Walk into Keches Law Group on any morning when one of their attorneys is preparing for trial, and you'll be mesmerized by the flurry of activity and the final mile of preparation, usually happening in the conference room that overlooks the scenic Neponset river. The case has been worked. The exhibits are ready. The witnesses have been prepped. And someone, usually Sean Flaherty, the firm's Managing Partner, says around every trial: trials are the lifeblood of this firm. He means it. The settlements that protect Keches's clients exist because the other side knows the firm will try the case. The accountability that drives their results exists because their attorneys believe in juries. The trajectory of a client's life after a serious injury changes because someone, at some point, walks into a courtroom and asks twelve strangers to do the right thing.
But trials are scary, and they are unpredictable, and the lawyers who try them well are the ones who never stop respecting how much can turn on decisions made in the days before voir dire. The story was chosen. The opening was drafted. The witnesses were scheduled. And the twelve people in the box are still strangers deciding someones fate.
MockJury was born out of a desire to help make great attorneys even better. The trial lawyers I've worked with at Keches are the inspiration — their craft, their conviction, the seriousness with which they prepare — and the product exists to give attorneys like them, at firms of every size, a way to spend the limited time they have leading up to trial with something more useful than uncertainty. Not to eliminate the uncertainty, because no honest tool could. But to narrow it. To turn the unpredictable into something a little more predictable, in time to actually act on what you learn.
The cost problem comes first
The traditional answer to "how will a jury react to my case" is a focus group. You hire a consulting firm or do it yourself. People who roughly match the venue and have the time, sign up to participate. Their motivation is already likely biased. They may have an interest in the justice system, but they definitely have an interest in being paid. A half-day session is run in a hotel conference room with one-way glass or more likely, over Zoom. You watch (or moderate, which likely introduces your own biases accidentally). You take notes. Two weeks later you get a report. You get one shot unless you want to spend the time and money to do it again.
The bill is anywhere from two to fifty thousand dollars depending on the firm and the depth of the engagement. For a case with seven-figure exposure, that's a rounding error and worth every penny. For most cases on most plaintiff PI dockets, it's a number that just isn't justified. So the focus group doesn't happen.
Most cases get no jury research at all. Not because the lawyers don't believe in it, but because the math doesn't work. That gap, the distance between "I'd love to know how this lands" and "I can't justify fifteen thousand dollars to find out" is where we live.
Then the speed problem
Even when the math does work, the calendar usually doesn't. A focus group takes weeks to schedule. The trial date doesn't move. By the time the report lands, the opening is already memorized and the witness order is already filed. The research becomes something you review rather than something you can still act on.
What trial lawyers actually need is something they can iterate against. "Try it this way. No, try it that way. What if the contributory negligence framing came up earlier?" That's not a question a focus group can answer in any practical timeframe. It's a question a synthetic panel can answer in a few minutes, several times in an afternoon, until the framing actually lands the way it needs to.
Argyle, Busby, Fulda, Gubler, Rytting, and Wingate published work in Political Analysis in 2023 — "Out of One, Many: Using Language Models to Simulate Human Samples" — demonstrating that large language models conditioned on detailed demographic and ideological backstories produce response distributions that closely mirror those of actual humans matched on the same characteristics. That's part of the underlying research that makes any of this possible. Without it, MockJury is fortune-telling. With it, MockJury is applied research.
Making the unpredictable a little more predictable
Jury behavior is not random. It feels random to lawyers because the sample size of any given trial is twelve, and twelve is not a number that yields patterns. But across thousands of trials, across decades of jury research, patterns are everywhere.
Pennington and Hastie's story model, first published in their 1986 work on the cognitive structure of juror decision-making and developed across two decades after, established that jurors don't weigh evidence the way the rules of evidence assume they do. They construct stories. They take the facts they've heard and organize them into a narrative with a beginning, a motive, and a conclusion, and then they evaluate that narrative against the verdict options the judge has given them. If your case has a story, jurors will find it. If it doesn't, jurors will build one, and you may not like the one they build.
Burger's 1981 analysis on defensive attribution showed that jurors who feel similar to a party (situationally, demographically, in the texture of their lives) attribute responsibility differently than jurors who don't. Hans and Reyna's work on the fuzzy-trace model of damages decisions showed that jurors don't process dollar figures the way the verdict form imagines. They translate numbers into rough qualitative meaning: "a lot," "fair," "outrageous" and that translation depends on each juror's numeracy and worldview.
None of this is mystery. It's research, living in academic journals that most trial lawyers don't have the time to think about let alone apply to their case, and they shouldn't have to. What MockJury does is take that research and operationalize it into a tool that does it for them using the unique attributes of their case.
That response isn't a script. It's what a persona profile that pairs conservative-deferential temperament with high Just World Belief and moderate distrust of large institutions (along with 100+ other variables) sounds like when asked the question cold. A different persona from the same venue would answer very differently. All answers help you frame your case, in a fraction of the time and for a fraction of the cost of traditional mock jury.
What it isn't
An exact prediction. MockJury models how a representative panel from your venue is likely to react to your framing, your questions and help you surface issues you didn't think about. It does not tell you what the actual twelve people in the box will do, because nobody can. It tells you where your case is strong, where it's brittle, and which arguments are landing differently across the panel, fast enough that you can iterate before voir dire rather than after without wasting time and money.
What we owe the clients
On the surface this tool is built for attorneys, but it's really built for your clients. They have the most at stake in any trial and MockJury is another arrow in your quiver to ensuring you set them up for success. If you've read this far, you clearly care about your clients and your own trial preparedness.
We're in private beta with a small group of plaintiff PI firms. If you want to be on the early-access list, the form on the homepage is the way.