Skip to content
EncryptTechBD
Home Industries Our Work Blog Pricing About Contact Discuss Your Project
বাংলা (BN)
Website & Infrastructure 6 min read

Core Web Vitals diagnosis: a step-by-step guide

A repeatable sequence for finding which template, script or image is actually hurting LCP, INP and CLS, using field data first and lab tools second.

E
EncryptTechBD · 0 comments
Close-up of a silver laptop keyboard
Photo: Math / StockSnap

Most performance conversations start in the wrong place. Someone runs a speed test on the homepage, gets a number, forwards it to a developer, and a week later the number has changed but nothing in the business has. Core Web Vitals diagnosis works better as a repeatable sequence: find the pages that matter, read what real visitors experience, then reproduce and fix one cause at a time.

Google's Core Web Vitals are three measurements: Largest Contentful Paint (LCP) for loading, Interaction to Next Paint (INP) for responsiveness, and Cumulative Layout Shift (CLS) for visual stability. The published target is 2.5 seconds or less for LCP, 200 milliseconds or less for INP, and 0.1 or less for CLS, assessed at the 75th percentile of real page visits. Those numbers tell you when a page passes. They do not tell you why it fails, and that is the work this guide covers.

Prerequisites

  • Access to analytics or a real-user monitoring tool that records page-level performance by device.
  • A Google Search Console property, which groups URLs by similar performance.
  • A staging copy of the site and permission to deploy changes during a quiet window.
  • A list of the ten to twenty URLs that carry revenue or enquiries, not just the homepage.
  • Knowledge of your visitor mix: mostly mid-range Android on mobile data, or desktop on office broadband. The answer changes the diagnosis.

The step-by-step sequence

  1. Group URLs by template. Product page, category page, blog post, checkout and campaign landing page usually behave differently. Performance problems repeat by template because the same scripts and layout load on every page of that type. Output: a sheet listing template, example URL, monthly sessions and owner.
  2. Read field data before lab data. Search Console's Core Web Vitals report and your own analytics show what visitors experienced over the last 28 days. Lab tools simulate one device and one connection; they explain causes but they should not decide priority. Output: each template marked pass, needs improvement or fail, by device.
  3. Choose the worst template, not the worst URL. Fixing one product page helps one page. Fixing the product template helps every product. Output: one template selected with a written reason.
  4. Reproduce the problem in the lab. Run PageSpeed Insights or Lighthouse against a representative URL with mobile emulation, then open the diagnostics rather than reading the headline score. Output: a saved report with the date and tool version.
  5. Diagnose LCP element by element. Identify the element that is the largest contentful paint, usually a hero image, a heading or a CSS background. Then break its timeline into four parts: time to first byte, resource load delay, resource load duration and element render delay. Output: the named LCP element and its dominant delay.
  6. Diagnose INP with real interactions. INP is not a page-load metric. Record a session while filtering a product list, opening a menu or adding to cart. Long tasks, heavy JavaScript on the main thread and large DOM updates are the usual causes. Output: the interaction, the long task and the script responsible.
  7. Diagnose CLS by source. Look for images without dimensions, embeds without reserved space, web fonts that swap with different metrics, cookie banners injected late and content inserted above existing content. Output: each shift source named with an element selector.
  8. Fix one cause per deployment. Compress and correctly size the hero image, preload it, stop delivering it through CSS, defer the blocking script, reserve space for the embed. One change per release means you can attribute the movement. Output: a change log with date, change and expected effect.
  9. Verify in the lab, then wait for the field. Lab verification is immediate; field data moves over a 28-day window. Do not declare success on day one. Output: a before-and-after lab report plus a diary entry for the field check.
  10. Add a guardrail. A performance budget in the build pipeline, or a monthly check of the same templates, stops the next campaign page from undoing the work. Output: the budget values and the person who reviews them.

Evidence checklist

  • Field data export for the last 28 days, per template and device.
  • Lab report naming the LCP element for each template.
  • Interaction recordings or trace files for the slowest INP interaction.
  • CLS source list with element selectors and the change that removed each one.
  • Deployment log matching each change to a date.
  • Post-change field data for the same 28-day window.

Common mistakes

  • Optimising the homepage while the product template carries the traffic.
  • Trusting a single lab score taken on a fast office connection.
  • Installing a plugin that claims to fix Core Web Vitals without knowing which metric it touches.
  • Lazy-loading the LCP image, which delays the very element being measured.
  • Removing a heavy third-party script from the homepage but leaving it on checkout, where it costs money.
  • Reporting a field-data improvement before the 28-day window has rolled over.

Metrics to report

  • Percentage of template impressions in the good band for each metric, by device.
  • LCP p75 in seconds, per template.
  • INP p75 in milliseconds, with the three slowest interactions named.
  • CLS p75 per template.
  • Number of third-party scripts on the checkout template.
  • Days since the last performance regression, and who owns the budget.

A note on local conditions

Bangladesh traffic is not evenly distributed across devices. A site that feels fast on a fibre connection in Dhaka can fail on a mid-range Android phone using mobile data in a district town. Segment by device and connection type before drawing conclusions, and test the checkout flow on a throttled connection, because that is where a slow interaction turns into an abandoned order.

Where this fits in our work

Performance diagnosis sits inside the maintenance and monitoring work we do under Managed Technology & Cybersecurity, from server response time through to front-end assets. Scope and commercial terms are set out on the pricing page. If you are planning the annual technology budget, the SME IT budgeting guide shows how recurring performance work fits alongside security and backups.

Sources

Services that deliver this

Want this implemented properly?

Pricing →

Every engagement starts with a written scope. If a requirement is outside it, we quote it separately before starting.

Responses (0)

No comments yet. Add the first one below.

Leave a response

Your email address will not be published. Required fields are marked *.

Keep reading

More field notes