A slow website causes you to lose customers before they even get a chance to read your offer. It increases the bounce rate, lowers conversion rates, and on mobile devices or with slow connections, the impact is magnified. If you run a small or medium-sized business in Barcelona or anywhere else in Catalonia, you already know this problem exists. The hard part is deciding which service to hire and how to tell if it’s worked.
The key is to match each symptom with the technical specialty that can resolve it. A server problem isn't fixed by compressing images, and an overloaded database isn't improved by a caching plugin.
Key points:
- Recommended thresholds: LCP ≤ 2.5 s, INP ≤ 200 ms, CLS ≤ 0.1.
- Google evaluates the 75th percentile of actual visits, not lab tests.
- First, identify the bottleneck; then optimize it.
- Require a written assessment that identifies priorities before approving changes.
- It takes about 28 days for the field data to reflect the improvements.
- With Odoo or Docker, check the server and PostgreSQL in addition to the front end.
Why does your website load so slowly?

A slow website is almost always caused by one of these four factors: the server takes too long to respond, the resources are too large, the browser has to process too much code, or the application is waiting for a response from a slow database. Each cause has different symptoms and requires a different technical approach.
The server responds slowly
If the time to first byte (TTFB—the interval between the request and the arrival of the first data) consistently exceeds 600–800 ms, the problem lies before the browser begins to display content. Typical causes include an overloaded shared hosting environment, the absence of page caching, or a server located far from your audience. Slow DNS resolution also adds milliseconds at this stage, although it has less of an impact.
The resources are too heavy
Images of 2–3 MB uploaded directly from the camera, background videos, or full font sets cause the page size to skyrocket. Even if the server is performing well, the LCP (Largest Contentful Paint, the moment the main element appears) still skyrockets on mobile devices.
The code crashes the browser
Too much JavaScript (carousels, chats, tracking pixels, visual builders) blocks the browser's main thread. The symptom is a website that appears to have loaded but does not respond when you click a button, which worsens the INP (Interaction to Next Paint, the speed of response to an interaction). If elements shift while the page is loading, the CLS (Cumulative Layout Shift) also suffers.
How to Diagnose a Slow Website
Diagnosing a slow website starts with Google PageSpeed Insights and the Core Web Vitals report in Search Console, and continues with server-side measurements. The goal is to identify the bottleneck before making any changes, because optimizing blindly wastes budget without improving the metrics that matter.
To get a reliable starting point, follow these steps:
- Use PageSpeed Insights to analyze your homepage and the two or three templates that generate the most revenue (product page, service page, contact form), always on the "Mobile" tab, because Google indexes the mobile version first.
- Check the Core Web Vitals report in Search Console, which groups URLs by template and shows which ones have good, average, or poor results.
- Measure the TTFB from various locations and at different times of day; if it spikes during peak hours, the server or database isn't up to the task.
- Open the Network tab in your browser's developer tools and sort the requests by size and duration to see which resource is slowing down the load.
- Record the results, including the date, device, and exact URL, because without that information, you won't be able to demonstrate any improvements later on.
Tools such as GTmetrix and Pingdom supplement this analysis with detailed waterfall charts of the loading process. Even so, PageSpeed Insights remains the primary benchmark because it combines lab tests and real-user data into a single report.
Step 3 is usually the most revealing for an SME. If the TTFB goes from 300 ms in the morning to 1.5 s at noon, compressing images won't make up for that delay. In that case, the issue should be addressed by a systems administration specialist, not a web designer.
How to Interpret Google PageSpeed Insights
PageSpeed Insights displays two sections: at the top, field data from real Chrome user visits (CrUX) over the past 28 days; at the bottom, a lab test using Lighthouse under simulated conditions. Both are useful, but they answer different questions, and confusing the two is the most common mistake when reading the report.
Field Data vs. Laboratory Tests
Field data reflects your users' experience and is what Google uses to evaluate a page. They are evaluated at the 75th percentile: to pass, at least three out of every four visits must fall within that threshold. Lab tests, on the other hand, are used to detect problems because they show which specific resource is causing the delay and allow you to repeat the measurement after each change.
A website with low traffic may not have enough real-world data. In that case, lab tests are your only initial guide, and you should configure the Web Vitals library to collect your own real-world data.
Which figures should you look at first?
The score from 0 to 100 is for reference only; the metrics are what matter. The recommended thresholds for Core Web Vitals are LCP ≤ 2.5 s, INP ≤ 200 ms, and CLS ≤ 0.1. A page might score 94 in lab testing but have an LCP of 2.7 s, which is slightly above the threshold. A high score does not guarantee that the page will pass.
Prioritize LCP and INP, because these are the metrics that have the greatest impact on perceived speed. CLS can usually be resolved with minimal effort: simply reserve space for images, ads, and iframes.
What should a web performance optimization service include?
A professional web performance optimization service consists of four components: a documented diagnosis, a plan with priorities based on impact, technical implementation on the server and front end, and follow-up measurement using real-world data. If a proposal jumps straight to installing plugins or changing the template, there is no diagnosis behind it.
Assessment and Priorities
The first document you should receive is a report detailing the baseline metrics, the identified bottleneck, and a list of actions ranked by impact and effort. A well-conducted audit first collects data, then analyzes it, and concludes with a list of actions in order of priority. Skipping the data collection step leads to incomplete diagnoses and inflated budgets.
Front-end Changes
Images should be served in modern formats such as WebP or AVIF and resized to their actual display size. Lazy loading (loading="lazy") is only recommended for images that are outside the initial viewport. The main image, which is usually the LCP element, should be loaded as soon as possible, using `fetchpriority="high"` or preloading, and never with lazy loading.
As for JavaScript and CSS, the task involves removing unused scripts, deferring non-critical ones, extracting critical CSS, and reducing HTTP requests. The usefulness of each third-party script (chat, maps, pixels) must justify its impact on the INP.
Cache and Infrastructure
A comprehensive service operates on three levels of caching: browser caching (Cache-Control headers with a long expiration time for versioned static resources), page caching (pre-generated HTML or HTML served from a reverse proxy such as Nginx or Varnish), and server or object caching (Redis for repeated queries). Added to this is an assessment of the hosting environment: PHP or Python version, HTTP/2 or HTTP/3, Brotli compression, and, if the audience is geographically dispersed, a CDN.
He demands that the proposal specify which levels it covers and which are excluded. A service that only affects the front end will not resolve a TTFB of 1.5 seconds.
How to Tell the Difference Between a Hosting Problem and an Image or Code Problem

The quickest way to identify the causes is to compare the TTFB with the rest of the load time. If the first byte takes longer than 600–800 ms, the problem lies with the server or the application. If the TTFB is good but the LCP is high, the bottleneck is in the images, fonts, or CSS that are blocking the load. If everything loads quickly but the website is slow to respond to clicks, JavaScript is the culprit.
Let's consider a store with a TTFB of 250 ms, an LCP of 4.1 s, and a cover image of 1.8 MB. The server isn't the problem. Converting that image to WebP so it’s about 150 KB and preloading it usually reduces the LCP to under 2.5 seconds without changing the hosting.
The opposite scenario would be a corporate website with images that are already optimized, a TTFB of 1.4 s, and an LCP of 3.2 s. Subtracting the TTFB, the rest of the load takes 1.8 s, which is a reasonable figure. In this case, the best course of action is to enable page caching or migrate to a server with dedicated resources.
When both symptoms occur at the same time, start with the server. Every improvement in TTFB reduces the LCP for all pages at once, while images are optimized on a page-by-page basis.
What Should You Check If Your Website Uses Odoo, PostgreSQL, or Docker?
If your website or online store runs on Odoo, its performance depends on the front end, the PostgreSQL database, and the server configuration. A service that focuses solely on web design won't review these layers, and that's where the causes of a high TTFB are often hidden.
Odoo and PostgreSQL Performance
In PostgreSQL, it's a good idea to review parameters such as shared_buffers (typically around 25% of the server's RAM), work_mem, and effective_cache_size, as well as the autovacuum status. Tables with many dead rows and unindexed queries slow down product pages and the checkout process. Enable pg_stat_statements to identify the the most time-consuming inquiries.
In Odoo, the number of worker processes must match the number of available cores (a common rule is two per core plus one), with memory limits defined for each process. A multithreaded deployment without configured workers is a common cause of slow performance under load.
Linux Servers and Docker Deployments
For Docker deployments, check the CPU and memory limits for each container, the storage used by PostgreSQL volumes (NVMe SSDs vs. slow network-attached drives), and whether there is a caching and compression reverse proxy in front of Odoo. Also check the Linux server’s swap memory usage: if the system is continuously transferring data from memory to disk, response times will skyrocket.
For this type of website, Synergia deploys Odoo on dedicated servers using Docker and PostgreSQL and offers web performance optimization, with on-site and remote support in Barcelona and an SLA with a two-hour response time. Its services cover both infrastructure and technical SEO.
Criteria for Choosing an Optimization Provider
Choose a provider that diagnoses the problem before providing a quote, that addresses the specific area where your bottleneck lies, and that commits to measuring results using field data. These three criteria allow you to rule out most generic proposals.
When comparing proposals, consider the following points:
- Written diagnosis including baseline metrics, analyzed URLs, and the identified root cause.
- Technical scope based on your symptoms: server and database if the TTFB is high; frontend if the LCP fails despite a good TTFB.
- A measurement plan that includes PageSpeed Insights, Search Console, and a tracking period of at least 28 days.
- Demonstrable experience with your platform (WordPress, Odoo, custom development) and documented access to the areas being modified.
The most common objection is the cost: many small and medium-sized businesses think a free plugin is enough. A caching plugin helps when the problem is related to caching, but it doesn’t fix slow PostgreSQL queries or a server with insufficient memory. Paying for a thorough diagnosis prevents you from investing in the wrong solution and usually pays for itself as soon as you recover mobile conversions.
Also ask the vendor to specify what they won't be working on. Knowing that the visual template or marketing scripts are off the table prevents surprises and allows you to coordinate with other teams.
How can you tell if the improvements have worked?

Improvements are verified by comparing the same pages on the same device under equivalent conditions and by confirming the changes in field data. A single lab test conducted after the change proves nothing if the initial test was performed on a different URL or with a different network profile.
For a reliable comparison, repeat the lab tests on the URLs listed in the diagnosis using the mobile view in PageSpeed Insights and at a similar time of day. Take three measurements and calculate the median, because the Lighthouse score can vary by several points from one test to the next.
Next, monitor the field data. Since CrUX aggregates data over 28 days, an improvement implemented on December 1, 2026, won’t be fully reflected until the end of the month. The Core Web Vitals report in Search Console lets you verify the fix by URL group and track the progress from poor to good results.
Round out your metrics with business indicators: mobile bounce rate, submitted forms, and conversions. Speed is a means to an end, and technical SEO and user experience both improve when the website loads quickly.
Where to Start
Before requesting quotes, spend half an hour measuring your homepage and your two key templates on PageSpeed Insights and note down the TTFB, LCP, and INP. With those three metrics, you’ll know whether you need a server specialist, a front-end specialist, or both, and you’ll be able to compare proposals effectively. If your website runs on Odoo or Docker, include a review of PostgreSQL and the server in your request.
Frequently Asked Questions
How long does it take to see the results of optimizing a slow website?
Laboratory tests show changes immediately, but Google's field data takes about 28 days to reveal the full effect. Plan to track the results for at least one month before evaluating the final outcome.
Does a good score on PageSpeed Insights guarantee a good ranking?
No. The score is based on lab tests, while Google uses field data from Core Web Vitals as a signal of page experience, which complements content and links without replacing them.
Will switching hosting providers fix a slow website?
Only if the problem is on the server—that is, if the TTFB consistently exceeds 600–800 ms. With a good TTFB, the slowness is caused by images or the code, and switching to a new hosting provider won't fix it.
Should we use lazy loading for all images?
No. Deferred loading is appropriate for images that are not part of the initial viewport. The main image should be loaded as soon as possible, with high priority or preloading, because delaying it directly worsens the LCP.

