Session replay analysis is one of the most misused tools in conversion optimisation. Teams install Hotjar or Microsoft Clarity, watch a few recordings, and then either over-interpret what they see or stop looking altogether. Heatmaps get screenshotted for quarterly decks and rarely acted on. At Choco Media, we use session replay analysis as a core part of every CRO engagement — but the value comes from knowing what to look for, not from the volume of recordings you watch. This guide covers exactly that.
If you run a website with at least 5,000 monthly visitors and you’re trying to improve conversion rates without guessing, this is for you. You’ll leave with a prioritised checklist of patterns to hunt for, an understanding of what heatmaps can and cannot tell you, and a repeatable process for turning qualitative insight into testable hypotheses.
We’ll cover both tools in tandem — heatmaps and session replay analysis work best together, not in isolation. By the end you should be able to run a focused review session in under two hours and come away with at least three actionable hypotheses.
Why most teams waste their session replay tools
The first mistake is volume. Watching hundreds of recordings without a question in mind produces noise, not insight. You end up with a general feeling that “users seem confused on the checkout page” but nothing specific enough to test. The second mistake is treating heatmaps as truth. Click maps and scroll maps are aggregates — they compress thousands of individual behaviours into a single image, which can hide as much as they reveal.
The third and most common mistake is confirmation bias. Teams already believe a particular element is the problem before they open the tool, and they watch recordings until they find one that confirms it. That is not research — it is storytelling.
- Start with a question, not a tool. “Why is the add-to-cart rate low on mobile product pages?” is a question. “Let’s check the heatmaps” is not.
- Define what you’re looking for before you press play. Rage clicks, dead clicks, scroll depth, form abandonment — pick one pattern per session.
- Set a time budget. 90 minutes of focused review beats five hours of open-ended watching.
The difference between heatmaps and session replay (and when to use each)
Heatmaps aggregate. A click heatmap shows you where hundreds or thousands of users clicked across a page — it is statistical and fast to read. Session replay is individual — you watch one user’s full journey through your site. Both are qualitative tools. Neither tells you whether a change will improve performance; that requires testing.
When heatmaps are most useful
Use click maps to identify elements users click that are not links (a signal of confusion or unmet expectation), elements users ignore that you expected them to engage with, and how click distribution changes between device types.
Use scroll maps to understand where users stop reading on long-form pages, whether your primary CTA sits above or below the average scroll depth, and whether your content hierarchy matches how far users actually get.
When session replay is most useful
Session replay earns its value in diagnosing specific friction — a form field that causes people to pause, a mobile layout that breaks navigation, a multi-step checkout where users stall on step three. Watch recordings filtered to users who almost converted: they added to cart but did not buy, or they started a form but did not submit.
“The recordings that teach you the most are the ones where users clearly wanted to complete the action but could not — or gave up in frustration. Filter for rage clicks on your highest-intent pages first.”
- Heatmaps: pattern-level; best for navigation, content hierarchy, CTA placement
- Session replay: behaviour-level; best for friction diagnosis, form UX, mobile issues
- Use them in sequence: heatmap surfaces the anomaly, session replay explains it
The six patterns worth hunting for in heatmaps
Most CRO practitioners have a shortlist of heatmap patterns that reliably point to conversion problems. Here are the six we look for in every review.
1. Click confusion on non-interactive elements
When users click text, images, or UI elements that are not links, they expect them to be interactive. A product image they expect to expand. A headline they expect to navigate. A bullet point they expect to reveal more. Every significant click on a non-interactive element is a broken expectation — and broken expectations erode trust and slow conversion.
2. CTA blindness
Your primary call-to-action receives fewer clicks than secondary or tertiary elements on the same page. This usually means the CTA does not visually register as the primary action — it blends into the background, it uses language that does not match user intent, or it is positioned below where most users stop scrolling.
3. Scroll cliff
A sharp drop in scroll depth — where a significant percentage of users stop — at a point above your key content or CTA. This is one of the most actionable signals in a scroll map. If 70% of users never see your pricing section, your pricing section cannot convert them.
4. F-pattern or Z-pattern deviation
On text-heavy pages, users typically scan in F-shaped or Z-shaped patterns. When your click distribution deviates significantly from these patterns, it suggests your page structure is working against natural reading behaviour. Attention analysis tools (available in Hotjar and Clarity) can surface this automatically.
5. Mobile vs desktop divergence
A CTA that performs well on desktop but receives almost no clicks on mobile usually means the button is below the fold on smaller screens, or that the touch target is too small. Comparing device-segmented heatmaps side-by-side is one of the fastest ways to find mobile conversion leaks.
6. Navigation abandonment
Users who arrive on a landing page and immediately click away via the main navigation are telling you the page did not match what they expected to find. High navigation click rates on targeted landing pages usually signal a traffic-quality problem or a headline that overpromises.
- Click confusion on non-interactive elements
- CTA blindness — primary action getting fewer clicks than secondary elements
- Scroll cliff above key content or conversion elements
- Mobile/desktop divergence in CTA visibility
- Navigation abandonment from landing pages
- F/Z-pattern deviation on content-heavy pages
What to look for in session recordings
Raw session recordings are overwhelming unless you filter deliberately. Here is the filtering stack we use before pressing play.
Filter 1: rage clicks
Rage clicks — repeated rapid clicks on the same element — are the clearest signal of frustration in your data. They indicate a user expected an element to respond and it did not. Filter for recordings containing rage clicks on your primary conversion pages and watch every one of them. It is a small dataset and the signal-to-noise ratio is high.
Filter 2: high-intent abandonment
Users who reached a key conversion step (cart page, checkout step one, contact form, pricing page) but did not complete. These are your most valuable recordings. The user had enough intent to get to the door — watch what made them leave.
Filter 3: long session duration with no conversion
Users who spent significantly above average time on site but did not convert. This pattern often reveals confusing navigation, unclear value propositions, or price/feature comparison paralysis. It is a different problem from rage-click frustration, but equally worth diagnosing.
Filter 4: mobile, specific entry page
Segment by device type and entry page. A landing page that converts well on desktop but poorly on mobile deserves its own recording session — the friction will usually be visible within 10–15 recordings. Our conversion rate optimisation process treats mobile as a separate audit track for exactly this reason.
- Rage clicks on conversion pages
- High-intent abandonment (reached checkout, form, pricing)
- Long sessions without conversion
- Mobile users on specific high-traffic entry pages
- Users who visited 4+ pages without converting
Turning observations into testable hypotheses
An observation is not a hypothesis. “Users seem confused on the pricing page” is an observation. “Changing the pricing table layout from three side-by-side columns to a vertical comparison will increase plan selection clicks by 15%+ because mobile users are currently scrolling past the tier differences” is a hypothesis.
A good hypothesis has three parts: the change, the expected outcome, and the reason. The reason is the part most teams skip — but it is the part that makes the hypothesis falsifiable and that teaches you something even if the test loses.
For each pattern you identify in heatmaps or session replay, write the hypothesis in full before moving to test design. This forces you to articulate the mechanism, not just the fix. If you cannot explain why the change should work, the change probably is not grounded in enough evidence.
We typically aim for three to five hypotheses per review session. More than that and you are either working on a very broken page or you are not being selective enough about what constitutes a real signal versus noise. For more detail on how we structure the full testing cycle, our guide to A/B testing for small sites covers prioritisation and statistical significance.
- Change: what specifically will be different
- Expected outcome: what metric you expect to move, by roughly how much
- Reason: the mechanism — why this change should produce that outcome
- Write all three before you design the test variant
Tools comparison: Hotjar vs Microsoft Clarity vs FullStory
The three tools we encounter most in client work are Hotjar, Microsoft Clarity, and FullStory. Here is an honest comparison for teams choosing between them.
Microsoft Clarity is free, which makes it a reasonable starting point for sites under 50,000 monthly sessions. The heatmaps and session replay are solid. The main limitation is data retention (30 days) and the lack of funnel analysis. For basic pattern identification it does the job. We have used it on projects where budget constraints ruled out paid tooling and gotten useful data.
Hotjar is the most widely used paid option. The funnel and form analysis features add meaningful context that Clarity lacks — you can see exactly which form field causes abandonment, not just that abandonment happened. The pricing has increased significantly over the last two years; at current rates, meaningful session volume (10,000+ recordings per month) costs €99–200/month depending on tier.
FullStory is enterprise-grade — better event autocapture, stronger data export, and deeper integration with analytics stacks. It is the right tool when you are running a high-volume e-commerce operation or need to connect session data to revenue attribution. For most SMB clients, it is overkill.
- Microsoft Clarity: free, 30-day retention, good for baseline analysis
- Hotjar: best form analytics, most intuitive for CRO teams, €99–200/month for meaningful volume
- FullStory: enterprise, best for high-volume e-commerce with analytics stack integration
The 90-minute session replay review process
Here is the exact process we run when starting a heatmap and session replay review for a new client engagement.
Minutes 0–15: Define the question. Pick the one page or funnel step with the biggest revenue impact and the most obvious gap between traffic and conversion. Write the question you are trying to answer before opening any tool.
Minutes 15–35: Heatmap review. Pull click maps and scroll maps for that page, segmented by device. Document every anomaly against the six patterns listed above. Do not interpret yet — just note what is unexpected.
Minutes 35–75: Session recordings. Apply filters in order: rage clicks first, then high-intent abandonment, then long session/no conversion. Watch 15–20 recordings per filter set, taking timestamped notes. Stop when you are seeing the same pattern repeatedly — that is your signal.
Minutes 75–90: Hypothesis writing. For each pattern you identified, write a full hypothesis. Stack-rank by estimated impact and implementation difficulty. The top two or three go into the test pipeline.
This process works for landing pages, product pages, checkout flows, and lead-gen forms. The question in step one changes; the rest of the process is consistent. If you are looking at a more complete optimisation stack, our CRO service page covers how we layer qualitative and quantitative analysis. You can also get in touch if you would like us to run a review on a specific page.
- 0–15 min: define the question and page
- 15–35 min: heatmap review against the six patterns
- 35–75 min: filtered session recordings, timestamped notes
- 75–90 min: hypothesis writing and stack-ranking
The tools are widely available and mostly affordable. The gap between teams that get value from session replay analysis and those that do not is almost never about the tool — it is about having a defined process for turning observations into hypotheses, and hypotheses into tests. Start with one question, one page, and one filter. The pattern becomes repeatable faster than you would expect.