What makes a website truly fast (hint: it’s not your Lighthouse score)

Everyone wants a 90+ Lighthouse score. It looks good on a website, it’s easy to screenshot for a slide deck, and it feels like progress. But here’s the uncomfortable truth: those numbers don’t always reflect how fast your website actually feels to customers.

In 2025, speed is about experience, not synthetic test results.

The illusion of speed

Lighthouse, PageSpeed Insights, and similar tools are great for catching technical issues. They measure things like First Contentful Paint, Largest Contentful Paint, and Time to Interactive. Useful, yes. But they’re synthetic tests that run in controlled conditions with a predefined network speed, a clean browser profile, and no competing background scripts or traffic.

Meanwhile, your customers are browsing on shaky Wi-Fi, with dozens of tracking scripts, product recommendations, and personalization engines running at once. They don’t care if your Lighthouse score is 98; they care whether adding an item to their cart just works.

A page that scores high but lags during checkout still feels slow. And feelings decide conversions.

What real performance looks like

So, what should we measure instead?

Real user monitoring (RUM) gives a better picture. It tells you what actual visitors experience across devices, browsers, and regions. Beyond that, the true indicators of speed live in the moments that matter most: when a customer tries to do something.

Here are a few of those metrics:

• Add-to-cart time: from click to confirmation
• API latency: especially for cart, pricing, and stock endpoints
• Checkout responsiveness: how long it takes to load delivery and payment options
• Server-to-server round trips: how many calls a single action triggers
• Stability under load: does performance hold up during campaigns or sales?
  

Real performance also needs to scale, both with traffic and with catalog size. A fast site with 100 products should stay just as fast with 100,000. Response times shouldn’t grow as the database expands, and latency should remain stable even as more customers browse, search, and buy at the same time.

These are the kinds of numbers that directly correlate with conversion rate, not the ones in a green circle on Google.

Why Lighthouse doesn’t capture modern commerce

Today’s composable commerce setups are far more complex than the old all-in-one monoliths. Your frontend (maybe built in Next.js or Remix) talks to multiple backends: cart, pricing, shipping, CMS, promotions, personalization, and more. Lighthouse primarily measures initial page render and doesn’t reflect multi-step, API-driven flows—the parts of the journey where performance often breaks down.

You can have a static product page that loads in 400 ms, but if adding that product to the cart takes 1.8 seconds because of a slow API call or inefficient orchestration between services, that customer still experiences a slow site.

In other words: fast rendering does not equal fast experience.

Modern performance is a system property. It depends on how efficiently your services talk to each other, how globally distributed your infrastructure is, and how well your edge handles dynamic data. That’s something no single Lighthouse score can tell you.

When caching hides the truth

Caching is often used to make a site appear fast. Product data, prices, and stock levels are preloaded so the next customer gets an instant response. On paper, that looks great. In reality, it can hide inefficiencies or mask slow technology underneath.

A cached product page might open instantly, but what happens when a user tries to buy the last item in stock and it’s already gone? Or when the cache serves a price that’s no longer valid? Speed that depends on old data isn’t real speed—it’s borrowed time.

At Brink, we believe caching has its place, but it should never be used to paper over latency or outdated system design. True performance comes from a responsive architecture that delivers live, accurate data at scale. That’s the difference between “feels fast” and is fast.

Time to add-to-cart: a better benchmark

If we’re serious about understanding real speed, we need a metric that reflects what actually happens when a customer interacts with our site. Something that captures not just how fast a page paints, but how quickly a click becomes an action.

There isn’t an official term for this yet, but we all love an acronym, so let’s call it TTAC: Time to Add-to-Cart.

TTAC would measure how long it takes from the moment a customer clicks “Add to cart” until the cart confirms that the item is there. It’s simple to explain, easy to test, and brutally revealing.

Because TTAC includes:

• Frontend responsiveness
  
• Network latency
• API performance
• Database writes
• Orchestration between services
  

In other words, it reflects the true system speed. If your Lighthouse score is 95 but TTAC is 1.6 seconds, you don’t have a fast site—you have a fast illusion.

Redefining “fast” in 2025

A modern definition of fast looks like this:

• Responsive: every tap and click gives instant feedback.
• Predictable: latency stays stable even under heavy load.
• Global: customers in Tokyo and Berlin experience consistently quick interactions.
• Resilient: one slow microservice doesn’t drag down the whole system.

This is what real performance looks like—the kind that builds trust and drives sales.

How we think about it at Brink Commerce

At Brink Commerce, we look beyond vanity metrics. Our distributed, serverless architecture across AWS regions keeps API responses around 100 ms globally, meaning every add-to-cart or checkout interaction stays consistently fast across regions.

That said, we don’t hate seeing great Lighthouse scores, especially when our agency partners and customer in-house teams consistently hit 90+. We’re a little vain too. We just know that what makes those scores possible—and sustainable at scale—often happens under the hood.

Because in the end, performance isn’t about pleasing Google. It’s about giving your customers a seamless, friction-free experience that feels fast everywhere, not just in a lab.