A fast website lets people read the essential content, get a response to a click and reach the contact page without frustrating delays. Start with a real journey on a phone, then compare your observations with PageSpeed Insights. Keep real-user measurements separate from a single laboratory test.
Three questions instead of one score
You do not need to know how a website is built to identify a problem. Open a specific service page and ask: when can I see its main content, when can I take action, and do the elements stay in place? These questions cover three aspects of the experience: loading, responsiveness and stability. A page can appear loaded but fail to respond to the menu for a moment. It can also respond quickly while moving a button when an image appears.
This is different from assessing the offer's wording or checking whether all the necessary pages exist. Here, you are examining how the finished interface behaves. A slow first visit does not mean that the entire website needs rebuilding. First establish a repeatable symptom, where it happens and the conditions in which it occurs.
Run a short test on a real phone
Choose three URLs: the homepage, your most important service and the contact page. Use your normal phone, but record the model, browser and connection type. Check Wi-Fi and mobile data separately. A slower phone or weaker signal may reveal a problem that is not noticeable on an office computer. However, one visit in an area with poor reception is not proof of a website fault.
- Open the URL in a new private window. Note whether you first see useful content, an empty screen or a decorative animation.
- Read the first screen and try opening the menu. Look for a delayed response, not just a spinning loading indicator.
- Scroll to the images and price information, if available. Check whether newly loaded elements move the content you are reading.
- Go to the form and tap a field. Opening a form does not prove that submission works; testing delivery requires a separate, agreed test enquiry.
- Visit the same URL again. Record the difference between the first and a subsequent visit, because caching changes the conditions.
Write a specific observation, such as “the menu only opens after a second tap”, rather than “the website is sluggish”. If possible, record your screen without personal information. The recording shows the symptom to the developer; it does not replace technical timing measurements or an investigation of the cause.
Read PageSpeed Insights without overinterpreting it
PageSpeed Insights combines two types of information. Real-user data, where available, covers the previous 28 days. A laboratory test loads the page under specified conditions and helps investigate causes. The results can therefore differ without either being wrong. After a change today, the historical data will not immediately become a report covering only the new version.
Check whether the report describes the requested URL or the whole origin. With insufficient data, the tool may not show real-user results for an individual page. Missing data means neither a good nor a bad result. Do not compare mobile and desktop tests without explaining the difference. Keep the URL, date, mode and measurement conditions so that the next report answers the same question.
What the three main metrics mean
According to the Web Vitals documentation, the current Core Web Vitals are LCP, INP and CLS. Good thresholds are no more than 2.5 seconds, 200 milliseconds and 0.1 respectively. Real-user assessments use the 75th percentile, not the single best visit. In practice, showing one successful run on a fast laptop is not enough.
| Area | What the customer notices | What to ask the developer |
|---|---|---|
| LCP: main content | A long wait for the largest relevant element on the screen. | Identify the LCP element and explain why it appears late. |
| INP: responsiveness | Tapping a menu or button produces a delayed interface response. | Investigate expensive tasks during real interactions. |
| CLS: stability | Content jumps when an image, font or another element loads. | Identify the shifts and reserve space for elements that load later. |
These are diagnostic pointers, not a remote diagnosis. A large image may slow loading, but the issue could also involve the server or how resources are fetched. Do not automatically remove every script: some support the form, analytics consent or other necessary functions.
A fault report that can be verified
For each symptom, prepare one spreadsheet row containing the exact URL, device, connection, first or repeat visit, action taken, result and report link. Add the date and whether the behaviour occurred again. Separate observation from hypothesis. “The image appears late” is an observation; “the hosting is underpowered” remains a hypothesis that needs testing.
My recommendation: fix repeatable obstacles on the enquiry journey first, other delays second and cosmetic score improvements last. This is a recommended work order, not a promise of increased sales. A better speed measurement alone does not demonstrate more enquiries; that requires separate data and comparable conditions.
Accept improvements without chasing 100
Ask for before-and-after reports for the same URLs and a description of the changes. Repeat the normal customer journey: arrival, menu, service and contact. Confirm that images, information and necessary functions have not disappeared. If the score improved only because the form was removed, the optimisation has not passed a meaningful acceptance check.
Keep immediate technical acceptance separate from observing real-user data. Laboratory testing can help check a change straight away; a 28-day dataset will still include older visits. Do not require a guaranteed score of 100 on every run. Require documented measurements, resolution of the agreed faults and preserved functionality. That is a stronger basis for a decision than a single coloured score.
Sources and limitations
Metric thresholds and definitions: Web Vitals — Chrome team documentation. Real-user data, the 28-day period and laboratory diagnostics: Google — About PageSpeed Insights. This guide does not present any company's results or an audit of your website. Identifying causes requires examining the specific site.
Need to define the scope of a website redesign?
Describe the repeatable symptom, website URL and test conditions in the brief. This starts a conversation about the work required; it is not an automatic performance diagnosis. The brief form is in Polish.
Open the website brief (in Polish) ↗