TTFB versus Lighthouse: measure the real server

Lighthouse is useful. Fixed network, fixed CPU, pretty score. TTFB is something else: how fast the first byte hits the wire. A server-side crawler feels that. Ours does. Googlebot does when it fetches your HTML. Your laptop on office wifi lies in your favour.

What I see most: cold PHP, one heavy homepage query, no cache header on HTML that could be cached. Or three redirects before any body. Hosting in one region, visitors in another, no edge. An object cache that empties after a deploy and bills the first visitor.

Under 200 ms on HTML is a decent target. Then Gzip or Brotli, HTTP/2, page weight. Those are separate problems. A green lab score with a slow first byte is not a win. It is a poster. Field data (CrUX) can look green while a crawler from another country waits 800 ms. Measure both.

If TTFB doubles after a plugin, you want that the same day. Not via a review about a “slow site”. Log the first byte on a few URLs, not only home. Checkout and a deep blog post lie less than your hero.

A lab run from your laptop measures your office. A crawler measures the server. Those two numbers may differ. If they sit far apart, trust the crawler. It is closer to what Googlebot feels when it shows up at 4 a.m., without your warm cache.