Try Our Free Review Simulation and Find Out How Much You Can Save
How much can I save using a TAR 2.0 system with Continuous Active Learning? That’s the question that many of our clients were asking before they had a chance to try Insight with Predict for themselves.
They were understandably nervous about using our next-generation technology, particularly in the early years before independent research proved that CAL beat TAR 1.0 systems hands down. But for our clients who’ve tried CAL, seeing is believing. For those of you who’ve yet to see the savings CAL delivers, we’ve created the TAR Challenge and invite you to participate. See for yourself how much time and money you would have saved on your last review.
How Does It Work?
We run a simulation at no cost or obligation to you. Send us the documents and a load file from your last review (one where you paid your team to review all the documents).
How many documents did your team review on the case? Of those, how many were irrelevant? That’s the delta of inefficiency. If your goal is to find relevant documents as quickly as possible, then every irrelevant document your team reviews is a waste of time and money.
Once we get your documents, we will load them into Insight and run a simulation to show you how the review would have gone if you had used Predict.
The process is simple:
- We load the documents into Insight and create a Predict “graph” database.
- At this point, Predict will know nothing about your case, what makes a document relevant, or the tags your team made during the review.
- We then choose a few relevant documents to act as initial “seeds” for training purposes. We can do this any way you like, from choosing a few outright to creating a synthetic document or picking some at random. We don’t really care how you start the process.
- From there, the process is automatic. First, Predict ranks the documents based on your initial training seeds.
- Next, we pull down a virtual batch of documents chosen by Predict, say about 50 of them. Then we tell Predict what the judgments were on those documents (relevant or not-relevant).
- Predict uses the new judgments as additional training seeds for a further ranking.
We go through this process until all documents are reviewed. At this point, we can draw an accurate “yield” curve that will show you how far into the review you had to go to find 80% or 90% of the relevant documents (or any other percentage you decide on).
As the French would say, “Voila.” You now know exactly how many documents you could have cut out of the review and still find an acceptable number of relevant documents. Courts have approved 75% recall back in the days of TAR 1.0 but with Predict we regularly cross the 90% recall level, while reviewing fewer documents than required for TAR 1.0 processes.
You can see your savings on what we call a gain curve. A gain curve provides a simple way to plot the success of a predictive ranking project.
The X axis shows how many documents the team has reviewed, from zero to 100%. The Y axis shows the percentage of relevant documents the team has found.
With a linear review, you can expect the team to find the same percentage of relevant documents as it has reviewed. In other words, if you review 10% of the population, you can expect to find 10% of the relevant documents. Same for 50% or 80%. And, 100 for 100%.
Linear review is represented by the red diagonal line.
With a good predictive ranking engine, the system will move relevant documents to the front of the line. You can see it in the yield curve. The further up and to the left of the linear review line, the more you can save.
Will I Always Save Time and Money?
Good question. The answer is almost always a resounding “Yes.” CAL systems have proven themselves over and over again to beat linear review, often by wide margins. And they beat TAR 1.0 systems as well. Indeed, we are happy to do a simulation against a review using a TAR 1.0 system as well.
What’s the Catch?
There isn’t one really. We are so sure you will save money using predictive ranking, we will show you on our dime. And write up a report that looks a lot like these samples.
When you finally want to learn how much you might save with TAR 2.0 Continuous Active Learning process, give us a call. We are looking forward to showing you what we can do.
- Simulation 1: Insight Predict Reduces Review By More Than 57%
- Simulation 2: Insight Predict Cuts 260,000 Documents from Review
- Simulation 3: Comparing Family vs. Non-Family Review and Expert vs. No-Expert Training