Every second of your ad, measured on real faces.
EchoDepth is ad pre-testing with facial coding. Real viewers watch your ad at home on their own webcam, and we show you the moments where it lands, drags or confuses before you buy the media.
Action Units, this frame
Neutral. Watching.
This viewer is AI-generated, not a panel member. The Action Unit values are real measurements of this clip, taken frame by frame with an open-source face model, so you can see the kind of output a test produces. Our production pipeline, the Emotional Response Curve and the Cavefish Emotional Index are explained in the methodology.
What is ad pre-testing with facial coding?
For brand and insight teams, creative agencies and media planners who need to know if a cut works before it airs.
Ad pre-testing with facial coding shows a finished ad, animatic or set of edits to a sample of real viewers before launch and measures how their faces respond, moment by moment. Each viewer watches at home on their own webcam, after giving consent. Software reads the facial muscle movements that make up an expression, known as Action Units in the Facial Action Coding System (FACS), on every frame of the ad. Those movements are converted into valence, arousal and dominance and averaged across every viewer in the study, 15 or more per audience segment, so you can see second by second where the ad builds enjoyment, where interest drops and how people react when the brand appears. Because the result is measured from real people rather than predicted by a model trained on past ads, every point on the curve can be traced back to the viewers and facial movements that produced it.
EchoDepth plots that response as the Emotional Response Curve, second by second, and scores what viewers say about the ad afterwards with the Cavefish Emotional Index (CEI), a 0 to 100 measure banded Strong, Good, Watch or Act. The methodology sets out exactly how both are worked out.
Measured, not predicted.
Predicted response
A growing number of ad testing tools score your creative with no viewers at all, using models trained on past ads. That is quick and useful for screening lots of rough ideas. But nobody actually watched your ad, and the score can't tell you why.
Measured response, with EchoDepth
A sample of real people matched to your audience watches the cut. We record what their faces do, frame by frame, and plot it against your timeline. Every score breaks down to the moments and Action Units behind it, so you can see why a scene works and decide what to change.
How an ad test works
Everything runs remotely in the viewer's browser. No lab, no app, no hardware.
Send us the ad
A finished cut, an animatic or several edits. TV, online video, vertical social formats or cinema.
We recruit viewers
From our viewer panel, matched to your audience on age, gender, region and category use.
They watch at home
Each viewer gives consent, then watches on their own webcam. We keep the scores, never video of their face.
You get the readout
The Emotional Response Curve, the Cavefish Emotional Index, peak and drop moments and the brand moment, with a debrief call.
What's in the report
Every number traces back to a moment in your ad and the facial movements behind it. Nothing is a black box.
- Emotional Response CurveSecond by second, from faces
- How viewers' faces responded at every moment of the ad: valence, arousal and dominance from 44 FACS Action Units, averaged across viewers. How the curve is built
- Cavefish Emotional IndexCEI, 0 to 100
- How viewers feel about the ad afterwards, read from what they write or say. The average of five readings (Advocacy, Goodwill, Engagement, Drive, Assurance), banded Strong 80+, Good 60–79, Watch 40–59 or Act below 40. About the CEI
- Valence
- Positive to negative. Where the ad pleases, and where it irritates or confuses.
- Arousal
- Calm to activated. How much energy a moment generates, whether good or bad.
- Dominance
- In control to overwhelmed. Low dominance often marks a busy or confusing scene.
- Peaks and drops
- The seconds where response rises or falls most, with the Action Units that caused it.
- Brand moment
- How viewers respond in the seconds around your logo, product or end frame.
- Day-after response
- A short check-in the next day: how viewers feel about the ad and the brand once the moment has passed, and what they still remember.
- Stated vs felt
- The Cavefish Emotional Index (what viewers said) set against the Emotional Response Curve (what their faces did while watching).
- Action Unit detail
- Any score breaks down to the facial movements behind it, such as AU6 with AU12, the cheek raise that marks a felt smile.
What can you test?
Anything a viewer can watch in a browser, from a storyboard animatic to the final cut. Testing two or three edits side by side is the most common brief.
- TV and connected TV
- 30s, 20s and 10s cuts, tested in full and as cut-downs.
- Online video
- YouTube pre-roll, bumpers and brand films.
- Vertical social
- TikTok, Instagram Reels and YouTube Shorts, shown at native aspect ratio.
- Cinema
- Long-form spots where the emotional arc matters most.
- Animatics and storyboards
- Test the idea before you shoot it.
- Stills and packaging
- Key visuals, outdoor and pack designs. FMCG testing
Is facial coding legal for ad testing?
Yes, when viewers give informed consent. Every EchoDepth viewer agrees to a specific study before their webcam turns on, and can stop at any time.
Video of the viewer's face is processed as it streams and is not stored. We keep Action Unit and emotion scores, linked to a panel ID rather than a name.
The EU AI Act, plainly
The EU AI Act treats facial coding as emotion recognition. It bans emotion recognition in workplaces and schools. That is why EchoDepth only codes the faces of consenting research viewers, never employees or students. Our workplace products analyse written language only.
How much does ad testing cost?
We quote per study. The price depends on how many ads or edits you test, how many viewers you need and how hard your audience is to reach.
Tell us about your ad, your audience and your deadline and we will reply within one working day with a fixed quote. If you test regularly, an annual plan gives you a set number of studies at a lower rate per test.
Become a paid ad tester
Join the EchoDepth tester panel and we'll invite you to studies that match your profile. You watch a few short ads at home with your webcam on, answer some quick questions, and get paid when you finish.
- Who can joinAnyone 18 or over with a laptop or desktop computer and a webcam.
- What you doWatch ads in your browser, then answer a short survey. Most studies take 10 to 20 minutes, and some send two quick questions the next day.
- What we recordYour webcam feed is analysed as you watch to measure expressions. We don't keep video of your face.
- How you're paidEach invitation shows the payment before you start. You're paid for every study you complete.
- Most wanted nowTeachers, people living in Wales (Welsh speakers especially) and sports fans. Everyone 18 or over is welcome to join.
- Your choiceYou can decline any study, stop part-way, or leave the panel and have your data deleted.
Questions
For brands and agencies
What is ad testing?
Ad testing shows an ad to a sample of people from its target audience before or after launch to find out how it works. Pre-testing happens before the media spend, so you can choose between edits, fix weak moments or drop an ad that doesn't land. EchoDepth pre-tests ads by measuring viewers' facial response second by second, alongside a short survey.
How does facial coding work for ads?
Each viewer watches the ad on their own webcam. Software detects the face on every frame and scores the facial muscle movements defined in the Facial Action Coding System (FACS), such as AU12, the lip-corner pull of a smile. EchoDepth turns those Action Units into valence, arousal and dominance, then averages across viewers to give a response curve for every second of the ad.
Is facial coding accurate?
Automated coding of Action Units is well established, and peer-reviewed work has found that automatic facial coding of people watching commercials relates to their own ratings of the ads and brands. But a facial expression is not a direct readout of a feeling, and people watching alone often show subtle expressions. That is why EchoDepth reports results across many viewers, with the Action Unit detail behind each score, and pairs it with a survey. The UK ICO has warned that emotion AI can be immature, so we publish our method and its limits.
How many viewers do I need to test an ad?
EchoDepth recommends 15 or more viewers per audience segment for each ad or edit. Every group you want to report on separately, such as under-35s and over-35s, is its own segment and needs its own 15. We agree the sample with you based on your audience and the decision you need to make.
What's the difference between facial coding and eye tracking?
Facial coding measures emotional response: how people feel as they watch. Eye tracking measures visual attention: where on the screen they look. They answer different questions and are often used together. EchoDepth measures facial response.
What is the Cavefish Emotional Index (CEI)?
The Cavefish Emotional Index (CEI) is a 0 to 100 score of how people feel, read directly from what they write or say. It is the average of five readings, each 0 to 100 where higher is healthier: Advocacy, Goodwill, Engagement, Drive and Assurance. Scores fall into four bands: Strong (80+), Good (60–79), Watch (40–59) and Act (below 40). In an ad test, the CEI is read from what viewers say and write about the ad after watching. It is separate from the Emotional Response Curve, which shows second by second what their faces did.
Does EchoDepth store video of viewers' faces?
No. The webcam feed is processed as it streams and video is not stored. We keep Action Unit and emotion scores, linked to a panel ID rather than a name.
For testers
How do I become a paid ad tester?
Fill in the short form on this page to join the EchoDepth tester panel. When a study matches your profile, we email you an invitation showing what it involves and what it pays. You watch the ads on your laptop or desktop with your webcam on, answer a few questions, and get paid when you finish.
Is getting paid to watch ads legit?
With EchoDepth, yes. It's free to join and we will never ask you to pay anything. We're Cavefish Ltd, a UK company (number 15127122) registered with the Information Commissioner's Office (ZB915633). Each invitation tells you the payment before you start.
Does the webcam record me?
Your webcam feed is analysed while you watch so we can measure facial movements, such as a smile or raised eyebrows. Video of your face is not stored, and the advertiser never sees you. We keep the measurements, linked to a panel ID rather than your name.
How much do ad testing studies pay?
Every invitation shows the payment and the expected length before you agree to take part. You are paid for each study you complete.
Who can join the tester panel?
Anyone 18 or over with a laptop or desktop computer and a working webcam. Right now we especially want teachers, people living in Wales and sports fans. We sometimes exclude people who work in advertising, marketing, PR or market research, or live with someone who does, because they watch ads differently from most people.
How long does a study take?
Most studies take 10 to 20 minutes: a few short ads followed by some quick questions. Some studies also send two short questions the next day, to see how you feel about the ad once the moment has passed. That takes a minute.
Can I leave the panel?
Yes, at any time. You can decline any study, stop part-way through, or ask us to delete your account and data. We keep your readings (the measurements of your expressions and your answers), never video, and we delete them when you leave the panel.
Other research from Cavefish
The same emotion science, applied outside advertising.