Digital Quality

Slow is a churn reason too.

Nobody writes a review about latency. They just stop opening the app. Core Web Vitals at the 75th percentile on web, cold start and screen render on mobile — measured from real sessions rather than from a test rig.

app.cxorbi.com/performance
30 Days ⌄
Web RUMSyntheticApp

Web Performance

Core Web Vitals (RUM) at the 75th percentile, by device and page group.

LCP p75good ≤ 2.5s
Mobile2.84s

p95 4.12s · n=1284

Desktop1.92s

p95 2.61s · n=2140

INP p75good ≤ 200ms
Mobile212ms

p95 402ms · n=1180

Desktop148ms

p95 240ms · n=2012

CLS p75good ≤ 0.1
Mobile0.083

p95 0.184 · n=1284

Desktop0.041

p95 0.092 · n=2140

Worst page groups

Page groupSamplesLCP p75INP p75CLS p75
/checkout4,8213.41s184ms0.112
/account/kyc3,1094.62s318ms0.058
/portfolio2,8842.18s142ms0.031
/research2,4102.94s196ms0.074

Right-click a page group to open it in Journeys or Heatmaps.

p75, not the average

An average hides the bad days. The 75th percentile is where a real user's frustrating session actually lives, and it is the bar the web vitals thresholds are set against.

Mobile counts too

Cold start and screen render times per release, so a startup regression is caught by the data rather than by a reviewer.

Attached to consequences

A slow screen is only worth arguing about if it costs something. Every measurement resolves back to the sessions it happened in.

What Performance Monitoring does

Web vitals from real users

Measured in the field on actual sessions, which is the only version of the number that predicts anything.

  • Core Web Vitals reported at p75
  • Broken down by page or mapped page group
  • Segmented by device class, so a desktop average stops hiding mobile

Mobile app performance

The measurements a mobile team is actually judged on.

  • Cold start timing per release
  • Screen render times across the app
  • Comparison release over release, so a regression is attributable

From a number to a decision

The point of a percentile is finding the people behind it.

  • Move from a slow screen to the sessions that abandoned on it
  • Check whether the slow page is also the page losing the funnel step
  • Watch a replay of a session that waited

Better with the rest of it

Every module reads the same identity, events and sessions, so an investigation that starts here does not have to be rebuilt to continue somewhere else.

Frequently asked questions

Is this synthetic monitoring or real user monitoring?

Real user monitoring. Every figure comes from an actual session on an actual device. There is no synthetic test runner and no uptime probe — this measures what your users experienced, not what a lab machine did.

Why the 75th percentile?

It is the percentile the Core Web Vitals thresholds are defined against, and it represents a bad-but-not-rare experience. An average is dominated by fast sessions and will tell you everything is fine while a quarter of your users struggle.

Does it cover mobile apps?

Yes — cold start and screen render times on iOS and Android, reported per release so a slowdown can be traced to the build that introduced it.

Ready to start?

Bring the whole user story into one workflow

Free plan availableNo credit card requiredWeb, iOS & Android