You've recorded the all-hands video. Layoffs, reorg, pivot to a new strategy. You've rehearsed your tone, trimmed the runtime, added a slide deck. Now you're about to send it to 300 employees.
Stop. Test it first.
Most leaders send internal videos cold. They assume clarity because they know the context. Employees don't. They react in real time, second by second, and you won't see those reactions until it's too late.
Facial coding gives you a preview. Show the video to a small opt-in panel, capture their micro-expressions, and you'll know exactly where people get confused, angry, or check out. Then you can re-cut before the real rollout.
Why all-hands videos fail (and why you don't know it)
Internal comms teams obsess over messaging docs and talk tracks. The video itself? Often an afterthought. Someone records it in one take, exports, sends.
The problem: employees don't read your intent. They read your face, your pauses, your word choice. If you look uncertain at 0:47 when you say "we're confident in this direction," they notice. If you rush through the severance details at 2:13, they feel dismissed.
You won't hear about this in the post-announcement survey. People are too worried about their jobs to give honest feedback. But their faces tell the truth in the moment.
What second-by-second emotion timelines catch
Facial coding tracks micro-expressions frame by frame. You get an emotion timeline that shows exactly when your audience reacts and how.
Common patterns in layoff and pivot videos:
- Confusion spikes when you use vague language ("rightsizing," "strategic realignment"). If 60% of your test panel shows confusion at the same timestamp, that's a rewrite.
- Anger or contempt when the tone feels dismissive. Saying "this was a tough decision" while smiling slightly? People catch it. The data will show a contempt spike.
- Sadness or fear during severance details. If sadness holds for 30+ seconds, your panel is overwhelmed. You may need to break that section into two parts or add a reassuring follow-up.
- Engagement drops (neutral affect, looking away) during long explanations. If attention falls at 3:40, cut or reframe that segment.
You can't fix these issues after you've sent the video. You can fix them in the edit.
How to run a pre-test with facial coding
This isn't a full-scale research project. You need speed and signal, not a 200-person sample.
Recruit 15 to 25 employees who match your audience. If you're announcing layoffs in engineering, test with engineers. If it's a company-wide pivot, sample across functions. Make it opt-in and anonymous.
Show them the video once. Capture their facial expressions with front-camera recording (with consent). The AI codes their micro-expressions using the Facial Action Coding System (FACS), which maps muscle movements to emotions like anger, fear, sadness, surprise, and contempt.
You get a second-by-second timeline. Overlay all participants' timelines and you'll see where emotions cluster. If 18 out of 20 people show confusion at the same moment, that's your edit point.
Run this 48 to 72 hours before your planned send date. You'll have time to re-record a section, adjust pacing, or add a clarifying slide.
What to do with the data
Don't try to eliminate all negative emotion. Layoffs are sad. Pivots are uncertain. Trying to engineer fake positivity will backfire (your test panel will show contempt or disbelief).
Instead, look for mismatches between your intent and their reaction:
- If you're trying to reassure but you're triggering fear, your tone or word choice is off.
- If you're explaining a decision but you're causing confusion, you're using jargon or skipping steps.
- If you're showing empathy but you're getting anger, you're not acknowledging the real impact.
Prioritize the biggest spikes. If sadness is high but steady, that's expected. If anger spikes suddenly at one timestamp, that's a red flag.
Re-record or re-edit those moments. Test again if you have time. Two iterations usually get you to a stable emotional response.
Real example: confusion at the "why now" section
A leadership team recorded a 4-minute video announcing a shift from B2B to B2C. They tested it with 18 employees. Confusion spiked at 1:32, right when the CEO said, "Market dynamics have shifted, and we're positioned to capture this opportunity."
The team reviewed the transcript. "Market dynamics" was vague. "Positioned to capture" was jargon. Employees didn't understand why the shift was happening or what it meant for their roles.
They re-recorded that section: "Our B2B revenue has been flat for two quarters. We've seen organic interest from individual users, and we think we can grow faster by serving them directly. This means changes to our product roadmap and some new hires in consumer marketing."
Second test: confusion dropped by 65% at that timestamp. Sadness and fear were still present (expected for a big change), but anger and contempt stayed low. They sent the revised video.
When to test (and when to skip)
Test if:
- The news is high-stakes (layoffs, leadership changes, pivots).
- You're worried about tone or clarity.
- You have 48+ hours before the send date.
- Your audience is large (100+ people) and you won't get real-time feedback.
Skip if:
- The video is purely informational ("here's the new PTO policy").
- You're sending it to a small team and you'll discuss it live right after.
- You don't have time to re-edit (better to send and follow up than delay bad news).
If you're on the fence, test. The cost of getting it wrong (trust erosion, rumor spirals, disengagement) is higher than the cost of a 24-hour delay.
Why leaders resist pre-testing
Three common objections:
Try it: Schedule an EmotionTrac demo and see second-by-second emotion tracking in action. Or visit Innovation for more information.
"We don't have time." A 20-person facial coding test takes 2 to 3 hours to set up and run, plus a few hours to analyze. If you're planning a layoff or pivot, you've already spent weeks on the decision. An extra day to test the comms is worth it.
"We'll look weak if we re-record." No one sees your test panel or your drafts. Employees only see the final video. If it's clear and respectful, they'll trust you more, not less.
"People will be upset no matter what we say." True. But there's a difference between "upset because this is hard news" and "upset because leadership seems clueless or dismissive." The first is unavoidable. The second is fixable.
What happens if you don't test
You send the video. Within an hour, Slack channels light up with confusion or anger. Someone posts a sarcastic reaction. Your head of comms gets 40 DMs asking for clarification. You schedule a follow-up all-hands to "address concerns," which makes it look like you didn't think through the first one.
Or worse: people stay quiet. They don't trust you enough to ask questions. They update their LinkedIn profiles instead.
Pre-testing doesn't prevent all of this. But it prevents the unforced errors, the moments where you accidentally sound tone-deaf or unclear. Those are the moments that break trust.
Want to test your next all-hands video before you send it? EmotionTrac for Innovation gives you second-by-second emotion timelines from real employees, so you can edit with confidence.
How to build pre-testing into your workflow
Make it standard practice for high-stakes comms. When you're planning a layoff or pivot announcement, add "test video with facial coding" to your project timeline, right between "record draft" and "send to company."
Assign someone to recruit the test panel (HR or internal comms, not the exec team). Keep it anonymous. Share the emotion timelines with the leadership team and your comms lead. Make edits. Send.
After a few cycles, you'll get faster. You'll learn which phrases trigger confusion, which pauses feel too long, which expressions read as uncertain. Your first drafts will get better, and you'll need fewer re-records.
But don't skip the test. Even experienced leaders miss things. The camera catches what you can't see in the mirror.
Sources and further reading
- EmotionTrac for Innovation: Facial coding and emotion timelines for product and comms testing.
- Höfling, T. T. A., & Alpers, G. W. (2023). "Moral emotions and facial expressions: A systematic review." Frontiers in Neuroscience, 17. DOI: 10.3389/fnins.2023.1125983