Researchers have demonstrated a way to speed up page loads by overlapping the downloads of a page’s objects rather than fetching them conservatively and one-by-one. In tests on a couple of hundred popular sites, the system — dubbed Polaris — reduced overall download time by roughly a third.
Why pages slow down
Modern web pages are built from many pieces: HTML, JavaScript, stylesheets, images, fonts and more. A browser fetches those pieces over the network and then evaluates them before the page can be fully rendered. The issue is not just the size of those files, but how many separate network round-trips the browser needs to make. On high-latency links, especially mobile networks, each additional round-trip can add tens or hundreds of milliseconds, so the delays stack up.
Browsers play it safe because they often don’t know how the various objects on a page depend on each other until they’ve been fetched and evaluated. That conservative ordering avoids breaking the page, but it also forces unnecessary waiting: the browser may pause downloading some resources until others are finished, even when those resources could have been fetched in parallel.
What this approach does
The research team’s system automatically tracks how objects on a page interact — which ones read or update data from others — and builds a dependency graph for the page. With that graph in hand, the system can schedule downloads in a smarter order and overlap requests where dependencies allow, cutting the number of required round-trips.
Think of it like a business trip: if you discover new cities to visit only after you reach each one, you’ll end up zig-zagging. If you had the full itinerary before you left, you could plan the most efficient route. The dependency graph gives the browser that itinerary, letting it fetch what it can in parallel and avoid needless back-and-forth.
The research showed the biggest wins on larger, more complex sites and on high-latency networks — the places where round-trip delays matter most.
Practical takeaways for webmasters and users
The research itself is an example of how thinking about dependencies, not just resource size, can yield big improvements. Even if you don’t deploy the exact research system, the underlying ideas point to practical optimisations you can apply today:
- Reduce round-trips: combine or inline critical resources where sensible so the browser needs fewer separate requests for the first render. - Make dependencies explicit: use resource hints (preload, preconnect) and sensible ordering so browsers can start fetching what’s needed early. - Prioritise critical assets: ensure the files required for the initial view are fast and small; defer non-essential scripts and assets. - Use modern transport features: HTTP/2 and newer protocols reduce the cost of many small requests by allowing multiplexing, though dependency-aware scheduling can still add value.
The research demonstrates that understanding how a page’s parts interact — and using that understanding to overlap what can be fetched in parallel — is a powerful way to cut perceived load time. For sites that feel sluggish, especially on mobile, reviewing resource dependencies is a useful place to start.