Attention heatmaps have become standard in creative review, and they are widely misunderstood. Teams present them as though a room full of people watched the ad and a camera recorded their eyes. That is not what happened.
This is an explainer of what the technology actually does, what the colours mean, and where the output stops being trustworthy. It sits inside pre-spend creative intelligence: tools that improve the launch decision before media budget is committed.
A worked video example: opening, demonstration and end card
This diagram explains how to review a map. It is not an analysis result or a claim about where actual viewers looked.
Imagine a skincare ad that opens on a dropper, demonstrates application and ends with a branded pack shot. The useful question is not simply “where is the brightest patch?” It is “does the likely focal area support what this moment needs to communicate?”
| Moment | Inspect alongside the map | Possible editing decision |
|---|---|---|
| Opening | Is the focal object meaningful, or does decorative motion obscure the benefit? | Preserve the useful visual, but bring the benefit into the opening if it is missing. |
| Demonstration | Can the viewer connect the action to the product and claim? | Simplify competing overlays; keep the product-use evidence readable. |
| End card | Are brand, offer and next action legible at phone size? | Increase contrast or reduce competing elements if the action is lost. |
These are conditional recommendations. A reviewer still needs to inspect the actual frames, brief and placement before choosing an edit. A bright region over a bottle does not establish that someone read its label, recognized the brand or wanted to buy it.
For video, report coverage as carefully as the map
A map from one frame cannot describe an entire scene, and a few selected scenes cannot establish attention throughout a video. Record the exact timestamp and whether the frame represents the opening, a scene transition, the product demonstration or the call to action. If a scene has no available map, call it a coverage gap instead of implying it was analyzed.
Where multiple frames are available, inspect them in sequence. An isolated attention peak may disappear before the benefit becomes readable. Avoid comparing colors across frames as if they share an absolute scale unless the underlying method actually uses one. Predicted saliency is not an observed retention curve.
Turn the finding into a reviewable brief
Write four things: the timestamp to change, the specific visual or wording to edit, the element to preserve and how the team will judge the revision. For example: “In the opening, move the benefit beside the product demonstration; preserve the demonstration itself; check legibility at mobile viewing size.” Do not invent an uplift percentage.
See the RoastIQ workflow, watch the product demo, or discuss a review of your own creative. For the difference between model evidence and customer evidence, read how to interpret a creative benchmark.
What an attention heatmap is
An attention heatmap is a colour-coded overlay on a frame. Warm areas (red, orange) mark zones predicted to draw high visual attention. Cool areas indicate lower relative predicted saliency, not proof that nobody will look there.
The overlay comes from a saliency model. The model was trained on datasets where researchers recorded real human eye movements across thousands of images, then learned the visual patterns that predict where gaze concentrates: contrast, faces, motion, text, edges, and isolation against a background.
When you feed it a new frame, it produces a probability surface. Nobody looked at your ad. The model is estimating, from learned visual regularities, where people would probably look.
Prediction versus measurement
This is the distinction that matters, and it is the one most often blurred in sales material.
Eye tracking is measurement. Real participants, a calibrated camera, recorded fixations. It produces ground truth about those specific people in those specific conditions, and it costs time and money to run.
Saliency prediction is inference. A model estimates a likely attention distribution without recruiting participants for that specific creative. It generalises from population-level patterns, so it has no opinion about any individual viewer.
Both are useful. They are not interchangeable, and a heatmap presented as eye tracking overstates the evidence behind a decision. There is a fuller account of this boundary in visual attention prediction vs. eye tracking: what we can and cannot claim.
How to use a RoastIQ attention map
RoastIQ combines creative diagnostics with available predicted saliency evidence. Inspect the timestamp and source frame shown with each map. Coverage depends on the evidence available in that report; a map on a selected frame does not mean that every frame or scene was analyzed.
Use the overlay alongside the original creative and the written diagnosis. If an opening, demonstration or end card is missing from the available attention evidence, record that gap. Do not extrapolate a selected frame into an observed retention curve or a guaranteed effect on the score.
What the heatmap answers well
- Is the brand cue in a high-attention zone? A logo in a low-saliency region is worth checking for visibility; the map alone does not prove that viewers missed it
- Does the product reveal land where attention already is? If the reveal happens in a low-attention frame region, the sequencing is fighting itself
- Is attention concentrated or scattered? Several strong regions may compete for attention; inspect whether they support or distract from the intended message
- How does attention move over time? The timeline view matters more than any single frame, because ads are watched, not viewed
What the heatmap cannot answer
- What one person actually looked at. The output is a distribution, not an observation
- Why attention landed somewhere. The model predicts location, not motivation. A face draws attention whether or not it helps your message
- Attention in a real feed. The model sees your frame in isolation. It does not know what sat above or below it while someone scrolled
- Audio-driven gaze shifts. A visual model does not account for a sound cue pulling the eye
- Whether attention converts to anything. Attention is a precondition for the message landing, not a substitute for the message being good
That last point is where heatmaps get abused. A creative can score beautifully on predicted attention and still fail, because attention was never the problem. Heatmaps diagnose visibility. They say nothing about persuasion.
How to read one without over-reading it
Treat the heatmap as a strong prior, not a verdict. A useful sequence:
- Look at where the warm zones sit before you look at scores
- Ask whether the elements that carry the commercial job (brand, product, offer) sit inside them
- Compare the available timestamps for changes in the predicted focal regions, and note uncovered scenes
- Only then compare against the Beat the Skip and Get Noticed scores
If the overlay and the scores disagree, inspect their inputs, timestamps and definitions. Neither is observed viewer behavior, and disagreement is a reason to investigate rather than choose the more convenient result.
What validation actually exists
Saliency models are benchmarked on academic datasets such as MIT300 and SALICON, where performance means agreement with aggregated human fixation data on those datasets. That is a legitimate measure. It is not a claim about accuracy on your particular creative, and it transfers less cleanly to advertising content with rapid cuts, heavy text, or unusual composition.
Saliency validation and downstream creative-score validation answer different questions. Saliency benchmarks compare predicted maps with recorded fixations. SaliencyLab’s documented platform-native attribute model is evaluated against TikTok CTR and YouTube views; those results do not validate the five-KPI composite or establish heatmap accuracy on a specific ad.
The validation and benchmark methodology gives the dated cohort sizes, held-out results and fitted confidence intervals separately. None of these results establishes sales, ROAS, brand recall or a causal improvement from a recommended edit.
The practical position
An attention heatmap is a cheap, fast, repeatable answer to a narrow question: where is this creative likely to be looked at, and is that where the important things are?
Asked that question, it is genuinely useful and worth running on every cut. Asked to predict business outcomes, it will happily produce a colourful picture that means nothing of the sort.
If you want to see one in context rather than in the abstract, the RoastIQ walkthrough shows how creative evidence supports a review.
A coverage-first reading aid
The diagram explains sampling; it is not a generated heatmap or recorded gaze. TranSalNet’s original paper describes an image-saliency model. Applying image predictions to selected video frames does not add evidence for the intervals between those frames. Use the scene worksheet to record coverage, observations and proposed edits.