Table XI

MIT’s Polaris can cut page load time by 34 percent — what you need to know

All that gorgeous photography and stylized copy that makes your product sell so well — it’s also what’s taking precious seconds to load when you call up your site on a smartphone. You need rich media to properly showcase what you’re selling, but you also need your site to be speedy enough to load quickly on small screens and conform to web standards. Amazon estimates that every extra 100 milliseconds it takes a page to load cuts its profits by 1 percent. It’s not just that users won’t wait around for 10, 20 seconds for your site to load — Google uses site speed in its search rankings. Slow sites will take an SEO hit, hurting the odds of your customers finding you organically — i.e. without paid advertising.

Polaris is the latest in a long list of technologies designed to keep your pages loading as fast as possible. The new server-based framework was created by researchers at MIT’s Computer Science and Artificial Intelligence Lab and Harvard. Polaris works by mapping the dependencies between the objects on your site — think Javascript and HTML files, images, and so on. Normally a browser loads the objects on your webpage piece-by-piece, making return trips to the server when it finds that one object is dependent on another — i.e. when executing a Javascript file reveals that additional images need to be fetched and rendered, in MIT’s example. Each trip to fetch additional objects increases the load time. By identifying the dependencies in advance, Polaris is able to keep your browser from making multiple calls to the server, speeding up your page.

MIT tested Polaris on 200 of the world’s most-visited websites, and reduced load time by an average of 34 percent. The framework does the most good for complicated sites like or — where an abundance of content creates multiple dependencies.

Polaris differs from other efforts to speed up sites. It doesn’t rely on data compression to reduce the size of information transferred, it just seeks to reduce the number of trips it takes to transfer it. It can work on any browser, meaning it could potentially give all users a faster page load-time. And unlike Google’s AMP project and Facebook’s Instant Articles, it doesn’t require changes to the content. For it to live up to its potential as a universal solution, though, two things need to happen:

This week, MIT PhD student Ravi Netravali, the first to author a paper on Polaris, will present his findings. If browsers are intrigued and start adopting the framework, having Polaris running to map your site dependences could quickly become a competitive advantage.
At Table XI, we assess every opportunity to help our partners grow their businesses. We’ll be keeping tabs on Polaris, and if and when it makes sense, we’ll incorporate it on our clients’ sites. If you want to learn more about Polaris, or experiment with running it on your site, email me at mike.hostetler [at]