Loading tutorials…
Loading tutorials…
Session recordings are powerful — until you have 50,000 of them and no time to watch any. The difference between useful and useless is filter discipline. Here is the system that surfaces the 10-20 sessions per week that actually inform decisions.
Who this is forOperators who installed Clarity, opened Recordings, scrolled forever, and closed the tab. If you have hundreds of recordings and no idea where to start, this tutorial gives you the filter and routine that surfaces the sessions worth watching.
What you'll need
Step 1
Settings → Recordings → Enabled. Verify masking matches your privacy policy.
Settings → Recordings. Confirm "Enable session recordings" is ON (default).
Verify your sampling rate. Default is 'All sessions' — for high-traffic sites (>100K sessions/month), drop this to 50% to keep recording counts manageable. You will still get statistically valid samples.
Re-check your masking config from the account setup tutorial. Recordings replay with the masking applied — if you set Strict at install, recordings show grey boxes where input fields were.
If you need to UN-mask specific elements for analysis (e.g., search query inputs, product titles), add their CSS selectors to 'Unmask elements' in Settings → Privacy. This applies forward — past recordings stay masked.
Click Save.
Step 2
Settings → Smart Events. Define click, page visit, and form submit events for your conversion path.
Settings → Smart Events. Smart Events are Clarity's answer to GA4 events — custom triggers you can filter recordings against.
Create one Smart Event per conversion: "Purchase completed" (page visit on /order-confirmation), "Signup completed" (page visit on /welcome), "Lead form submit" (click on submit button OR page visit on /thanks).
Use CSS selectors for click events when possible — they survive copy changes. Use URL patterns for page visits.
For each Smart Event, set the "Type" — Conversion, Funnel step, or Custom. Mark anything that represents real business value as Conversion.
Save. Smart Events apply forward — past recordings will not be retroactively tagged, but new ones will.
Wait 24-48 hours, then verify Smart Events are firing: Recordings → filter by "Smart Event: [your event]" → confirm sessions show up.
Step 3
Recordings → Filters → save preset. Start with: 60s+ duration + 3+ pages + NOT bot.
In the Recordings tab, click "Filters" (top of the page).
Add: Session duration > 60 seconds (filters out bounces and bots).
Add: Page count >= 3 (real engagement, not single-page hits).
Add: Device = Desktop (or Mobile — run separate watchlists per device).
Add: NOT a bot (Clarity auto-detects most bots; this filter is belt-and-suspenders).
Click "Save as preset" → name it "Watchlist — Engaged Desktop." Repeat for Mobile.
This preset becomes your default weekly review queue. Most weeks, this filter surfaces 50-200 sessions — manageable to scan.
Step 4
New filter preset: Rage click = Yes OR Dead click = Yes OR Excessive scrolling = Yes.
Recordings → Filters → New preset.
Add: Rage click = Yes (user clicked rapidly in frustration).
OR: Dead click = Yes (user clicked something that did not respond).
OR: Excessive scrolling = Yes (user scrolled up-down repeatedly looking for something).
OR: Quick back = Yes (user immediately navigated back).
Save as preset: "Friction — Struggle Sessions."
These are your highest-value recordings to watch. Even 5 sessions from this filter typically reveal a fixable UX problem.
Step 5
Filter preset: Smart Event = [your conversion event]. Compare to "Friction" filter to find lift opportunities.
New filter preset.
Add: Smart Event = "Purchase completed" (or your primary conversion).
Add: Device = Desktop (or split).
Save as preset: "Converters — Purchase Desktop."
The value here is comparing converter sessions to friction sessions. What did converters do that strugglers did not?
Specific tactic: pick a struggling page (high rage clicks) and watch 5 converter sessions + 5 friction sessions on that page. The behavioral difference is your test hypothesis.
Step 6
15 min every Monday: 3 sessions from Friction filter, 3 from Watchlist, 1 from Converters. Notes go in a shared doc.
Block 15 minutes on your calendar every Monday morning (or whatever day starts your work week).
Step 1 (5 min): Open Recordings → Friction filter → sort by most recent. Watch 3 sessions at 4x speed. Note any new patterns.
Step 2 (5 min): Open Watchlist filter → sort randomly (Clarity has a "Random" sort option). Watch 3 sessions at 4x speed.
Step 3 (3 min): Open Converters filter → watch 1 session at 4x speed to remind yourself what success looks like.
Step 4 (2 min): Log any new patterns in a shared "Clarity findings" doc with the date. Patterns that recur 3+ weeks become CRO test candidates.
This routine produces 5-10 actionable hypotheses per quarter. Skipping it produces zero.
Step 7
Use Clarity's built-in share link (top right of a recording). Default is "share read-only with link" — review before sending.
In any recording, click the "Share" icon (top right).
Choose "Copy link" — generates a link viewable by anyone with Clarity access to that Project (or anyone with the link if "Public link" is enabled, which it usually should not be).
For PII-sensitive recordings: verify masking is applied in the playback before sharing. If you UN-masked something for your own analysis but want to share more broadly, re-apply mask temporarily.
For external sharing (designers, contractors): add them as Viewers in Settings → Team. Avoid Public links — they cannot be revoked once shared.
For documentation/training: download the recording as MP4 (Share → Download). MP4 captures the visual replay but not the interactive timeline.
Common mistakes
Watching recordings without any filter
What goes wrong: You scroll through endless 5-second bot sessions and 12-second bounces, give up after 20 minutes, and conclude 'recordings are useless.' Lost: the ability to make data-driven CRO decisions for the next quarter. Opportunity cost: $5K-15K in untested CRO hypotheses.
How to avoid: Build the three filter presets (Watchlist, Friction, Converters) on day one. Never open Recordings without applying one.
Configuring Smart Events with brittle CSS selectors
What goes wrong: Dev team refactors a component → your Smart Event selector breaks → you stop seeing conversions in the filter → you waste 2-3 weeks before noticing → CRO testing roadmap is delayed a month.
How to avoid: Use `data-clarity` attributes on every element you track. Ask your dev team to treat these as a contract — do not rename without coordination.
Watching at 1x speed
What goes wrong: A 5-minute recording at 1x takes 5 minutes. Watching 10 recordings = 50 minutes. Nobody has 50 minutes/week sustained — so you stop watching entirely.
How to avoid: Always watch at 4x speed. Pause when you see something interesting. Jump to specific events using the timeline markers. A 5-min recording takes 75 seconds at 4x.
Recording without any data retention plan
What goes wrong: Default 30-day retention means your seasonal recordings (Black Friday, end-of-quarter pushes) are deleted before you can review them in context. The most valuable recordings are gone before you analyze them.
How to avoid: Push retention to 90 days minimum (free tier max is 13 months — use it). For critical campaigns, download key recordings as MP4 for permanent archive.
Sharing public recording links externally
What goes wrong: Public links cannot be revoked. A contractor leaves with the link → your competitor sees user behavior on your site → you have no recourse. Plus potential GDPR exposure if any PII was visible.
How to avoid: Add external collaborators as Viewers in Settings → Team. Use named access, not public links. If a Public link is absolutely necessary, rotate the recording or delete it after the external review.
Recap
Done — what's next
How to set up Clarity heatmaps that actually inform decisions
Read the next tutorial
Hand it off
Recordings are useful only if someone actually watches them every week. Most founders intend to and never do — which is fine, but it means the data is producing $0 of value. A CRO specialist who watches weekly and surfaces 3-5 fixable patterns per month typically pays for themselves in one A/B win. From $14-16/hr on EverestX.
Get a CRO specialist watching
10-15 is the sweet spot. Below 5: you are not seeing enough variety. Above 20: diminishing returns — patterns repeat. Pick the 15 most relevant via filters and watch each at 4x speed.
Yes — click Share → Download to export as MP4. Note that MP4 download captures visual only, not the interactive timeline (no jump-to-event, no scroll heatmap overlay). For deep analysis, keep recordings in Clarity.
Common causes: (1) bot traffic that hits a page and leaves before JS fully loads, (2) user clicked an outbound link immediately, (3) page errored and loaded a redirect, (4) ad-blocker terminated Clarity mid-recording. Use the 60s+ duration filter to exclude these.
Yes — Clarity records mobile sessions including touch events, scroll patterns, and orientation changes. Replays correctly render the mobile viewport size. Tap targets and rage taps are tracked the same as desktop clicks.
Default 30 days, configurable up to 13 months in Settings → Data → Retention. Push to 90 days minimum. Recordings beyond your retention window are permanently deleted — no recovery.
Microsoft Clarity
Clarity generates heatmaps automatically — but the default view aggregates desktop + mobile + tablet across every URL variant. That is unusable. Here is the configuration that produces heatmaps you can actually act on.
Microsoft Clarity
Rage clicks and dead clicks are the most actionable signals Clarity gives you — every one represents a user actively frustrated with your site. This walkthrough shows the diagnostic sequence to fix them.
Microsoft Clarity
Out-of-the-box Clarity gives you "all sessions, all devices, all sources." That is useless. Segmentation is where Clarity goes from "looks cool" to "actually changes decisions." This is the segment library specialists build.
Microsoft Clarity
You installed Clarity. You watched some recordings. You ran an A/B test or two. So why is conversion rate still flat? This is the honest framework: when DIY CRO hits its ceiling and a specialist starts paying for themselves.