Running Your Own Website Check From Start to Finish

A loading bar that takes longer than two and a half seconds to paint the largest visible element already fails the Core Web Vitals thresholds — and that single number is where most small-business sites stumble on the first run. If nobody has touched the website since the agency invoice was paid, chances are nobody knows where it actually stands technically. The steps below check that without writing a line of code, in the order that actually produces useful results.

1. Measure Core Web Vitals with PageSpeed Insights

Google PageSpeed Insights is free and only needs a domain. It returns two kinds of data: a lab score from Lighthouse, simulated on Google's servers, and, if traffic is high enough, real field data from the Chrome user base. Three figures matter: Largest Contentful Paint (LCP, target under 2.5 seconds), Interaction to Next Paint (INP, target under 200 milliseconds) and Cumulative Layout Shift (CLS, target under 0.1), all defined in web.dev's INP documentation since INP replaced the older First Input Delay metric in March 2024. Skip this step and there is no way to tell whether a problem exists at all, or whether the site just feels slow to whoever is looking at it.

Run the test separately for desktop and mobile. The gap can be wide, because phones compute slower than the laptops most sites get built on. A joinery's site scoring one second on desktop and four seconds on mobile has a real problem — one that never shows up on the office monitor, even though that is exactly where most people search for a phone number on their way to a job.

2. Check mobile usability in Search Console

Google retired its standalone Mobile-Friendly Test tool; the check now lives inside Google Search Console, free once the domain is verified via a DNS record or an HTML file. The Mobile Usability report there flags text that is too small, elements placed too close together, or content wider than the screen. Without Search Console access this entire category of problems stays invisible, since none of it ever appears on a desktop screen. Setting it up via DNS usually takes a few minutes, provided someone still has access to the domain settings at the registrar — if not, the step can drag on for days while that access gets tracked down.

3. Crawl the structure

Screaming Frog SEO Spider is a desktop application whose free version, according to the manufacturer, crawls up to 500 URLs — plenty for most small-business sites, not nearly enough for a shop with several thousand product pages. The crawl surfaces broken internal links, duplicate or missing meta descriptions, and page titles repeated across the site because a template got copied without changes. After a website relaunch, old link targets pointing nowhere are a common find; which redirects tend to get forgotten in that process is covered in the piece on redirects lost after a relaunch. Skip the crawl, and only the pages someone still remembers get checked — not the ones quietly producing dead links for years.

4. Read the waterfall chart

PageSpeed Insights says something is slow. A waterfall chart shows why. GTmetrix offers a free analysis without login and lists every single request with its load time — often it's one oversized header image or a third-party script holding everything up. The fact that a 1.5-second delay cost one electrical contractor roughly €14,000 a year in estimated losses is a reasonable argument for taking this extra look, even after PageSpeed already gave a rough score. The waterfall chart doesn't fix anything by itself, though — it only points to where a look at the code or the hosting plan is worth the time.

5. Test the contact channels in person

Technical checks only cover half the picture. Send a message through the contact form, a WhatsApp message, and an inquiry via Google Messages — ideally from a second phone number or device, not from an inbox that already knows the sender. This reveals whether form submissions land in a spam folder, whether a WhatsApp number is even still active, or whether an inquiry simply goes unanswered because nobody checks that old inbox anymore. Which channel actually gets through for which kind of business is laid out in detail in the comparison of WhatsApp Business, contact forms, and Google Messages. A technically fast site that drops every inquiry hasn't gained anything.

6. Bring the findings together

Four tools, four interfaces, four different ways of presenting numbers — at this point, one consolidated report is more useful than four open browser tabs. The free website check from Log-System Development bundles load time, basic SEO signals, and core technical metrics into a single overview, which suits anyone who doesn't want to repeat this whole routine every month but wants one clear status instead. For whatever follows from that — a website actually built to generate inquiries, or individual tasks worth shortening with AI — it's worth looking at what Log-System Development builds directly with developers, skipping the usual agency markup. None of this replaces a proper review of your own, though: no automated tool, including this one, judges whether a privacy policy or an imprint actually meets legal requirements. That part stays manual, or goes to a lawyer.

Where this usually goes wrong

The check tends to stop after the first run. PageSpeed scores shift by several tenths of a second between measurements, because server load and network conditions play a role — one bad score isn't proof of a problem, and one good score is no guarantee either. Testing on three different days before touching anything on the site produces a more honest picture. And fixing the issues found without testing again afterward wastes half the effort: whether a compressed image actually pulls LCP under 2.5 seconds only shows up in the second run. Anyone who considers the job done after the first report has usually read the symptoms, not fixed the cause.