The expensive version of a slow website is not a red circle in a slide. It is a customer on a mid-range Android, on 4G, on the A46 or the ring road, who asked Google for a job you can do — and left before the heading finished painting.
I will not quote a bounce-rate study I did not run, and I will not paste a Lighthouse number I invented for a demo firm. Those screenshots sell fear. I will translate Core Web Vitals into the language I use with owners, and say how a slow page costs an East Midlands business.
I am Harry Ashton. Hashton Agency is a Nottinghamshire studio. I build the site; there is no performance team behind a curtain. Live work is on proof. The crawl argument is in If it has no URL, search cannot find it. This is the speed argument: if the URL exists and still will not open, you have published a door that sticks.
Your customer is not on office fibre. Photo: Ono Kosuki / Pexels.
Speed is a customer problem first
Core Web Vitals are Google’s names for three feelings a person already has.
Largest Contentful Paint (LCP) is how long until the main thing arrives. On a plumber’s homepage that is usually the heading, the hero image, or the phone-and-form block. If the “main thing” is a four-megabyte photograph of a van, the customer is looking at a blank tab. They do not say “poor LCP”. They say the site is broken and try the next result.
Interaction to Next Paint (INP) is how long after a tap the page reacts. They hit “Get a quote”, “Book”, or a menu. If a chat widget, a tag manager, and a page-builder bundle are still arguing, the tap feels dead. INP replaced an older metric that only cared about the first input. Owners should care because the first useful tap is often the only one they get.
Cumulative Layout Shift (CLS) is the jump. The heading loads, then a cookie wall, then a webfont, then a banner, and the button they were about to hit has moved. They meant to call you. They opened a newsletter signup. They will not try again.
Google publishes thresholds for those three and uses field data — real Chrome users — more than a lab run on a quiet office connection. I will not pretend I know your field numbers from here. Open the live URL on the phone you actually carry, on the network you actually get between jobs, and be honest about whether you would wait.
The Nottingham page puts it the way I mean it: speed on a mid-range Android, not a trophy. The same sentence applies in Derby, Leicester, Mansfield, and Newark. East Midlands traffic is vans and car parks, not a fibre socket in a pitch meeting.
Where the money goes when the page is late
A slow page costs you in ways that never show up as a single invoice.
They call someone else. Local search is a shortlist. If your result is the one that spins, the shortlist shrinks by one. You do not get a notification that this happened. You get a quiet Tuesday.
Paid clicks land on a door that sticks. If you run Search ads, you are paying for the visit whether the form appears or not. I will not invent a cost-per-lead for a trade I have not seen. I will say the obvious thing: buying traffic for a page that will not open is paying to demonstrate that you are hard to hire.
Search has less reason to prefer you. Speed is not the only ranking signal and anyone who says “fix Vitals, win page one” is selling a story. It is one of the signals, and it is a quality signal for the visit itself. A crawlable page that will not paint is still a bad document. Technical SEO without performance is a sitemap of disappointments. I treat performance as a build constraint in services, not a later upsell.
Your own staff stop using it. When the public site is painful, people email a PDF, quote from a WhatsApp broadcast, or send an Instagram link. The official URL becomes decoration. That is how you end up with three prices and two versions of the same service.
You pay twice. Page builders and theme stacks are cheap to launch and expensive to make fast. The usual path is: launch, complain about leads, buy a plugin, buy another, then pay someone to rebuild. The second bill is the one that should have been the first.
None of that requires a study I did not run. You can watch it happen on your own phone.
Why East Midlands SME sites are often slow
The pattern is boring, which is why it persists.
A hero that is actually a film. Autoplaying background video, or a PNG exported from a poster. The owner likes how it looks on the office iMac. The customer is on a phone on the ring road.
A page builder’s idea of a homepage. Every section is a block. Every block brings CSS and JavaScript. The homepage becomes a tour of the vendor’s marketplace: sliders, counters, testimonials carousels, “meet the team” grids that never became pages. I wrote about unique URLs in the local search notes. The speed version is: every unused widget is weight.
Third parties with a key in the header. Chat, reviews, ads, maps, heatmaps, three fonts from two foundries, a cookie banner that loads a second banner. Each script can be defended in isolation. Together they make INP worse and they often block the thing the customer came for.
Images nobody resized. A gallery of workshop shots straight from a phone, served at full resolution, with no width and height, so the page jumps as each one arrives. That is LCP and CLS in one folder.
JavaScript as the page. Content that only exists after hydration. You can see it on a good connection. A crawler, and a tired phone, get a shell. That is a speed problem and a URL problem at the same time.
Hosting and cache as an afterthought. A cheap shared box, no cache headers, a plugin fighting a CDN. I will not name-shame a host. I will say that “it is fast in the office” is not a measurement.
EO Charging had to stay usable for depot staff who cannot wait on a spinner. A five-page trade site has no excuse to be heavier than an operational dashboard.
How to check without a trophy screenshot
You do not need my stack to see the problem. You need the same phone your customers have.
- Open your live URL in a private window, on mobile data, not the workshop Wi-Fi.
- Time, roughly, until you can read the service you sell and a way to contact you. If you are embarrassed, that is data.
- Tap the main button. Does it react, or does the chrome freeze?
- Did the layout jump? Did a bar eat the first tap?
- Disable JavaScript if you can. Can you still read the offer? If not, you have a demo.
- If you use Search Console or a CrUX-based report, look at mobile field data over a recent period. Treat it as weather, not a score to game. I will not publish a target I have not measured on your host.
PageSpeed Insights is fine as a torch. It is a bad contract. Agencies that sell “we will get you a 100” are selling a lab. Labs are useful for finding a 3MB image. They are not your customers.
If you want a written pass on the current URL — what is heavy, what I would cut, what is a rebuild — send the domain or book a slot. You will get causes, not a traffic-light graphic.
What I actually change
When I take a speed brief, the work is unglamorous.
Images get real dimensions, modern formats, and a size that matches the slot. Fonts get subset and a fallback that does not shove the heading. Third parties get a trial: if I cannot justify the script against a tap on a phone, it goes. The homepage loses sections that do not earn their bytes. The HTML already contains the words a crawler and a human need.
The stack I default to is custom code — typically Next.js — because I can decide what ships. That is in services. I will also tell you when a template is enough and the honest fix is compression and fewer widgets, not a rewrite. Custom that reimplements a heavy theme is theatre.
I will not promise a lab score. I will not promise page one. I will promise the page is built so a person on 4G can read the offer and act, and that the field vitals have a chance to be good because the document is small and stable.
A keyboard user and a slow phone are both failed by the same pile of scripts. Semantic HTML and a contact path that works with JavaScript off are speed work as much as SEO work.
What I will not do
I will not screenshot a Lighthouse run on a throttled desktop and call it evidence. I will not sell a monthly “performance retainer” that is a plugin licence. I will not hide a chat widget in the quote and then charge you to remove it.
If the business needs video, we give it a URL and a poster — not autoplay on the landing page a van just opened. Reviews do not need five platforms in the head.
East Midlands, here, is not a keyword. It is the network conditions your customers actually have between Nottingham, Derby, Leicester, and the towns in between. Build for that and the lab usually follows. Build for the lab and the van still leaves.
Bring the current URL. Tell me which page you need to be fast — homepage, a service, a booking flow. I will tell you what is actually costing you.