How to use Engagement Analysis
A practical guide to engagement analysis for online stores: which signals predict conversion, what benchmarks to hit, and how to turn weak engagement into the next A/B test.
Engagement Analysis
Measuring the depth and quality of on-site interaction — scroll, clicks, video, dwell — to predict conversion intent before it happens.
Engagement analysis goes beyond pageviews and sessions to ask how deeply visitors actually interact with a page. The core signals are scroll depth, interactive-element usage (size pickers, swatch swaps, accordion opens, image-zooms), video play and completion rates, content-block visibility, and active dwell time per section.
Treat it as the leading-indicator layer of behavioral analytics: conversion rate tells you what happened yesterday, engagement tells you which page templates, product pages, and traffic sources are about to convert — or about to bleed budget. It is the bridge between traffic acquisition and revenue, and the input most teams need before they can write a credible A/B test hypothesis.
Most stores measure traffic and conversion, then wonder why the middle of the funnel feels like a black box. Engagement analysis fills that gap by quantifying what happens between the visitor landing and either converting or leaving.
Done well, it answers questions raw analytics cannot: did anyone actually read the size-and-fit block? Does the founder video earn the space it takes? Which PDP sections do high-intent buyers linger on versus skip? Those are the questions that produce testable hypotheses.
What to measure (and what to ignore)
The useful engagement signals fall into four buckets: scroll depth, element interactions, media playback, and active dwell time. Each one maps to a different reader behaviour, and the combination is what reveals intent.
Scroll depth at fixed thresholds (25%, 50%, 75%, page-end) tells you whether the page is being consumed. Element interactions — swatch clicks, size selections, accordion opens, image zooms, review-tab opens — tell you which decision factors matter to this audience. Video play, 50% completion and 90% completion separate ambient autoplay from genuine attention.
Active dwell time is the one people get wrong. Raw time-on-page includes idle tabs and abandoned sessions; active dwell only counts seconds where the visitor was scrolling, moving the mouse, or interacting. On a product page that distinction can be the difference between a 12-second skim and a 90-second consideration.
Stop tracking generic "engaged sessions"
GA4's default engaged-session definition (10+ seconds OR a conversion OR 2+ pageviews) sets such a low bar that 80%+ of sessions qualify. It is not a useful CRO signal. Build your own engagement score from scroll + interaction + active-dwell, weighted to your funnel.
How engagement signals predict conversion
The pattern shows up consistently across apparel, beauty and electronics stores: each engagement layer the visitor crosses roughly doubles their conversion probability. The first scroll past 50% matters more than any single click below the fold.
The chart below tracks conversion rate by engagement tier on a representative Shopify apparel store. "Bounced" sessions barely register; "deep + interactive" sessions — visitors who scrolled past 75%, opened at least one accordion, and used a swatch or size selector — convert at more than 20x the bounce rate.
Conversion rate by engagement tier — Shopify apparel store
The practical takeaway: optimising for scroll-past-50% on a product page typically moves more revenue than tweaking the buy-box, because it expands the pool of visitors who reach the buy-box at all. Engagement is the upstream lever.
Benchmarks by page type and vertical
Engagement benchmarks vary more by page type than by vertical. A category page should be browsed quickly; a long-form PDP for a €180 jacket should reward depth. Compare your numbers to the right template, not to a site-wide average.
The table below is a working reference for online stores between €1M and €15M in revenue. Numbers will shift with traffic mix — a paid-social-heavy store sees lower depth than an organic-search-heavy one — but the relative shape holds.
Engagement benchmarks by page type (online retail, €1M-€15M revenue)
| Page type | Scroll-to-50% rate | Avg active dwell | Interaction rate | Median scroll depth |
|---|---|---|---|---|
| Homepage | 55-70% | 18-28s | 20-30% | 60% |
| Category / PLP | 45-60% | 22-35s | 35-50% | 55% |
| Product page (PDP) | 60-78% | 45-90s | 55-70% | 72% |
| Blog / content | 40-55% | 60-150s | 10-20% | 50% |
| Cart | 85-95% | 15-25s | 75-85% | 90% |
| Checkout step 1 | 90%+ | 20-40s | 95%+ | 95% |
If your PDP scroll-to-50% sits below 50%, the issue is almost always above the fold — price framing, image clarity, or a hero that overstays its welcome. If interaction rate is low but scroll is healthy, the controls (swatches, size pickers, review tabs) are either invisible or unconvincing.
Turning engagement signals into experiments
Engagement analysis is only useful when it produces test hypotheses. The pattern: find a block with high visibility but low interaction (it's seen but ignored), or low visibility but high interaction when seen (it's hidden but effective). Both are clear test candidates.
Worked example: on a beauty SKU PDP, the ingredients accordion sat at 78% scroll-depth visibility but only 6% open rate, while reviews — placed lower — drove 31% opens whenever they were reached. Hypothesis: swap their order. Result on a two-week test: +9% add-to-cart, +4.8% conversion rate. The engagement data wrote the brief.
The hypothesis shortcut
For every PDP block, plot visibility (% of sessions that scroll to it) against interaction rate (% who engage once visible). Anything in the high-visibility / low-interaction quadrant is a redesign candidate. Anything in the low-visibility / high-interaction quadrant is a reorder candidate. That two-by-two replaces about 80% of brainstorming.
Frequently asked questions
Behavioral analytics is the umbrella discipline — it covers everything from session recordings to funnel analysis to event-stream querying. Engagement analysis is a specific layer within it focused on depth-of-interaction metrics on individual pages. Think of engagement analysis as one chapter inside behavioral analytics.
A well-built engagement script adds 8-20KB and runs after the main paint, so impact on Largest Contentful Paint is typically under 50ms. Bloated heatmap-style tools that record every mouse pixel are a different story — those can add 200ms+ and should be replaced with sampled event tracking.
Scroll depth measures how far down the page the visitor reached. Active dwell measures how long they were actively engaged (scrolling, moving, clicking) while there. A visitor can scroll to 100% in three seconds and then leave — scroll depth looks great, active dwell exposes the truth.
For a standard PDP in the €1M-€15M revenue band, aim for 60-78% of sessions reaching the 50% scroll mark, with median scroll depth around 72%. Below 50% scroll-to-half means the above-the-fold is failing to earn the next screen.
Send custom events for scroll thresholds (25/50/75/100), interactive-element clicks, video play and 50%-complete, and a periodic active-dwell ping. Store them in GA4, BigQuery, or your analytics platform. You get the same diagnostic power as a heatmap with a fraction of the page weight.
Only if the video earns its placement. Track play rate, 50% completion, and conversion lift among video-viewers vs. non-viewers. If viewers convert at less than 1.3x the non-viewer rate, the video is taking attention without giving it back and should be moved or removed.
For directional reads on a single page, 1,000-2,000 sessions is enough. For segmenting by traffic source or device, plan for 5,000+ per segment. Engagement metrics are noisier than conversion rate at low volume because they include extreme values (60-minute idle tabs).
As a secondary metric, yes — it catches problems weeks before they show up in conversion rate. As a primary KPI, no. Goodhart's law applies: optimise the team on engagement and they will ship attention traps that don't convert. Pair it with revenue per visitor.
Engagement scoring by ad set is one of the fastest ways to spot misaligned creative. If a Meta ad set drives high traffic but the median scroll depth is half the site average, the landing page-creative match is broken and ROAS will follow within days.
55-70% of sessions should fire at least one interactive event on a standard product page (swatch, size, accordion, zoom, review tab). Below 40% suggests the interactive elements are either hidden, unclear, or visually inert — usually an icon-only design problem.
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.