Home / Insights / Attention vs emotion
Attention vs emotion in advertising: what each measures and when you need both
Attention and emotion measure different things. Attention measurement tells you whether people looked at an ad and for how long: eye tracking follows where their gaze lands, and attention vendors model how much viewing a format or placement typically earns. Emotion measurement tells you how people responded while they were looking: facial coding reads the facial movements that make up an expression, and surveys ask people what they felt. An ad can hold attention and leave viewers cold, or move the few people who notice it. Attention is the gate, because an ad nobody sees cannot work. Emotion is what the brand takes away once someone is watching, and long-running analysis of the IPA Effectiveness Databank links emotional brand associations with long-term brand effects. Most teams need both: attention data to buy and place media well, and emotion data to judge and improve the creative itself.
Updated 17 October 2026 by Jonathan Prescott, Founder, Cavefish.
What attention measurement captures
Attention measurement answers one question: did people actually look at the ad, and for how long? It grew out of the gap between an ad being served or viewable and an ad being seen.
Eye tracking
Eye tracking records where a person's gaze falls on the screen, frame by frame. It used to need lab hardware. Today it can run on an ordinary webcam: Lumen says its software turns a phone or computer webcam into an eye-tracking camera, calibrates to each person's eyes and processes everything on the device. It also logs what was on screen, so it can separate what was viewable from what was actually viewed.
Attention models and media metrics
Eye-tracking panels are small, so attention vendors use them to train models that estimate attention across placements nobody watched in the panel. Amplified Intelligence reports active and passive attention, which together make up total attention, along with attention second by second and a predictive product, AttentionAI. Adelaide reports a 0 to 100 media-quality score called AU, built from a machine-learning model trained to outcomes. These metrics are mostly used to plan and buy media: which formats, sites and placements earn the most attention per pound.
What emotion measurement captures
Emotion measurement answers a different question: once someone is watching, how do they respond, and when?
Facial coding
Facial coding scores the facial muscle movements that make up an expression, defined as Action Units in the Facial Action Coding System (FACS), on every frame while the viewer watches. EchoDepth converts those Action Units into valence (positive to negative), arousal (calm to activated) and dominance, and reports them second by second. The method and its limits are set out in our methodology.
Self-report
Surveys ask viewers what they felt. System1 describes FaceTrace as showing the second-by-second emotional response to a creative, with viewers who "self-select how the ad made them feel". Self-report is cheap and easy to explain, but it records what people say they felt, after the fact.
Predicted emotion
Some tools now predict emotional response without new viewers. iMotions Predictive AI scores Attention, Expressiveness and Valence against norms without recruiting a panel for each test, with a median turnaround of under three minutes from upload, according to iMotions.
Attention vs emotion, side by side
| Question | Attention measurement | Emotion measurement |
|---|---|---|
| What it answers | Did people look, and for how long? | How did they respond while looking? |
| How it is captured | Eye tracking, in a lab or by webcam, plus models trained on eye-tracking panels | Facial coding by webcam, surveys, or models trained on past responses |
| Typical output | Attentive seconds, attention curves, media-quality scores | Emotion curves, valence and arousal, emotion indices |
| Main use | Planning and buying media, judging formats and placements | Judging and improving the creative itself |
| Blind spot | Can't tell a delighted viewer from an irritated one | In a forced-viewing test, can't tell whether the ad would be noticed in a real feed |
Why one without the other misleads
Attention without emotion
A loud, confusing or irritating ad can hold the eye. Attention data would score it well: people looked, for a long time. Facial coding would show something else, such as lowered brows (AU4) through a cluttered product demonstration, or a flat response at the brand reveal. Long attention on a scene nobody enjoyed is not a win.
Emotion without attention
The reverse also happens. A slow-building film can move the people who watch it to the end, but in a social feed most people may scroll past before the emotional payoff arrives. Pre-tests, including EchoDepth's, ask viewers to watch the ad, so they measure response given attention. They cannot tell you whether the ad would stop someone scrolling. That is a question for attention data and in-platform results.
What the effectiveness evidence says
In The Long and the Short of It, a meta-analysis of 996 campaigns from the IPA Effectiveness Databank, Les Binet and Peter Field link emotional brand associations with long-term brand preference and rational messages with short-term behavioural response, and recommend balancing the two. Attention gets an ad the chance to do either. Emotion largely decides what it does with that chance.
How to use them together
- Test the creative for emotion before launch. Use facial coding and a short survey on the finished cut or animatic to find the weak seconds, check the brand moment and choose between edits.
- Check the opening seconds separately. For social formats, read the first two or three seconds of the emotion curve on their own. A flat opening is a warning sign for scroll-past, even before you have attention data.
- Use attention data to plan media. Choose formats and placements that earn enough attention for the ad's emotional arc to land. A 30-second story needs a placement where people actually watch for 30 seconds.
- Verify in market. After launch, attention and in-platform data show whether the ad is being seen. Brand tracking shows whether it is working.
Where EchoDepth fits
EchoDepth measures facial response, not gaze. We code 44 Action Units on consenting viewers' webcams and report valence, arousal and dominance second by second as an Emotional Response Curve, alongside the Cavefish Emotional Index, which is read from what viewers say afterwards. We do not track where on the screen people look. If you need gaze or attention data, pair an EchoDepth study with an attention provider. Some platforms, such as RealEye, combine webcam eye tracking with basic facial coding, but cover far fewer facial movements.
More on how a study runs is on the ad testing page, and on how the index is built in what an emotional engagement score is.
Questions
Is facial coding the same as eye tracking?
No. Eye tracking records where on the screen a person looks. Facial coding records how their face moves while they watch, which shows how they respond. Both can run on a webcam, but they answer different questions: attention versus emotional response.
Which matters more for an ad, attention or emotion?
You need both, in that order. An ad that is not seen cannot work, so attention is the gate. Once it is seen, emotion largely decides what the brand gets from it. Binet and Field's analysis of the IPA Effectiveness Databank links emotional brand associations with long-term brand effects.
Does EchoDepth measure attention?
No. EchoDepth measures facial response: 44 FACS Action Units, converted into valence, arousal and dominance and plotted as an Emotional Response Curve. It does not track gaze. If you need attention data, pair an EchoDepth study with an attention provider.
What are active and passive attention?
They are terms used by the attention vendor Amplified Intelligence, which reports active and passive attention as the two parts of total attention. Check each vendor's own definitions before comparing numbers, because attention metrics are not standardised.
Can a pre-test tell me if my ad will stop the scroll?
Only partly. Pre-tests ask viewers to watch the ad, so they measure response given attention. The emotion curve in the first two or three seconds is a useful warning sign, but whether people stop scrolling in a real feed is best judged with attention data and in-platform results.
Sources
- Binet, L., & Field, P. The Long and the Short of It: 10 key principles of success. IPA. Meta-analysis of 996 campaigns, 700 brands and 83 categories. IPA presentation (PDF)
- Lumen Research. How does the eye-tracking work? lumen-research.com
- Amplified Intelligence. Attention data. amplified.co
- Adelaide. Attention metrics (AU). adelaidemetrics.com
- System1. Ad testing platform. system1group.com
- iMotions. Predictive AI. imotions.com
- Höfling, T. T. A., & Alpers, G. W. (2023). Automatic facial coding predicts self-report of emotion, advertisement and brand effects elicited by video commercials. Frontiers in Neuroscience, 17, 1125983.
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.