Making Facebook 2x Faster
Everyone knows the internet is better when it's fast. At Facebook, we strive to make our site as responsive as possible; we've run experiments that prove users view more pages and get more value out of the site when it runs faster.
Context & Ripple Effects
Facebook is publishing its own engineering case for speed: internal experiments convinced the team that a faster site means users view more pages and extract more value per visit, so the company has rebuilt front-end delivery to run the site twice as fast. Coming from a platform at Facebook's scale, that framing turns page-load time from a back-end chore into a stated product priority.
The timing lands amid measurable momentum: Nielsen put average U.S. Facebook usage up nearly 10% in January alone, past seven hours per month, while Compete told the San Francisco Chronicle that Facebook already routes 13 percent of U.S. traffic to big portals including Yahoo, MSN and AOL. A faster site compounds both numbers — more sessions per user and more outbound referrals flowing through Facebook's front door.
First-order effects
- Users get roughly half the wait on every page, and Facebook gets more pages viewed per visit — the company's own experiment data says speed directly lifts consumption, stacking onto the January usage growth Nielsen recorded.
- Front-end performance becomes an explicit engineering commitment, with Facebook staking credibility on the 2x figure it just published.
Second-order effects
- Faster page turns deepen Facebook's role as a traffic gateway: with Compete already crediting it for 13 percent of U.S. referrals to Yahoo, MSN and AOL, every additional pageview widens the gap between Facebook's distribution power and the portals dependent on that flow.
- Rival high-traffic consumer sites now face speed as a visible competitive benchmark — Facebook has published its number, inviting comparison against any slower competitor's experience.
Third-order effects
- If the speed-drives-engagement pattern holds at scale, latency stops being an infrastructure metric and becomes an engagement metric — large platforms will run the same experiments Facebook ran and treat load time as a lever on session depth and ad inventory.
- Publishing performance results as engineering content, as Facebook does here, points toward performance becoming part of how major web companies compete for developer and user trust, not just something measured internally.
The trend: Consumer web platforms are converting load-time engineering into measurable engagement gains, turning site velocity into a published competitive metric.