Most speed work is a plugin install and a screenshot of a score. Mine is measured against your real site, one change at a time, and the site stays up while it happens.
Every speed service sells the same thing: install a cache plugin, run a testing tool, send you the green number. The number moves. The site often does not. What I measure is what your server actually does when a browser asks it for a page, because that is the part your visitor sits through.
It is usually several small ones stacked, and the order they get fixed in is the whole job. Cache goes last because it hides everything else: turn it on first and every later change measures the same, since you are timing the server’s hard drive instead of your site. Most people start there, because it is the step that shows a fast number by tomorrow.
LiteSpeed on the server and Cloudflare in front of it. That is the whole list, not a stack of plugins layered on each other. If a problem cannot be solved with those two, it does not get solved by adding a third: it gets documented, and told to you straight.
A shop that caches the wrong page shows one customer another customer’s basket. It does not throw an error and it does not appear in any log: the site looks fine while it leaks. So nothing goes live on a shop until the test has actually been done, with a real basket, not a checkbox ticked in a settings screen.
Comercial Yobeda is a WooCommerce shop carrying 279 products. It had no page cache at all, so every single visit rebuilt the whole page from scratch. Same server, same day, same method, measured before the work and after it. That is the only comparison worth showing you, because it is the only one where both numbers describe the same machine.
No before-and-after on these three, because there was no before worth showing: they have been kept on this footing since they launched. So this is simply what they answer in when I measure them, from my office in Spain, all on the same day and the same way. Read it as that and nothing more: it is not a promise about your visitor’s phone on a bad connection.
This will sound strange in a market where everyone sells it as an add-on: if I built your site, or you are on a care plan, the speed work is already in it. Not a package, not an upsell, not a performance tier. It is included because a site that takes four seconds to answer is not doing its job, and because keeping every site I run on the same footing is exactly why I can show you the numbers above instead of a case study from three years ago.
“Daniel set up our store and SEO completely. Within 3 months we were selling consistently online — he handles everything.”
“Our old website was years behind the business. Daniel rebuilt it from scratch and now runs our SEO — one point of contact, and things move without chasing.”
Because a score and a speed are different measurements. Tools score how a page is built against a checklist. I measure how long your server takes to answer. I have cut a shop’s answer time by three quarters, and I cannot tell you what that did to its score, because I do not measure scores. Ask what the site does, not what it rates.
It can, if it is done all at once. Which is exactly why it is not. One change, verify, next change, and the settings most likely to break something are the ones left off by default rather than the ones led with. On a shop, the basket and checkout are tested with a real product before anything is left switched on.
Usually no. Most of what makes a site slow lives inside the site, not underneath it. If your hosting really is the ceiling, I will say so and show you the measurement that proves it, instead of selling you a migration.
Days, not weeks — but it depends on how many different page templates there are, not how many pages. Two hundred product pages built from one template get tested once. Six different layouts are six rounds, and I will tell you which you have before starting.
Your own before and after numbers, the list of what changed, and the exclusion list, which is the part that keeps the site from breaking the next time somebody touches it. No tool screenshots, because they are not what was measured.
Where the speed work sits, and what else is in the same plan.
One reply with what it does now, in plain numbers, and whether the problem is worth fixing. If it is not, I will tell you that too.
