather than keywords or phrases. This goes beyond the “bag of words” approach and leverages 300 data points (vectors) to examine the entirety of a document, taking into account the order, meaning of words, and sentence structure to arrive at intelligent insight. As a result, DISCO AI understands correlations like man is to king as woman is to queen as well as differentiate between “man-eating dog” vs. “dog-eating man.”
As the model evolves, the accuracy and results are continually A/B tested to ensure the most accurate model is deployed. The underlying architecture is not just smart — it also gets smart more quickly. DISCO AI capitalizes on both fastText and ElasticSearch to deliver fast, always-on AI for all your tags continuously learning, training, and providing predictions in minutes — not hours or days. The algorithm can begin making informed recommendations with a fraction of the coding decisions required in older ediscovery platforms.
Employing DISCO AI and engaging with the interactive data visualization in DISCO can quickly uncover key custodians, timeframes, and concepts to inform your negotiation process during the 26(f) conference. If you have not yet conducted your own examination of the data with a robust tool like DISCO, how can you have a meaningful discussion about your data with opposing counsel?
On your mark, get set, cull
The best F1 teams master the art of finding the most efficient path to the finish line, taking curves at the apex, and drafting off the competition to eliminate every bit of unnecessary driving to finish first. Similarly, DISCO leverages advanced front-end culling to eliminate noise before the review and triangulate it on the most relevant information at super speed. While DISCO does employ traditional culling parameters (custodians, data ranges, and search terms), the secret sauce lies in the advanced computational power of DISCO AI.
This powerful technology helps legal practitioners identify likely relevant data quickly and amplifies each coding across the dataverse — think of it like drafting off the earlier decisions to amp up your speed. DISCO helps practitioners quickly identify “junk data” and cull it from the review. Finally, DISCO AI identifies when coding decisions are at odds with the suggestion of the algorithm and prioritizes them for future quality control validation. This is like the F1 car having better instrumentation and sensors than the other teams to identify errors and course-correct quickly. In both cases, better real-time information prevents time-consuming errors.
Move the finish closer
If your finish line comes in fewer miles than the competition, the odds of winning greatly improve. DISCO AI prioritized review exhausts the relevant population sooner, with fewer turnovers and less validation required — meaning that review can end early with full confidence that all relevant documents have been uncovered and reviewed.
In the first half of 2020, across the millions of documents reviewed by DISCO Managed Review and leveraging DISCO AI, only 51.16% of documents required human review. Compared to legacy approaches and other TAR 1.0 or even TAR 2.0 workflows, these results are substantially better (benchmarks from industry peers are 65-70%). DISCO has moved the finish line closer for our clients.
So what does an ediscovery supercar look like in action?
Although the fastest to the finish line in ediscovery may not be met with cheering fans and bottles of champagne, finding the most impactful evidence faster can materially impact case outcomes and save you big time.
DISCO AI on a large-scale matter
A client with a large-scale class action came to DISCO needing to make a production from over 22 million documents they had collected. In only three weeks, all relevant documents were identified by reviewing less than 1% of the documents and the team completed the review with an average throughput of 148 documents per hour (triple the industry standard) and at a 90% reduction in cost for the client.
The team started with the foundation, using best practices like keywords and data ranges to refine the dataverse to 5.7M documents. While most providers would have dived into a costly review at this point, DISCO did not stop there. The team applied DISCO AI and layered in insights data visualization to identify priority custodians, concepts, and narrow windows of time. Next, the team focused on the documents with middle of the road relevancy scores (the docs the AI was struggling to categorize) to supercharge the development of the algorithm. This approach quickly eliminated over 5M documents as irrelevant, taking the document population to 484,160. Of the remaining documents, the team had to review only 30% (159,980) before determining with statistical certainty all relevant data was identified.
In the end, the client saved millions because DISCO shortened the road by 99%.
DISCO AI for smaller cases
Many people assume that AI is great for the behemoth cases but maybe not worth the effort or not as impactful for smaller cases. This could not be further from the truth. The advanced multi-vector approach DISCO AI employs enables case teams to see informed predictions in a fraction of the data volume of legacy technology. Since there is no additional cost, it is a no-brainer to leverage DISCO AI.
DISCO was engaged recently by WeWork for a small 32,000 document review on a tight timeline. An offshore provider submitted an estimated cost of $26,000 and the client wanted to know if we could beat the offshore estimate. Leveraging the same processes applied in massive cases, DISCO’s team reduced the review population to just 2,729 docs for review and completed the review in 20 hours and 42 minutes. The client only incurred 5% of the original cost estimate and had important insights into this critical data set in days instead of weeks.
Win every race from 5 to 5,000 miles
In the past, there were only certain cases that could benefit from an AI-powered workflow as a result of limitations in the underlying technology. The underlying technology that powers DISCO AI is much more sophisticated and able to begin making informed and insightful recommendations in a fraction of the time legacy technology requires. In as few as 50 coding decisions, the DISCO AI algorithm can provide informed recommendations. And, unlike with TAR 1.0, there is no special process, additional cost, statistical sampling, or data processing that has to be done to take advantage of AI. Every DISCO case is AI-ready at no cost and the process starts by simply switching on the AI setting in the system.
As with F1 racing, it is not solely about finding the optimal path, but about having the best driver with the fastest car on that path. What is the net result of this supercharged technology? Faster, better-informed reviewers who average review speeds 60% faster than the industry average across the board — and who sometimes achieve superhuman rates as high as 148 documents an hour!
Regardless of your matter type, budget, or the number of documents the benefits of leveraging AI-powered DISCO Managed Review are undeniable. Whether your matter is massive or minuscule, DISCO AI and effective project management can dramatically improve the entire process. Better results, in less time for less money, is a winning proposition whether you are handling a front page of the wall street journal litigation or a more manageable contract dispute. Whatever race you are facing in your next ediscovery matter, DISCO is here to help you get across the finish line.