How to use Attention Analysis
Attention analysis infers what visitors are looking at using mouse movement, scroll depth, and dwell time — the closest practical proxy for eye-tracking. Here's how to read the signals and act on them.
Attention Analysis
Inferring what visitors are paying attention to from mouse, scroll, and dwell signals — without eye-tracking hardware.
Attention analysis is the practice of estimating where a visitor's focus lands on a page by combining three behavioural signals: mouse movement and hover paths, scroll depth and scroll velocity, and dwell time on individual elements. Together they approximate what an eye-tracker would record, with none of the lab cost or sample-size limits.
It sits inside the broader discipline of behavioral analytics. Where session replay shows you a single visitor's story and heatmaps show you aggregate clicks, attention analysis answers a different question: of everything on this page, what did people actually process before they bounced, scrolled past, or converted?
Eye-tracking studies have shown for years that mouse position correlates with gaze position roughly 70-85% of the time on content-heavy pages. That correlation is the foundation of every attention map you've ever seen — it isn't perfect, but it's good enough to spot the elements your visitors ignore.
For a Shopify or WooCommerce store, that matters more than it does for a content site. If shoppers never read the returns policy on your product page, that's a conversion risk. If they hover over the size chart for eight seconds and then leave, that's a fit problem you can fix with copy.
What attention analysis actually measures
The first signal is mouse trajectory. Most modern attention tools log cursor X/Y coordinates 10-20 times a second and infer two things from the trace: where the cursor lingered (a proxy for reading) and how it moved between elements (a proxy for scanning order).
The second is scroll behaviour — not just how far visitors scrolled, but how fast. A visitor who scrolls past your hero in 400ms is skipping it; a visitor who pauses at 60% scroll depth for six seconds is reading. Scroll velocity is the underrated half of the equation.
The third is element-level dwell time. Modern tools tag DOM elements (image carousels, review widgets, accordions) and measure how long each one is in the viewport with low cursor velocity nearby. That gives you a per-element attention score rather than a fuzzy page-level heatmap.
Attention ≠ engagement ≠ intent
A high-attention element isn't automatically a high-performing element. Visitors stare at confusing things too. Always pair attention data with a conversion outcome — bounce rate, add-to-cart rate, or checkout completion — before drawing conclusions about what to change.
How accurate is mouse-based attention vs eye-tracking?
The honest answer is: accurate enough for testing decisions, not accurate enough for academic research. Mouse-gaze correlation is strongest on pages where the cursor is used for reading aids — hovering over images, dragging size sliders, expanding accordions. It's weakest on pages skimmed passively or scrolled with a trackpad swipe.
The chart below shows how the correlation degrades by page type. Product detail pages and configurator flows track gaze closely; long-form blog posts and landing pages with sticky video track it less well because the eye wanders independent of the cursor.
Mouse-gaze correlation by page type
The practical takeaway: trust attention data most on transactional pages where the cursor does work, and treat blog/homepage attention maps as directional. On a Shopify PDP with a size selector, swatch picker, and quantity input, you're well above 80% — solid ground for a test hypothesis.
Typical attention patterns on e-commerce pages
Across thousands of stores, certain attention patterns repeat. The product image gallery and price block always dominate above-the-fold attention. Reviews capture a second spike if they're visible without scrolling. Returns and shipping copy gets near-zero attention on the PDP itself but spikes during checkout.
The benchmark table below shows typical attention share — the percentage of total page dwell time spent on each element — across the most common DTC verticals. Use it as a sanity check before declaring your own numbers anomalous.
Typical attention share by element across e-commerce verticals (PDP, desktop)
| Page element | Apparel | Beauty | Electronics | Home goods |
|---|---|---|---|---|
| Product image gallery | 38% | 31% | 27% | 34% |
| Price + add to cart | 14% | 12% | 11% | 13% |
| Reviews / ratings | 12% | 19% | 22% | 11% |
| Variant selectors (size/color) | 11% | 8% | 6% | 9% |
| Product description | 9% | 13% | 18% | 14% |
| Specifications / ingredients | 4% | 11% | 9% | 8% |
| Shipping & returns | 3% | 2% | 3% | 4% |
| Cross-sells / related | 6% | 4% | 3% | 5% |
Notice the vertical splits: beauty and electronics buyers spend far more attention on reviews and specs than apparel buyers, who lean on imagery. If your beauty PDP shows apparel-style attention distribution — heavy on imagery, light on reviews — you've likely buried social proof below the fold.
Turning attention data into A/B tests
Attention analysis is most valuable when it generates hypotheses you can test. The workflow is simple: find an element with surprising attention (much higher or lower than the benchmark), form a hypothesis about why, and run an experiment to validate it. The signal points; the test confirms.
A worked example: an apparel store finds that its size guide accordion has 14% attention share — well above the 11% variant-selector benchmark — but add-to-cart rate is below average. Hypothesis: shoppers are stuck on sizing. Test: replace the accordion with an inline size recommender. If attention drops AND add-to-cart rises, the hypothesis held.
Pair attention with intent signals
The strongest hypotheses come from layering attention data over exit points. An element with high attention right before a bounce is almost always a friction point — confusing copy, missing info, or a broken expectation. That's where to test first.
Frequently asked questions
No. A standard heatmap shows where users clicked or moved their cursor in aggregate. Attention analysis goes further by weighting those signals with dwell time and scroll velocity to estimate actual focus, not just activity. Most modern attention tools render their output as a heatmap, but the underlying model is more sophisticated.
It's one technique within behavioral analytics, alongside session replay, funnel analysis, and event tracking. Where event tracking tells you what visitors did and session replay shows you how, attention analysis tells you what they processed in between. Used together, they form a complete behavioural picture.
It depends on the implementation. Heavy heatmap scripts that log every mouse coordinate at 60Hz can add 80-150ms to page load. Modern attention tools sample at lower frequencies and batch events, typically adding under 30ms. Always check Lighthouse scores before and after installing any tracking snippet.
For a PDP getting 500+ daily sessions, two weeks is usually enough to see stable attention patterns. Lower-traffic pages need a month or more. Be cautious of seasonal traffic — attention patterns during a flash sale don't represent your everyday baseline.
Partially. Mobile lacks the cursor signal entirely, so attention has to be inferred from scroll velocity, viewport dwell, and tap targets alone. Mobile attention maps are directional but noticeably less precise than desktop ones — treat them as supplementary, not primary.
It can be, but the implementation matters. Anonymous, aggregated attention data is generally low-risk. Session replay that captures form input or personally identifiable elements requires explicit consent and careful masking. Check that your tool masks input fields by default and respects cookie consent state.
Scroll-depth alone tells you how far visitors got. Attention analysis adds the time and engagement dimension — a visitor who scrolled to 80% in two seconds is very different from one who took two minutes. The combination is what makes the data actionable for testing.
No, and it shouldn't try. Attention data tells you what visitors looked at, not why. A confused stare and a careful read look identical on a heatmap. Pair attention findings with five user interviews or a short on-site survey when you need to understand intent.
For an above-the-fold primary CTA, 8-14% attention share is healthy on a PDP. Below 5% usually means the button is competing visually with other elements (a busy gallery, a promo banner). Above 20% can signal hesitation — visitors are looking at the button but not clicking it, which is its own diagnostic.
If you import historical session data into your analytics tool, you can usually identify the top two or three attention anomalies within the first day. Without historical data, you're waiting on the cold-start period — typically a week of traffic before patterns stabilise enough to act on.
Get an AI expert review of your site
Paste your URL — Metricuno's AI runs the same heuristic checks a senior CRO consultant would, scoring your page and prioritising the fixes that'll move conversion fastest.