Home / Insights / Emotional engagement score
What is an emotional engagement score? (and how EchoDepth measures it)
An emotional engagement score is a single number, or a line of numbers, that summarises how strongly and how positively an audience responded to a piece of content. In advertising it is usually reported second by second as a curve across the ad, and sometimes as one overall score. Vendors build these scores in different ways: some ask viewers what they felt, some code facial expressions from webcams, and some predict the response with a model trained on past ads, without new viewers. EchoDepth reports two separate measures. The Emotional Response Curve is plotted second by second from the facial movements of consenting viewers, coded against 44 FACS Action Units and converted into valence, arousal and dominance. The Cavefish Emotional Index (CEI) is a 0 to 100 score read from what viewers write or say about the ad afterwards, averaged across five readings. Together they show what people felt while watching and what they took away.
Updated 17 October 2026 by Jonathan Prescott, Founder, Cavefish.
What an emotional engagement score measures
Engagement on its own is a loose word. In media it often means attention or interaction: views, watch time, clicks. An emotional engagement score is narrower. It describes the response itself: whether people reacted, how strongly, and whether the reaction was positive or negative.
Most scores combine two things. Intensity says how much a moment moved people. Direction says whether that movement was pleasure or displeasure. A score that only measured intensity would rate an irritating ad as highly engaging, which is why direction matters.
Reported second by second, the score becomes an emotion curve: a line across the running time of the ad that shows where response builds, where it falls away and how people react at the moments that matter, such as the brand reveal.
How vendors build them
There is no standard emotional engagement score. Each provider builds its own, from different inputs.
| Provider | Headline scores | How emotion is captured |
|---|---|---|
| System1 | Star Rating, which System1 says predicts long-term market share growth from the types of emotions felt; Spike Rating; Fluency Rating | Self-report: in FaceTrace, viewers self-select how the ad made them feel |
| iMotions Predictive AI | Attention, Expressiveness and Valence, scored against norms | Predicted by a model trained on past facial-coding data, with no new respondents per test |
| Element Human | Stopping Power, Staying Power and Brand Uplift, plus Emotional Response and Key Moments | Facial coding among other methods, with emotion broken into seven categories |
| EchoDepth | Emotional Response Curve, with valence, arousal and dominance; Cavefish Emotional Index (CEI) with five readings | Facial coding of 44 Action Units on consenting viewers' webcams for the curve; what viewers write or say after watching for the CEI |
The differences matter when you compare numbers. A self-reported score, a predicted score and a measured facial score can all be called emotional engagement and still disagree about the same ad.
How EchoDepth measures it
EchoDepth keeps apart two things that many scores blend together: what viewers' faces did while they watched, and what they say about the ad afterwards.
The Emotional Response Curve
The curve is built from faces. Each viewer's face is coded on every frame for Action Unit intensity, from 0 to 1. The Action Units are converted into valence, from negative to positive, arousal, from calm to activated, and dominance, from overwhelmed to in control. Valence and arousal are combined for each second and averaged across viewers, so the line rises when a moment is both positive and energetic and falls when it turns negative. A shaded band shows how far viewers agree. How the curve is calculated is set out in the methodology.
The Cavefish Emotional Index
The Cavefish Emotional Index (CEI) is built from words, not faces. After watching, viewers write or say what they made of the ad, and EchoDepth reads those answers on five readings, each scored 0 to 100, where higher is healthier:
- Advocacy: whether they have anything good to pass on.
- Goodwill: how much hostility is in what they said. Higher means less.
- Engagement: whether they are still invested, or have resigned themselves to it.
- Drive: whether they are still trying to fix it themselves.
- Assurance: whether they know where they stand. Higher means less unease.
The CEI is the average of the five readings across everyone in the study. It falls into one of four bands, the same bands every EchoDepth module uses: Strong (80 and above), Good (60 to 79), Watch (40 to 59) and Act (below 40). The definition is the one published on cavefish.ai.
Felt and stated, side by side
Because one measure comes from faces and the other from words, reading them together shows where the two agree and where they part. An ad that lifts the curve but leaves a CEI in the Watch band may have entertained people without leaving them anything good to say about it. A flat curve with a Strong CEI suggests an ad that people liked on reflection more than in the moment.
A worked example
Three seconds into the clip on our homepage, the viewer is smiling. The measured Action Units are AU12 0.94, AU6 0.29, AU4 0.08, AU1 0.01 and AU2 0.04: a lip-corner pull with a cheek raise, the pattern of a felt smile. That raises both valence and arousal, so the response curve climbs at that point. A real study averages the curve across every viewer, 15 or more per audience segment, and reads the CEI from what those viewers say afterwards.
How to read an emotion curve
- Opening seconds
- Does the curve lift early? For social formats, the first two or three seconds deserve a look on their own.
- Peaks
- The seconds where response rises most. Check which Action Units caused them: a cheek raise with a smile (AU6 with AU12) is a felt smile, while a smile alone is often polite.
- Drops
- The seconds where response falls. Lowered brows (AU4) during a busy scene often point to confusion rather than dislike.
- Brand moment
- Response in the seconds around the logo, product or end frame. A peak that has faded before the brand appears is a common fix.
- Confidence band
- How much viewers agree. A wide band at a peak means the audience is split, which is worth reading by subgroup.
- Overall score
- Useful for comparing edits, but it hides the shape. Two ads with the same overall response can have very different curves.
What a score can't tell you
An emotional engagement score describes response, not results. We have not yet published a validation of the response curve or the CEI against in-market sales or brand tracking, and we do not yet publish norms for ads. Until we do, compare edits within a study and against your own earlier ads, rather than reading a single number as good or bad in the abstract. A face is also not a feeling: expressions and inner states are linked but not one to one, which is why the curve is read across many viewers and alongside a survey. The full list of limits is in the methodology.
A score also says nothing about whether the ad will be noticed in the first place. For that, see attention vs emotion in advertising.
Questions
What is a good emotional engagement score?
It depends on how the score is built. On the Cavefish Emotional Index, 80 and above is Strong, 60 to 79 is Good, 40 to 59 is Watch and below 40 is Act. EchoDepth does not yet publish norms for ads, so for the response curve the most useful comparison is between edits in the same study and against your own earlier ads.
Is an emotional engagement score the same as an attention score?
No. Attention scores measure whether and how long people looked. An emotional engagement score measures how they responded while looking. An ad can score well on one and badly on the other, so most teams use both.
How is the CEI different from System1's Star Rating?
System1 says its Star Rating predicts long-term market share growth based on the types of emotions felt, which viewers report themselves through FaceTrace. The Cavefish Emotional Index is read from what viewers write or say about the ad, across five readings, and describes how they feel about it rather than forecasting market share. EchoDepth's second-by-second measure, the Emotional Response Curve, comes from facial coding rather than self-report.
What is second-by-second ad analysis?
It means reporting the response to an ad for each second of its running time rather than as one score at the end. The result is an emotion curve that shows where the ad builds, where it loses people and how they react at the brand moment. EchoDepth's version is the Emotional Response Curve, built from facial coding.
Can I get a response curve and a CEI for my ad?
Yes. Every EchoDepth ad test reports both. Send us your ad, your audience and your deadline through the contact page and we will reply within one working day with a sample plan and a fixed quote. A typical study uses 15 or more viewers per audience segment.
Sources
- EchoDepth methodology, including how the response curve is calculated and its limits. echodepthinsight.com/methodology
- Cavefish. CEI vs NPS: the Cavefish Emotional Index. cavefish.ai
- System1. Ad testing platform. system1group.com
- iMotions. Predictive AI. imotions.com
- Element Human. elementhuman.com
- Russell, J. A., & Mehrabian, A. (1977). Evidence for a three-factor theory of emotions. Journal of Research in Personality, 11(3), 273–294.
- Barrett, L. F., Adolphs, R., Marsella, S., Martinez, A. M., & Pollak, S. D. (2019). Emotional expressions reconsidered. Psychological Science in the Public Interest, 20(1), 1–68.
Descriptions of other providers are taken from their own websites on 17 October 2026 and summarise their main public offer.
Test an ad with EchoDepth
Send us a cut, an animatic or a few edits and we'll reply within one working day with a sample plan and a fixed quote.