A green performance score is reassuring. It is not a complete account of what customers experience when they open your site, tap a menu and try to submit a form. Core Web Vitals are useful because they separate three everyday frustrations: waiting for the main content, waiting for a response and watching the page move under your finger.
This guide is for Australian business owners who need to commission sensible fixes, not become browser engineers. Start with the journey that earns an enquiry or completes an order. A fast homepage is not much comfort when checkout remains painful.
The three metrics, without the jargon fog
Google’s Web Vitals documentation defines the current metrics and recommended good thresholds: Largest Contentful Paint (LCP) at 2.5 seconds or less, Interaction to Next Paint (INP) at 200 milliseconds or less, and Cumulative Layout Shift (CLS) at 0.1 or less. Assess the 75th percentile of visits, segmented by mobile and desktop. These are not averages and they are not targets for just your fastest visitor.
LCP asks when the largest visible content element has rendered. A prominent heading or hero image may be that element. A slow initial server response, delayed image discovery or render-blocking work can hold it back. That is why optimising a logo that appears elsewhere on the page may leave LCP unchanged.
INP describes responsiveness across interactions during a visit. A page can look finished and still respond slowly when someone opens a navigation menu or changes a product option. Expensive JavaScript work and large rendering tasks are worth investigating. INP replaced First Input Delay in March 2024; an old report that only discusses FID is not a complete current assessment.
CLS measures unexpected visual movement. An image without reserved space, a late-loading banner or a changed font can shove a button somewhere else. Reserving dimensions and planning space for dynamic content are often better fixes than telling customers to wait until the page settles.
A lab test and a customer measurement answer different questions
A lab test gives you a controlled reproduction and diagnostic evidence. Field data describes real visits, including the devices and connections you do not control. Google explains both measurement approaches and notes that Lighthouse without user interaction cannot directly measure INP; Total Blocking Time is a useful lab proxy, not the same metric.
PageSpeed Insights can show both field information and a Lighthouse lab result. Read the labels before comparing them. A new or low-traffic page may not have enough eligible field data for a page-level assessment. That is missing evidence, not proof of failure. An origin-level result may describe the broader website rather than the exact URL you entered.
For a concrete investigation process, read the speed-audit walkthrough. Keep the test location, device profile, cache state and page version with the result. Otherwise two screenshots taken on different days can create a completely imaginary improvement.
Do not confuse a score with a ranking guarantee
Google Search Central recommends good Core Web Vitals, while explaining that they form part of the wider page-experience picture. Improving them does not guarantee a particular position in search. Relevant content, a usable page and a credible business remain important.
A perfect Lighthouse score is not a promise of sales either. Use performance evidence alongside completed enquiries, purchases and support complaints. If a form fails validation or visitors cannot understand the offer, a faster delivery of that problem will not fix it.
For WordPress, diagnose before installing another plugin
Separate the server wait from browser work. The hosting-performance guide covers server response, caches and resource limits. A change in hosting may help a slow uncached request; it will not automatically remove a large video, ten third-party scripts or a menu doing expensive work in the browser.
Test representative templates: homepage, service page, article, product, checkout and a logged-in page if relevant. Page caching is valuable for public content, but a cart or personalised account page needs separate testing. Do not cache private content just to make a benchmark look good.
For loading, identify the actual LCP element, then check how early its resource is discovered and whether it is delayed. For responsiveness, reproduce a real action and inspect the work blocking it. For layout stability, watch the page load and reserve space where content arrives late. Google’s LCP optimisation guide provides the technical sequence rather than a generic shopping list.
Commission a fix with a testable acceptance criterion
Ask your developer to state the symptom, likely cause, proposed change and how it will be measured. Specify the important templates and confirm that contact forms, navigation and checkout still work. A useful report includes before-and-after measurements under comparable conditions, not a single dramatic score.
Keep an accessible workflow in the acceptance test. Removing keyboard focus styles can make a page look tidier while harming users. Disabling every script can improve a synthetic result while breaking the form that generates enquiries. The accessibility requirements guide explains why a performance pass and an accessibility pass are separate questions.
Change one meaningful group of things at a time, keep a rollback path and check the result after release. Field measurements take time to reflect changed experiences; do not claim an instant field-data victory from a fresh lab test.
What to ask for in the final report
- The tested URLs, device profiles and cache states.
- The measured LCP, responsiveness evidence and layout-shift causes.
- The customer task the change improves.
- The exact changes and any remaining trade-offs.
- A functional test covering forms, checkout and keyboard navigation.
- A monitoring and rollback plan if later plugin or content changes regress performance.
If you want help deciding where to spend the next hour, contact Matt. I would rather identify one measurable bottleneck than sell a pile of plugins that makes the dashboard look busy. For the financial decision behind a fix, see the business-cost guide for prioritising the work.
Featured photograph: Luke Chesser / Unsplash.
