Website speed optimization services, compared on evidence

Agencies with a verified published price

How to commission website speed work

1. Measure the field data before you take a quote. Synthetic lab tests are useful for diagnosis and useless as a contract term, because they measure one run on one machine. Real user data across your actual visitors is what Google reads and what your customers experience. Establish that baseline yourself first, so you are not negotiating against a number the vendor chose.
2. Agree the pass mark in the three published metrics. Google documents the good thresholds on web.dev: Largest Contentful Paint at or under 2.5 seconds, Interaction to Next Paint at or under 200 milliseconds, and Cumulative Layout Shift at or under 0.1, assessed at the 75th percentile of page visits. Write those into the engagement with the pages and the device split they apply to.
3. Establish what is in scope and what is a rebuild. Compression, caching, a content delivery network, image formats, font loading and script deferral are tuning. Removing a page builder, replacing a theme, cutting third party tags or changing rendering strategy is engineering with real risk. Get the boundary in writing, because most disputes here are a tuning fee meeting a rebuild problem.
4. Fix regression, because speed decays. A site tuned once drifts back as marketing adds tags, editors upload large images and plugins update. Agree who monitors field data after the engagement, what the alert threshold is, and whether a rule exists that new third party scripts get reviewed before they ship. Otherwise you are buying the same work again next year.

What providers publish, and how to read it

Our index records what each provider states publicly: pricing or a starting fee where published, minimum engagement, named clients and the services claimed as in house. Performance specialists publish pricing more often than general agencies, because a fixed audit or a tuning package is a repeatable product. That makes the market comparatively easy to compare, and makes an unexplained absence of scope detail worth asking about.

Be careful with headline claims built on a single lab score. A number from one synthetic run on a fast connection is not evidence about your visitors, and it is trivially improved without helping anyone. Ask instead for a before and after in field data on a site the provider worked on, covering mobile as well as desktop, and ask how long the improvement held.

What actually makes sites slow

The causes repeat across almost every engagement. Oversized uncompressed images in modern layouts. Third party scripts for chat, heatmaps, testing, consent and multiple analytics tools, each adding main thread work. Render blocking stylesheets and fonts. A page builder that ships far more markup than the page needs. Slow server response, often shared hosting under load or an uncached database query. Layout that shifts as late loading elements arrive.

That list is worth knowing because it tells you what a competent quote looks like. A provider who inspects your site and names which of these apply, in what order, with an estimated effect on each metric, is doing the job. One that quotes a package before looking is selling a process and hoping your site fits it.

What it is worth, and when not to buy it

Speed work pays for itself in two ways: visitors who would otherwise abandon, and eligibility in a ranking system that reads page experience among many signals. It is worth commissioning when your field data misses the published thresholds on the pages that make you money, especially on mobile. It is not the highest value purchase when your pages already pass, or when the site is not being found at all, because a fast page nobody reaches earns nothing.

Sequence matters. If a rebuild is planned within a year, tuning the current site is often money spent twice, and the better purchase is a performance budget imposed on the new build. Ask any provider directly whether your money would be better spent elsewhere first. The honest ones say so, and that answer is a reliable signal about everything else they tell you.

Questions people actually ask

Does site speed affect rankings?

It is one signal among many rather than a lever, and Google's own documentation frames page experience as something it considers alongside relevance and quality. A slow page about exactly what someone wants can still outrank a fast irrelevant one. Buy speed for conversion and for the visitors who leave before the page paints, and treat any search benefit as secondary.

Why does my score differ between tools and runs?

Lab tools simulate one visit on one device and connection, so results move between runs by design. Field data aggregates real visits over a rolling window, which is why Google assesses the thresholds at the 75th percentile of visits. Use lab tools to find causes and field data to judge whether anything actually improved.

Can I just add a caching plugin and a CDN?

That fixes server response and delivery, which is real but usually not the binding constraint. Most failing sites are slow because of what the page loads in the browser: images, fonts, scripts and layout that shifts. Caching will not remove a chat widget or a heavy page builder, so measure first and treat the plugin as one tool rather than the answer.

How long does the work take?

Tuning an existing site is commonly days of work with results visible in field data over the following weeks, because the rolling window has to refill. Structural work such as replacing a theme or removing a page builder is a project measured in weeks and carries real regression risk. Agree which one you are buying before you agree a price.

Browse agencies by specialty

• SEO agencies with verified facts
• PPC agencies with verified facts
• Social media agencies with verified facts
• Content marketing agencies with verified facts
• Email marketing agencies with verified facts
• Performance marketing agencies with verified facts