Slow is a bug. It just has nobody assigned to it.
We make websites fast, and we do it the boring way: measure first, find what the time is actually being spent on, change one thing, measure again. You get the numbers from before and after, and the list of what we did to move them.
Any stack, any host. We do not need to have built it, and you do not have to move it.
What we do
Twelve places the time actually goes.
A slow site is rarely slow for one reason, and it is almost never the reason people guess. These are the twelve we look at, in the order the measurement puts them.
Speed optimisation
The whole job, and the only honest way to start it: a real measurement of a real page on a real connection, before anybody has an opinion about what is wrong.
Core Web Vitals
Largest Contentful Paint, Interaction to Next Paint and Cumulative Layout Shift — what a visitor actually experiences, and what search engines actually read.
PHP tuning
Version, opcode cache, workers, memory limits and the settings that ship as defaults on shared hosting and stay that way for the life of the site.
MySQL tuning
Buffer pool, connections, temporary tables and the slow query log — turned on first, because a database nobody is logging is a database nobody can defend.
Caching
Page, object and browser caches that actually hold, with rules for the pages that must never be cached. A cart served from cache is worse than a slow one.
CDN configuration
Static assets served from near the visitor, with cache headers, compression and TLS set properly at the edge — not a switch flicked on and left to guess.
Image optimisation
Modern formats, real dimensions, correct sizes for each breakpoint and lazy loading everywhere except the one image the visitor is waiting for.
Code optimisation
Blocking scripts, render-blocking styles, fonts that hide text while they load, and the third-party tag somebody added in 2021 that is still costing a second.
Database optimisation
The queries behind the slow page: indexes that are missing, joins that scan, tables that have grown by a factor nobody planned for, and rows that should have been deleted.
Mobile performance
Measured on a mid-range phone on a mid-range connection, which is the device most of your visitors have and the one nobody tests on.
Technical SEO performance
Crawlability, rendering, redirect chains, duplicate URLs and the field data search engines collect from real visitors rather than from a laboratory run.
Performance audit
The whole of the above written down: what is slow, why, what each fix would take, and what it is worth. Yours to act on with us or with anybody else.
Not sure where your time is going? That is what the audit is for, and it is the only way anybody could tell you.
Ask for an auditWhat gets measured
Three numbers a visitor can feel.
Core Web Vitals are not a score somebody invented to sell audits. They are three specific measurements of what it is like to use your site, and search engines read them from real visits to it.
LCP
How long until it looks loaded
- Slow server responses and no page cache
- A hero image with no dimensions, or lazy-loaded
- Fonts and stylesheets blocking the first paint
- Redirect chains before the page is even requested
INP
How long until it answers a tap
- Long JavaScript tasks holding the main thread
- Third-party tags running on every interaction
- Handlers doing layout work on every keystroke
- Sliders and menus rebuilt instead of shown
CLS
How much it moves while you read
- Images and embeds with no reserved space
- Banners and notices inserted above the content
- Web fonts swapping in at a different size
- Anything injected after the first paint
The thresholds for all three are published by Google and they change — First Input Delay was replaced by INP in March 2024, which is how a figure printed on a marketing page becomes wrong while nobody is looking. So you will not find one here. Your report carries your site's numbers, measured on the day, against whatever the thresholds are that day, from the laboratory and from real visits alike — because the two disagree more often than anybody selling speed admits.
The honest part
Making the number go up is easy. That is the problem.
A performance score can be improved without making anything faster, and most of the ways are one afternoon's work. Here is what we will not do, so you know what you are getting when the number moves.
Ways to score well and feel slower
- Lazy-loading the one image the visitor is waiting for, so the score improves and the page looks empty for longer
- Deferring the scripts a page needs, so it paints early and does nothing when tapped
- Removing a feature rather than fixing it, and not mentioning which one
- Tuning only the home page, which is the page the test tool opens and often not the page anybody lands on
- Serving a stripped-down page to the test agent and the real one to visitors
- Turning on aggressive caching that also caches the cart, the login and the prices
- Blocking analytics or consent tools so the laboratory run looks clean and the real one does not change
What we do instead
-
Measure
Laboratory and field data both, on the pages people actually land on, on a mid-range phone. Before anybody has an opinion.
-
Find where the time goes
A waterfall, a profile and the slow query log. Not a checklist of generic advice that happens to fit any site.
-
Change one thing
One change at a time, so the improvement can be attributed and the regression can be found. Bundled changes are how nobody ever learns what worked.
-
Measure again
Same pages, same device, same conditions. A number measured a different way is not a comparison, it is a new number.
-
Stop when it stops paying
There is a point where the next hour buys less than it costs, and we will tell you when we reach it rather than keep going.
-
Hand you the numbers
Before, after, and what each change was worth — including the ones that turned out to be worth nothing.
Send the address of the slow page — not the home page unless that is the slow one — and we will tell you what is costing the time.
Ask for an auditWhat comes back
A report you could hand to somebody else.
Which is the test of whether it is worth anything. None of the following is an upgrade or a tier — it is what an audit is.
The measurement, not a screenshot of one
Which pages, which device, which connection, and on what date. Repeatable by you or by anybody you show it to.
Laboratory and field, side by side
A test run and what real visitors actually experienced. When they disagree, that disagreement is the finding.
Findings in cost order
What is slow, why, and what each fix would take — sorted by what it buys, not by what is easiest to sell.
What we would not touch
The things that look wrong in a report and are not worth fixing on your site, said plainly so nobody sells them to you later.
The numbers after, in the same shape
Measured the same way as the numbers before. A comparison against a different method is not a comparison.
Yours either way
The audit is a deliverable, not a sales document. You can act on it with your own developer and never speak to us again.
How it goes
Five stages, and you can stop after the second.
The audit stands on its own. Nothing about it assumes you will have us do the work afterwards, and it is written so that somebody else could.
- 01 Access Read access to the site and, where the work needs it, the hosting. Nothing is changed at this stage.
- 02 Measure Real pages, real device, laboratory and field. This is the number everything afterwards is compared against.
- 03 Report What is slow, why, what each fix costs and what it buys. Yours, whether or not we go on.
- 04 Work One change at a time, on a copy first, with the site staying up and behaving the same.
- 05 Re-measure The same pages the same way, and the difference written down beside what caused it.
Straight answers
The questions worth asking before you write to us.
No, and anybody who says yes is selling a checklist. The whole method here is that the measurement comes first: the reason a site is slow is different every time, and it is almost never the thing people guess before looking.
We will not answer that before measuring, and the studios that do are quoting somebody else's site. What we can tell you after the audit is what each fix is worth in the numbers we measured, so you can decide which ones are worth doing.
No. Most of the work is in the site rather than under it, and we do this on hosting we do not sell and cannot see the inside of. If the server really is the limit we will show you the measurement that says so, and it is still your decision.
No. A score is a summary of measurements, and it can be improved without making anything faster — the section above this one lists the usual ways. The goal is the page feeling quicker to the people who use it, which is what the field data measures and what a laboratory run only approximates.
The work happens on a copy first, a restore point is taken before the live site is touched, and every change goes on one at a time so a regression has one suspect. Performance work that removes a feature is not performance work, and we will not do it without telling you which feature and asking.
Usually not on its own. What is usually the problem is a page builder rendering through forty plugins, a theme loading everything on every page, and images uploaded at camera resolution. All three are fixable without leaving WordPress, and we will say so if leaving it really is the cheaper answer.
Speed is one input among many and nobody outside the search engines knows the weights, so anyone promising you positions is guessing. What is true and checkable is that Core Web Vitals are collected from real visits to your site and are used, and that the technical work here — crawlability, rendering, redirect chains, duplicate URLs — is the part of SEO that is engineering rather than opinion.
Until somebody adds a plugin, a tag or a hero video. That is not cynicism, it is how sites work — which is why the audit tells you what to watch, and why continuous monitoring belongs with maintenance rather than being sold as part of a one-off speed job.
Free audit request
Send us the slow page. We will tell you where the time goes.
The address of the page that is actually slow is worth more than a paragraph describing it. You get a real answer from the person who would do the work.