Someone on your team runs a speed test, gets a green score, and the conversation ends. Meanwhile your sales team keeps hearing that the site felt sluggish, and your form fills keep drifting down. Both things are true. The test measured a simulation on a fast connection. Your buyer measured a mid-range phone on hotel Wi-Fi with twelve tracking scripts loading in the background.
Site speed is rarely a developer problem. It's an accumulation problem — the result of two years of small, reasonable additions nobody ever subtracted. Here's how to find out what your site actually costs you, and what to do about it.
Lab data is a rehearsal. Field data is the show.
Tools like PageSpeed Insights give you two kinds of numbers, and most people only read the first one. The lab score is a simulated load under controlled conditions. It's useful for debugging, useless as a verdict.
The number that matters is field data — real measurements from real visitors on real devices. Google collects this through the Chrome User Experience Report, and you can see it at the top of a PageSpeed Insights result when your site has enough traffic. If your lab score is green and your field data is yellow, your field data is right. That gap usually means your visitors are on slower devices and networks than the person running the test, or that something loads fine once it's cached and painfully the first time.
If your site doesn't get enough traffic to show field data, use real-user monitoring instead. Guessing from a laptop on office fiber is not measurement.
Loading isn't the whole story anymore
Core Web Vitals sound technical but translate cleanly into things a buyer feels:
- LCP (Largest Contentful Paint) — how long before the main thing on the page shows up. Google's good threshold is 2.5 seconds.
- INP (Interaction to Next Paint) — how long the page takes to respond when someone taps or clicks. Good is under 200 milliseconds.
- CLS (Cumulative Layout Shift) — whether content jumps around while loading and makes people mis-tap.
INP is the one most teams haven't caught up on. It replaced the older First Input Delay metric in 2024, and it's stricter in a way that matters: it measures the full delay from tap to visible response, not just the initial lag. A page can paint quickly and still feel broken because a script is hogging the main thread when someone hits your mobile menu or starts typing in a form field. That's the exact moment a lead decides whether to keep going.
Most slowness is bought, not built
Run a script census before you touch a line of code. Open your tag manager and list everything firing on your site. In most mid-market setups you'll find a live chat widget, two or three analytics tools, a heat-mapping tool someone trialed, retargeting pixels from an ad platform you stopped using, a review widget, a font service, and an A/B testing tool that's been running zero tests for a year.
Each of those was approved by a different person for a good reason. Nobody owns removing them. Together they're often the single largest contributor to a slow, janky experience — and unlike a code rewrite, deleting them takes an afternoon.
The rule we'd apply: every third-party script needs a named owner and a question it answers. If nobody can name both, turn it off and see who complains. Usually nobody does.
Fix templates, not pages
The second trap is optimizing the wrong things. Teams tune the homepage because it's the page leadership looks at most, while the pages that actually convert — service pages, the pricing page, the one landing page your ads point at — run on a template nobody has touched.
Work from your analytics. Find the handful of pages that produce qualified inquiries, check their field data specifically, and fix the template those pages share. Oversized hero images, render-blocking fonts, and an unoptimized form embed are template-level problems. Fixing one template often fixes forty pages.
When rebuilding is the honest answer
Sometimes improvement runs out of road. If your site is built on a page builder that generates bloated markup, if every plugin update breaks layout, or if your "quick fix" list has sat untouched for a year because nobody wants to touch the codebase — you don't have a performance problem, you have a platform problem. Patching it is more expensive than replacing it, just in smaller invoices.
That's a real decision with real tradeoffs, and it should be made with numbers: what the current platform costs in maintenance time, what pages produce revenue, and what you'd need to preserve in a migration.
Make it a business metric
Stop reporting the score. Report the thing the score is a proxy for: form submissions, qualified leads, bounce rate on your money pages, mobile conversion versus desktop. Speed work that doesn't move one of those wasn't worth doing.
If you're not sure whether your site needs a tune-up or a rebuild, that's a conversation worth having before you commit a budget to either. Infraxio audits sites the way operators look at any other system — what it produces, what it costs to keep running, and where the leverage actually is. Get in touch and we'll take a look at yours.
