Skip to main content

Try Base free for 14 days, no card or M-Pesa needed See pricing

Guide

How to prioritise Core Web Vitals fixes by page template

Last updated
Page tools
Agent setup

Compare Core Web Vitals for listing, category and editorial templates, pick representative URLs, find shared causes and decide which to fix first.

PageSpeed Insights says your site fails Core Web Vitals on mobile. Your developer asks which pages to fix first, and the honest answer is “it depends on the page”.

Most sites are built from a few templates: listings, category or search pages, and articles. Fix a template once and every page using it changes. So compare templates, not single pages.

The short version. Pick one representative URL per template, read its field data by device, and note whether the number is really for that URL or for the whole site. Look for a cause shared across templates, then fix first where the business impact is biggest. Confirm with field data over the next 28 days.

Field data decides, lab data explains

Field dataLab data
Where fromReal Chrome users, through the Chrome UX Report (CrUX).A simulated load on one device and one network (Lighthouse).
PeriodThe previous 28 days, rolling (Chrome).The moment you ran it.
Measures INPYes.No. Total Blocking Time is a rough proxy, not a substitute (web.dev).
Use it toDecide what fails for visitors and confirm fixes.Find why it fails, and test a change before release.

PageSpeed Insights shows both, and Google notes the two may differ because lab data uses a single device and fixed network (Google).

The thresholds

Each metric is judged at the 75th percentile of page loads, separately for mobile and desktop. A page passes only when all three are good (web.dev).

MetricWhat it measuresGoodPoor
LCPWhen the main content appears≤ 2.5 s> 4 s
INPHow fast the page responds to taps and clicks≤ 200 ms> 500 ms
CLSHow much the layout jumps≤ 0.1> 0.25

Between good and poor is “needs improvement” (web.dev).

Step 1 · Choose one URL per template

  • List your templates: for a property site, listing pages, category or search pages, editorial guides and the homepage.
  • For each, pick an indexable URL with real traffic. CrUX only reports pages that are publicly discoverable and have enough visitors (Chrome).
  • Pick a typical page, not an outlier: a listing with the usual number of photos, not the one with sixty.
  • Write down the URL so every later check uses the same one.

What one URL can’t tell you

LimitWhat to do
One page may not represent the template.Spot-check one or two more URLs of the same template in PageSpeed Insights.
The URL may have too little traffic for its own data. PageSpeed Insights then falls back to the whole origin.Note the scope. An origin number is not evidence about that template.
CrUX strips query strings, so filtered search URLs are counted as their base page.Choose the base search URL, and test filters in the lab.
Search Console groups pages with similar experience, judged on the worst metric (Google).Use its groups to check whether your template list matches Google’s.

Example: a property listings site

Illustrative example, not real data. An invented property site, Nyumba Hub (nyumbahub.example.co.ke). Field data on phones, 19 October – 15 November 2026.

TemplateDeviceScopeField-data periodRepresentative URLLCP · INP · CLSBusiness importanceNext diagnostic step
Whole sitePhoneOrigin28 days to 15 Novnyumbahub.example.co.ke2.9 s · 290 ms · 0.06Context onlyNone: use it as the baseline.
ListingPhoneURL28 days to 15 Nov/listing/2-bed-apartment-kilimani-48213/3.6 s · 180 ms · 0.04High: 80% of enquiries start hereLighthouse: when is the first gallery photo found and how heavy is it?
Category / searchPhoneURL28 days to 15 Nov/rent/nairobi/apartments/2.4 s · 380 ms · 0.05High: most Google clicks land hereRecord tapping filters in Chrome DevTools; Lighthouse can’t measure INP.
Editorial guidePhoneOrigin (no URL data)28 days to 15 Nov/guides/renting-in-nairobi/whole-site figuresMedium: early-stage readersLab check only; no field evidence for this template.
ListingDesktopURL28 days to 15 Novsame URL2.1 s · 120 ms · 0.03Lower: most visits are on phonesNone now.

LCP by template and device

  • Good
  • Needs improvement
  • Whole site, for context
  • No URL data: whole-site figures
  • Good: 2.5 s or less
  • Poor: over 4 s
Whole sitePhone · origin2.9 s
ListingPhone · URL3.6 s
Category / searchPhone · URL2.4 s
Editorial guidePhone · origin (no URL data)2.9 s
ListingDesktop · URL2.1 s

75th percentile of real Chrome visits (CrUX field data), 28 days to 15 November 2026. The editorial guide has too little traffic for its own data, so it repeats the whole-site figures.
Source: Illustrative example: invented figures, not real data.

Copy linkDownload SVG
Embed

INP by template and device

  • Good
  • Needs improvement
  • Whole site, for context
  • No URL data: whole-site figures
  • Good: 200 ms or less
  • Poor: over 500 ms
Whole sitePhone · origin290 ms
ListingPhone · URL180 ms
Category / searchPhone · URL380 ms
Editorial guidePhone · origin (no URL data)290 ms
ListingDesktop · URL120 ms

75th percentile of real Chrome visits (CrUX field data), 28 days to 15 November 2026. The editorial guide has too little traffic for its own data, so it repeats the whole-site figures.
Source: Illustrative example: invented figures, not real data.

Copy linkDownload SVG
Embed

Shared problem: Lighthouse on both the listing and the search page shows long main-thread tasks from the same map script, loaded before anyone opens the map. It sits on two high-value templates and plausibly hurts both LCP and responsiveness.

Decision: load the map only when it is opened, on every template that uses it. Then fix the listing’s first gallery photo for LCP. Leave the guides: their score is the whole site’s, not theirs.

Step 2 · Decide which template goes first

WeighFix earlier whenFix later when
Business importanceEnquiries, sales or most search clicks pass through it.Few visitors reach it, or they are early researchers.
Distance from “good”It is poor, or the fix is clear and it sits just past the line.It already passes.
Evidence scopeThe data is for that URL.It is origin data standing in for the template.
Shared causeOne fix moves several templates.The cause is unique to a low-value page.
DeviceIt fails on the device most visitors use.It fails only on the minority device.

Read it by template in Ndovu

  1. Open Site Audit → New audit. Turn on Include Lighthouse if you want lab tests too.
  2. In the finished audit, open Performance. Real-user Core Web Vitals (Chrome UX Report) shows the Whole site, the homepage and one row per template, with a Phone / Desktop toggle.
  3. Read the Data level column: URL is that page’s own data; Origin (no URL data) repeats the whole-site figures because the page has too little Chrome traffic.
  4. Below it, Lighthouse lab tests show diagnostics for a sample of templates, if the audit included Lighthouse.

Ndovu picks each template’s page for you: the indexable page with the most Search Console clicks, then inlinks. The section shows the CrUX collection dates, which are independent of when the audit ran. More in Audit settings, views and priorities.

Step 3 · Confirm the fix in field data

  • Check the lab result straight after release: it shows whether the change did what you expected.
  • Field data is a 28-day rolling window, so a fix shows gradually. Compare the same URL and device after a full 28 days from release.
  • Check the scope again. If the URL had its own data before, it should still have it.
  • In Search Console’s Core Web Vitals report, Start Tracking on a fixed issue begins a 28-day monitoring session.

A template sheet you can copy

Template:
Device:
Representative URL:
Scope (URL / origin / no data):
Field-data period:
LCP / INP / CLS (75th percentile):
Business importance (enquiries, clicks):
Shared with other templates?:
Next diagnostic step:
Fix released on:
Re-check after (28 days):
When you need to…Read
Run an audit and verify fixesRun a site audit and verify fixes
Check one URL quicklyCore Web Vitals checker
Rank the rest of the auditPrioritise a technical audit

Sources

Next steps

Was this helpful?