Core Web Vitals Audit

From a slow page to a proven fix.

Treo CWV Audit loads your page, interacts with it like a visitor, finds the causes, tests each fix on the live page, and hands the proven ones to your coding agent.

Private beta for Treo account owners50 audits a month

A real auditwww.daals.co.uk/products/maru-round-4-6-seater-extending-oak-pedestal-dining-table-silver-oakdesktop

Largest Contentful Paint LCP

2.35sGood→1.72s

TTFB0.14 sLoad delay0.31 sLoad duration1.15 sRender delay0.75 sLCP1.72 sDefer the Convert A/B testing scripts−0.67 s00.511.52 s

2 fixes tried1 proven11 loadsOpen this audit

01

It loads your page as a visitor would

The audit loads your page on a phone and on a desktop and shows one load you can read like DevTools: LCP and its four parts, the shifts, the frames. A URL is enough: no script, no setup, no access to your code.

In this audit

Largest Contentful Paint LCP

2.35sGood

on a desktop

The Overview screen: LCP 2.35 s with its four parts on one timeline, and the proven fix drawn as an arrow
Overviewwww.daals.co.uk, a product page, on a desktopaudited 19 Nov 2026
02

It interacts with the page like a visitor

A page is more than its load. The audit writes the scenarios a visitor would take on this page and walks them step by step: it types, it presses, it waits for the next paint. Every press is measured, so INP comes with the press that set it and the script behind it. For INP the audit finds the cause today; testing its fixes, as it does for LCP and CLS, comes next.

In this audit

Interaction to Next Paint INP

600msPoor

on a desktop, set by the press “Add to Basket”

The scenarios of the Overview screen: three scenarios, INP 600 ms set by the press Add to Basket, with the page before and after the press
Overview, the scenarioswww.daals.co.uk, a product page, on a desktopaudited 19 Nov 2026
03

It finds what holds the page back

A finding names its place on the waterfall, what it costs and the experiment that tested it. Click one, and its requests, tasks and shifts light up on the chart. The one open here is “17 requests block the render”; the fix that defers the Convert A/B testing scripts is proven.

In this audit

Findings

10

3 of them high on this load

The Waterfall screen with a finding open: 17 requests block the render, and the fix that defers the Convert A/B testing scripts, proven
Waterfall with a finding openwww.daals.co.uk, a product page, on a desktopaudited 19 Nov 2026
04

It tries each fix on the live page

A fix is applied in the audit’s browser as the page loads, and the page is loaded again and again, with the fix and without it. One fix made this page faster and is marked proven. One made it slower. Each has its number. So the sprint goes to the one that works, and none of it to the one that does not.

In this audit

Fixes tried

2

1 proven11 loads

The Experiments screen: two fixes as rows, one proven and one slower, each with its change in LCP
Experimentswww.daals.co.uk, a product page, on a desktopaudited 19 Nov 2026

Your page

Does a page of yours fail LCP, CLS or INP?

Name it. In the beta, Treo CWV Audit measures it, finds the cause and tests each fix on it.

Request access

Private beta for Treo account owners50 audits a month

Or see it on a real page first: live examples

05

It shows why the fix worked

Open any fix and see its load beside the untouched one, on one clock. The frames show the page painting sooner. The LCP parts show where the time went: here the render delay drops from 0.75 s to 0.21 s. The requests show what the fix changed.

In this audit

LCP with the fix

1.72s

2.35 s without it

The Compare screen: the load with the fix beside the untouched load, frame by frame
Comparewww.daals.co.uk, a product page, on a desktopaudited 19 Nov 2026
06

It hands the proven fix to your coding agent

A proven fix comes with the exact change and a prompt. Copy it into your coding agent, review what it writes, and ship it. Treo proves and shows; you decide. An API and a command line tool come next, so your agent can start an audit and read the proven fixes itself. Next

In this audit

Defer the Convert A/B testing scripts

−0.67s

Proven

A proven fix’s panel: Copy the prompt, its loads, the page with and without it, what it changes
A proven fix’s panelwww.daals.co.uk, a product page, on a desktopaudited 19 Nov 2026
07

Then you run your own experiments

The audit runs by itself first. After that it works for you: test your own fix, add your own scenario, and get the same proof, a number beside the untouched page. Any expert can tune it to their site.

In this audit

Your own tests and scenarios

In development

The Experiments screen: the fix the audit did not try, and why
Experiments, the fix not triedwww.daals.co.uk, a product page, on a desktopaudited 19 Nov 2026

What it does for each metric

Largest Contentful Paint LCP

  • Measured on one real load, split into its four parts
  • What holds it back, found on the waterfall
  • Each fix tested on your live page, with its number
  • A prompt with every proven fix

Cumulative Layout Shift CLS

  • Every shift, with its moment and what moved
  • During the load and during the scenarios
  • Each fix tested on your live page, with its number
  • A prompt with every proven fix

Interaction to Next Paint INP

  • Scenarios the audit writes and walks, press by press
  • The slow press: input delay, processing, presentation
  • A waterfall of the press, with the script behind it
  • Fixes for INP, tested as LCP’s and CLS’s areNext

A fix is proven in the lab, on your live page, in the audit’s own browser: nothing on your site changes, and your real users confirm the gain after you ship. Pages behind a login are not audited.

Questions about Treo CWV Audit

01What is a Core Web Vitals audit?
A study of one page’s LCP, CLS and INP that explains why they are slow. Treo CWV Audit also tests each fix on the live page and shows which ones work.
02How is it different from Lighthouse or PageSpeed Insights?
They list opportunities and estimate what each would save. Treo CWV Audit applies each fix to your page and measures it beside the untouched page.
03How do I fix a slow LCP?
Start from the cause, not from a checklist. The audit splits LCP into its four parts, finds what holds the largest one, and tests the fixes for it.
04Can it fix INP?
It finds the slow press and the script behind it, in scenarios it walks like a visitor. Tests of fixes for INP come next.
05Does the audit change my site?
No. A fix is applied in the audit’s own browser as the page loads. Your site and your code stay as they are until you ship the change.
06Which pages can it audit?
Any public page, by its URL, with nothing to install. Pages behind a login are not audited.
07Can my coding agent use it?
Yes. Every proven fix comes with a prompt your agent can apply. An API and a command line tool, treo-cli, are next.

Request access

One audit loads your page dozens of times in real Chrome, on a phone and on a desktop, to measure each fix. That takes real machines. So the beta is private only for Treo account owners.

  • Pricefree while the beta runs
  • Audits50 a month for an account
  • The reportfull functionality
Do you have a Treo account?

After the beta the audit will be part of a paid plan. Your answer tells us where it belongs.

We read every request and answer by email