Real user monitoring for websites

Real user monitoring that shows what got slower today, and why.

Epsilon Delta Gemini is a real user monitoring (RUM) tool for websites. It tracks Core Web Vitals, server and API time, JS errors, rage clicks and bots from real visits. AI-curated insights explain what changed.

Start free trial Take the dashboard tour
paid page inspections
9 Billion
growth in annual traffic
18%

Pulse

Your morning stand-up, already written.

Today vs yesterday for every product and page: what improved, what degraded, and the top offenders by metric.

Improved

  • Total Blocking TimeHome · 17% of traffic ▼ 120 ms (24%)
  • Largest Contentful Paint/product/:id · 44% of traffic ▼ 0.3 s (11%)
  • Rage clicks per 1k views/payments · 2% of traffic ▼ 1.1 (12%)

Degraded

  • Gemini Score/search · 9% of traffic ▼ 7 pts (15%)
  • Cumulative Layout Shift/basket · 1% of traffic ▲ 0.02 (6%)
  • Total Blocking Time/payments · 2% of traffic ▲ 410 ms (17%)
Open Pulse in the tour

Screen Painting

See the load the way your users saw it.

  1. 0.0 s
  2. 0.6 s
  3. 1.1 s
  4. 1.7 s · LCP
  5. 2.1 s

Frame-by-frame paint progress from real visits, so you can compare what users saw on slow and fast loads.

Open in the tour

Interaction

Name the script behind a slow click.

Input delay
22 ms
Processing
84 ms
Presentation
62 ms
  • tag-manager.js 4,800 poor
  • inline script 3,600 poor
  • button.add 610 ms
Open in the tour

Rage clicks

See where users lose patience.

Heatmap, top rage elements, repeat rage and INP while raging, split by page, platform and browser.

Open in the tour

JS errors

Fix the errors with real reach first.

  • TypeError: cannot read properties of null4.2% reach
  • Unhandled rejection: failed to fetch1.9% reach
  • ReferenceError: x is not defined0.7% reach

Grouped by message, with class, trend, first seen and a bot filter.

Open in the tour

SEO perf

Core Web Vitals, page by page.

PageTrafficLCPINPCLS
/product/:id54%3.4 s210 ms0.14
Home20%3.1 s230 ms0.42
/checkout2%1.8 s880 ms0.27
Open in the tour

Acquisition · New

Know which traffic source lands slowest.

ChannelShareLCP p75Pass CWVBounce
Paid search33%2.08 s76%81%
Direct29%6.32 s21%58%
Organic search17%2.44 s55%63%
Email & messaging1%3.18 s61%76%
Why LCP changed2.97 s → 2.86 s
  • Traffic mix −71 ms
  • Within channels −37 ms

Insight · Two thirds of the change came from who arrived, not from the site.

Core Web Vitals, bounce and score for first visits, by channel, campaign, landing page and ad click ID. Every change is split into traffic mix and the channels themselves, so a campaign start never looks like a site regression.

Open in the tour

Start with your question.

Screen Painting

Sample data

What did users see while loading?

How the page painted for real users, shown as speed perception.

FCP
1.1 s
LCP
2.8 s
Speed Index
2.1 s
Slowest to paint
  • /search 3.8 s
  • /product/:id 3.0 s
  • Home 1.9 s

Insight · Compare what users saw on slow and fast loads.

See it in the tour

Interaction

Sample data

Which click feels slow?

INP split into input delay, processing and presentation, with culprit scripts and target elements.

INP p75
250 ms
Processing
84 ms
Interactions
310K
Top culprit scripts
  • tag-manager.js 4,800
  • inline script 3,600
  • framework.js 2,500

Insight · Compare your field INP against the CrUX zone.

See it in the tour

Rage clicks

Sample data

Where are users frustrated?

A heatmap and named elements show what users keep tapping.

Rage rate
2.2 / 1k
Sessions
2.4K
Repeat rage
24%
Top rage elements
  • Image reel 320
  • input.search 210
  • button.add 160

Insight · One page causes 29% of all rage clicks.

See it in the tour

JS errors

Sample data

Which errors matter most?

Errors grouped by message, with class, reach, trend and first seen.

Error rate
240 / 1k
Affected views
11%
New
36
Top problems
  • TypeError: null properties 47K
  • Failed to fetch 28K
  • x is not defined 9K

Insight · Bot traffic can be excluded in one click.

See it in the tour

Server, Ajax and Resources

Sample data

Is it the backend or the network?

Connection phases, endpoints and assets, with cumulative load showing what costs most overall.

Response
0.58 s
API p90
1,040 ms
DNS
0.04 s
Where cumulative load concentrates
  • Images 90%
  • Scripts 7%
  • Other 3%

Insight · Images carry 90% of cumulative load: the clearest target.

See it in the tour

SEO perf

Sample data

Will this hurt my SEO?

Field Core Web Vitals at the 75th percentile, for your highest-traffic pages.

LCP
2.9 s
INP
260 ms
CLS
0.12
Share of good experiences
  • FCP 72%
  • INP 65%
  • LCP 60%
  • CLS 51%

Insight · Flagged pages combine high traffic with poor vitals.

See it in the tour

Acquisition

Sample data

Which traffic source lands slowest?

First visits by channel, campaign and landing page, with Core Web Vitals and bounce for each, and how much of a change came from traffic mix.

Landings
413K
Paid share
38.6%
LCP p75
2.86 s
LCP p75 by channel
  • Direct 6.32 s
  • Email & messaging 3.18 s
  • Organic search 2.44 s
  • Paid search 2.08 s

Insight · Two thirds of the LCP change came from the traffic mix, not the site.

See it in the tour

Insight details

Sample data

Why are vitals in this state?

Page weight, blocking resources, LCP element type, entry points, referrers and browsers.

LCP
1.7 s
CLS
0.07
TBT
520 ms
Page weight before LCP
  • Before FCP 190 KB
  • Above fold 74 KB
  • Before LCP 160 KB

Insight · Images drive LCP: optimise the hero image.

See it in the tour

Bot details

Sample data

How much traffic is not human?

Suspect bots, referrals, user agents, viewports and top IP ranges.

Suspect bots
810
No referral
82%
Top UA
62%
Top IP ranges
  • 203.0.113.* 4.8%
  • 198.51.100.* 1.6%
  • 192.0.2.* 1.4%

Insight · Keep crawlers out of your real-user numbers.

See it in the tour

Every Gemini screen, one link each

Each screen of the product tour has its own page. Open the one that matches your question.

What are Core Web Vitals?

Core Web Vitals are Google’s three field metrics for page experience: LCP for loading, INP for responsiveness and CLS for visual stability. A page passes when 75% of real visits meet the good threshold for all three.

LCP

Largest Contentful Paint

When the largest image or text block in view finishes rendering.

Good ≤ 2.5 sPoor > 4 s

INP

Interaction to Next Paint

How long the page takes to respond visibly to a click, tap or key press.

Good ≤ 200 msPoor > 500 ms

CLS

Cumulative Layout Shift

How much visible content moves unexpectedly while the page is in use.

Good ≤ 0.1Poor > 0.25

TTFB

Time to First Byte

Time until the first byte of the page arrives. Not a Core Web Vital, but it delays everything after it.

Good ≤ 0.8 sPoor > 1.8 s

Why Gemini over Lighthouse and PageSpeed

Capability Gemini Lighthouse PageSpeed Insights
Data source Real users, all browsers Single lab run Lab + Chrome-only CrUX
Freshness Selectable windows, down to 24 hrs On demand CrUX 28-day window
Slow-click culprit scripts Yes, in the field No No
Rage clicks and JS errors Yes No No
Day-on-day change report Pulse No No
Real-time alerts Yes No No

Based on public documentation for Lighthouse and PageSpeed Insights, November 2026.

Questions about Gemini RUM

What is real user monitoring (RUM)?

RUM measures performance from the browsers of real visitors as they use your site. Lab tools such as Lighthouse run one simulated test instead.

How is Gemini different from Lighthouse or PageSpeed Insights?

Lighthouse is a lab test and CrUX is Chrome-only over 28 days. Gemini reports real users across browsers, with windows as short as 24 hours.

Does Gemini work with single-page apps?

Yes. Speed details separates full page loads from soft route changes.

Will the Gemini script slow my site down?

[Script size and loading strategy.]

Is user data private?

[Data handling, PII policy, hosting region and compliance details.]

Last updated 6 Nov 2026 · Reviewed by [Name], [Role]

See your own Pulse tomorrow morning.

Start free trial Contact for a free demo