Website Speed Optimization: What to Check First
Investigate a slow website: measure loading, responsiveness and layout stability, identify the causes and check whether your changes actually help.

Contents

Investigate a slow website: measure loading, responsiveness and layout stability, identify the causes and check whether your changes actually help.
Share this article
Where Waiting Gets in the Way
You open a website on your phone. The main image has not appeared, the menu reacts late and the page shifts just as you tap the contact link. A page can be technically available while remaining difficult to use.
A universal three-second cutoff does not adequately describe that experience. Loading depends on the device, connection, cache and content. Instead of applying someone else’s bounce rate to your business, measure when the main content becomes visible and whether visitors can complete their task.
Consider Speed and Enquiries Together
A faster page can remove obstacles on the way to an enquiry. The effect on your business cannot be inferred from another company’s revenue statistic. An online store, a technical article and a local service page have different tasks and audiences.
Compare relevant page types and periods: can visitors reach the contact link, does the form work on a phone, and are there technical failures? Account for applicable privacy and consent requirements when collecting usage data. Changes in campaigns or content may also affect the number of enquiries.
Find the Actual Cause of the Delay
A CMS is not inherently slow, and a custom-built website is not inherently fast. Large images, additional scripts, slow server responses and expensive JavaScript are common places to investigate. Caching and embedded third-party content can make a substantial difference.
Start with the observed bottleneck. If HTML arrives late, inspect the server and its data access. If main content appears late, inspect images and blocking resources. If controls react slowly, investigate work in the browser. Measurements should inform the choice between improving an existing site and rebuilding it.
Use Caching and Rendering Deliberately
Static generation, server-side rendering and caching solve different problems. Content shared by all visitors can often be prepared in advance and distributed through a CDN. Personalized or frequently changing content needs suitable rules for refresh and access.
Prepared HTML does not solve every performance problem. A large hero image, expensive animation or excessive JavaScript can delay a static page too. Assess the entire delivery chain, from the first request and critical resources to the first usable interaction.
Three Metrics for a Usable Website
The Core Web Vitals cover distinct aspects of use. Google’s good thresholds, measured at the 75th percentile, are LCP within 2.5 seconds for the largest visible content, INP at or below 200 milliseconds for interaction responsiveness, and CLS at or below 0.1 for layout shifts. Assess mobile and desktop separately.
These metrics support diagnosis; they do not predict revenue. Good Core Web Vitals do not guarantee a top ranking. The page still needs to answer the visitor’s question well.
A Practical Technical Check
- Open the affected URL on a phone using a representative connection.
- Inspect which resource delays the main visible content in browser developer tools.
- For images, check file size, appropriate dimensions and loading priority.
- For delayed clicks, examine long JavaScript tasks and third-party scripts.
- Reserve space for images and dynamic content so controls do not move unexpectedly.
- After a change, measure the same page again under comparable conditions.
Lighthouse provides a controlled laboratory measurement. Real usage data adds evidence about different devices and connections. A single measurement on a fast development computer cannot replace that view.
Define a Useful Starting Point
Choose a complete visitor journey for the first investigation. For a service business, that might be a service page followed by a contact form. A product offering might also require an image gallery. A fast homepage tells you little about whether this particular task works without delays. First record the affected address, the device and the action that feels slow.
Distinguish a first visit from a repeat visit. On the second visit, images, fonts and scripts may already be in the browser cache. Both situations matter, but they answer different questions. Keep conditions as consistent as possible when comparing changes. A measurement on an office computer over the internal network is not a reliable comparison with a phone on an unstable connection. Repeat unusual results before using a single outlier to justify a large rebuild.
Make the choice of pages explicit too. Start where a visitor makes an important decision, then add an article or an image-heavy page. That lets you investigate different technical patterns without turning every measurement into a separate project.
Match Image Delivery to Visible Value
A large hero image needs different treatment from an illustration far down an article. The first image may be part of the main visible content. If the browser requests it late, the introduction stays unnecessarily incomplete. An illustration below the first screen usually does not need the same loading priority. Let the image’s actual position and purpose guide that decision.
Check the network log to see which file the browser really requests. A mobile layout does not necessarily need the same image width as a large desktop display. The delivered variants, displayed dimensions and result on the device all matter. A smaller file is useful only if the subject and any labels remain understandable. For a diagram, sharp lines may matter more than photographic detail.
Reserve space for the image before it appears. Otherwise text and controls can move as it loads. Reducing file size alone will not reliably fix that movement. Dimensions, aspect ratio and surrounding layout must work together. Check the same behavior for images that editors can add later, so the improvement survives everyday content changes.
Assess Scripts and Animation by Their Purpose
Forms, galleries and product demonstrations can all make useful use of JavaScript. Problems arise when many functions start at once or a long-running task keeps the browser busy. A visible button may look ready while responding only after a delay. Download size alone does not explain that behavior: executing the scripts also takes work.
Connect each substantial script to a function. Is it needed for the first visible screen? Does it need to run on every page? Could a rarely used feature be prepared when it becomes relevant? These are concrete implementation questions. A blanket instruction to remove every animation could discard a useful explanation without addressing the actual bottleneck.
Check embedded services separately. Calendars, videos and additional analytics can introduce resources and processing. Assess their effect under the consent states that actually apply. Preserve usability and accessibility while optimizing: a lighter menu still needs to work with a keyboard and touch, and movement should respect reduced-motion preferences. A faster page that prevents someone from completing the task has not achieved the intended improvement.
Caching Needs a Rule for Fresh Content
Caching avoids repeated work by reusing responses or files that have already been prepared. That is easy to understand for an unchanged font file. Product prices, opening hours or availability information add another requirement: how quickly does an edit need to appear? A fast response containing old information does not satisfy the website’s purpose.
Agree with the technical maintainer which content can be reused, for how long, and what triggers a refresh. Test the editing journey as well as loading: update content, wait for publication and open the affected page in a fresh browser context. Check multiple affected locations, such as a product listing and its detail page, if their responses are cached separately.
Restricted content needs separate consideration. Responses for signed-in visitors must not accidentally become shared public responses. Performance work should preserve those access boundaries. When a page combines public content and personalization, its architecture should clearly distinguish the parts that can be shared from those that depend on a particular visitor’s access.
Keep a Small Measurement Record
Before changing anything, record the problem you intend to solve. “Make the website faster” is too broad. A more useful observation is that the main image appears noticeably later than the text, even though it explains the introduction. That creates a testable question about when the image is requested and how it is prioritized. Choose an implementation change after identifying that question.
A short record can include the address, time, device, connection, baseline, change and follow-up measurement. Note whether the browser cache was populated and which external services were loaded. Change one related area at a time where possible. If images, fonts, tracking and server settings all change together, it becomes difficult to identify which intervention helped or caused a regression.
After the technical measurement, repeat the visitor task. Can people still use the navigation and submit the form? Is the content complete? Does readable text appear early? A better score is not a success if important content disappears or a function breaks. Record both the measurement and the practical check, then agree when future content changes should trigger another review.
Choose Between Improvement and Rebuilding
An existing website can often be improved selectively. If an oversized image or an unnecessary integration causes most of the delay, a complete rebuild is not yet a justified conclusion. Ask to see the bottleneck and discuss a limited intervention followed by a comparable measurement. That provides evidence for deciding what to do next.
A larger change can become reasonable when technical constraints affect maintenance, usability and performance together. Keep requirements concrete even then. Which content must remain? Which addresses should stay stable? How will publication work afterward? Who will check forms, redirects and mobile layouts following the transition? Performance is one requirement among several that make a website useful.
Planfold develops custom websites and discusses scope in terms of these tasks. A particular technology is not proof of a particular loading time. The relevant question is whether the agreed implementation works under understandable conditions and can be maintained afterward. A useful starting point for a conversation is one affected URL and a short description of the difficulty you observed.
Plan One Concrete Improvement
Start with an important page and an observed difficulty, such as a late main image or an unresponsive contact form. Record the baseline, the change and the follow-up measurement. That makes it possible to see whether the work achieved its purpose.
The guide to technical SEO places performance in the context of crawlability and page structure. If you want to modernize your website, discuss which improvements fit its purpose through Websites and Modernization.


