How to Test Expert Witness Explanation Videos Before Deposition or Trial

Why expert videos fail in front of juries

Your expert spent 40 hours preparing a video that explains how the brake system failed. The animation is polished. The script was reviewed three times.

Then you play it for the jury and watch six people check out at the 90-second mark.

You don't know why. The expert used simple language. The visuals matched the narration. But somewhere between the CAD rendering and the metallurgy explanation, you lost them.

Traditional focus groups catch some of this, but people are polite. They nod. They say "it made sense" because they don't want to look stupid. Facial coding (FACS) measures what their face does when they think nobody's watching: confusion at second 34, boredom at second 112, a brief spike of interest at second 203 that you can replicate.

The workflow

1. Define your proxy audience

You're not testing lawyers. You're testing people who match your venue's jury demographics.

If you're in a suburban county with a median age of 52 and a mix of college-educated and high-school-only adults, your panel should reflect that. EmotionTrac Legal lets you filter opt-in panelists by age, education, geography, and income. You can run 40 people in two days.

Most teams test 30-50 panelists. Smaller panels (15-20) work for early drafts when you're just hunting for catastrophic confusion. Larger panels (60-80) make sense if the case has eight figures on the line and you want tight confidence intervals on every scene transition.

2. Upload the expert video and set your questions

You upload the MP4. Panelists watch it once, front camera on, while the platform captures facial action units in real time. No surveys during playback (that's a cognitive load problem). You can add a few multiple-choice questions after: "Which part was hardest to follow?" or "Did the animation help or distract?"

The AI codes seven core emotions (joy, surprise, sadness, anger, fear, disgust, contempt) plus attention and confusion markers. You get a second-by-second graph that shows when the room goes cold.

3. Review the emotion timeline and flag problem zones

You're looking for three patterns:

Confusion spikes. If 60% of panelists show furrowed brows and head tilts at second 78, that's your jargon trap. Go back to the transcript. You'll find the expert said "strain-rate sensitivity" without defining it, or introduced a third variable without a visual cue.

Attention drops. Flat affect and reduced eye contact mean people zoned out. This happens during long static slides or when the expert repeats the same phrase structure four times in 90 seconds. Attention drops aren't always fatal (people recover), but if the drop happens right before your key conclusion, you have a sequencing problem.

Negative emotion clusters. A 12-second stretch where 40% of panelists show anger or disgust is either a sign that your visuals are too graphic (good to know before trial) or that the expert's tone sounds condescending (fixable with a re-record).

4. Drill into individual panelist video

The aggregate timeline tells you where the problem is. Individual playback tells you why.

You can watch any panelist's video synced to the emotion graph. If confusion spiked at second 112, you scrub to that moment and watch their face. Maybe they squinted at the screen because the font was too small. Maybe they looked away because the animation moved too fast. Maybe they leaned in, replayed the last five seconds in their head (you'll see a micro-expression of concentration), then gave up.

You do this for 5-8 panelists in the problem zone. Patterns emerge. If seven people squint at the same moment, it's a visual design issue. If seven people look confused after the same sentence, it's a scripting issue.

5. Edit and retest the weak sections

You don't reshoot the whole video. You fix the 30-second segment where confusion peaked.

Maybe you add a two-second pause after the expert introduces the term "fatigue crack propagation." Maybe you replace the CAD wireframe with a labeled photo. Maybe you cut the entire metallurgy detour because it added 90 seconds of boredom and contributed nothing to your liability theory.

Then you test the revised segment with a fresh panel (20 people is enough for a revision check). If confusion drops and attention holds, you're done. If the problem moved to a different timestamp, you iterate again.

What the data shows

Most expert videos lose 30-40% of attention in the first two minutes. The drop happens when the expert front-loads credentials or spends 90 seconds on methodology before explaining what they're about to show.

Jurors don't care about your expert's CV until they understand why the expert matters to them. If you reorder the video (start with the conclusion, then show the evidence, then explain the method), attention stays above 70% for the first four minutes.

Confusion spikes cluster around three triggers: unexplained acronyms, visual-audio mismatch (the expert says "left side" while pointing to the right side of the screen), and assumption gaps (the expert assumes everyone knows what a "cam lobe" is because it's obvious to engineers).

The fix rate is high. Teams that retest after edits see confusion drop by 50-70% on the second pass. The video doesn't need to be perfect. It needs to be clear enough that jurors can follow the thread from A to B without rewinding in their heads.

Common mistakes

Testing too late. If you test the expert video two days before trial, you don't have time to reshoot or re-edit. Test at the draft stage, when the expert is still available and the editor can turn fixes in 48 hours.

Ignoring the first 15 seconds. If panelists look confused or bored in the opening, they'll mentally check out even if the rest of the video is strong. The first 15 seconds set expectations. If the expert opens with "I'm going to explain the failure mode of the brake caliper assembly," you've told people this will be boring. If the expert opens with "Here's what happened in the three seconds before the crash," you've told them this will matter.

Over-testing demographics that don't match your venue. If your trial is in a rural county and your panel is 80% urban college graduates, the data will mislead you. A video that works for a high-education panel might bury a jury that's 60% high-school-only. Match your panel to your venue or you'll optimize for the wrong room.

Skipping the post-video questions. The emotion graph shows where people got lost. The post-video survey tells you whether they recovered. If someone looked confused at second 90 but still answered your comprehension question correctly, the confusion was temporary. If they looked confused and then failed the comprehension check, the confusion was structural.

Testing only one expert when you have three. If you're calling a biomechanics expert, an accident reconstructionist, and an economist, test all three videos. The confusion patterns will be different. The biomechanics video might lose people on anatomy terms. The economist video might lose people on chart design. You can't assume one fix applies to all three.

Sources

• Höfling, T., & Alpers, G. W. (2023). Automated facial action coding system validation. Frontiers in Neuroscience, 17.
• EmotionTrac Legal