TTFB Explained: What Time-to-First-Byte Measures and Why Single-Number Comparisons Mislead
Time to First Byte is a composite metric — DNS, TCP, TLS, server processing, return path. The single TTFB numbers in hosting reviews are not reproducible. Here is what TTFB actually measures and how to compare hosts honestly.
What TTFB actually measures
Time to First Byte is the elapsed time from the browser issuing a request to the first byte of the response arriving back. It is a composite of five things: DNS resolution, TCP handshake, TLS negotiation, server processing, and network return. On a static page, the network components dominate. On a database-backed page (WordPress with active plugins, a Rails app, a Django site), server processing dominates. The implication: the same host can post very different TTFB numbers on different sites.
Why single-number TTFB claims are not reproducible
Hosting reviews routinely publish a single TTFB figure for each host (a millisecond number on the managed-WordPress winner, a much higher number on a budget shared host). Those numbers are not reproducible because TTFB depends on the test site's code, plugin set, database size, cache state, the test client's geographic location, and server load at the moment of the test. We do not publish single TTFB figures because any single number we cited would not be reproducible by a reader running their own measurement.
How the major hosts describe their performance promise
Vendors talk about performance via infrastructure (Kinsta lists Google Cloud Premium Tier across 37 datacenters at kinsta.com/features; SiteGround lists Google Cloud at siteground.com/google-cloud-hosting.htm), via SLA (DreamHost publishes a 100% uptime guarantee at dreamhost.com/legal/100-uptime-guarantee/, Kinsta publishes a 99.9% SLA at kinsta.com/legal/sla/), and via cache architecture. None of these is a TTFB guarantee, and that is the honest framing.
How to actually compare hosts
Run your own measurement on the workload you actually care about. Use a tool like k6, Apache Bench, or hey from a region close to your audience; hit a representative URL on each host (during a free trial or during the refund window of a paid plan); record the p50 and p95 over a sustained run, not a single hit. Compare those numbers, not the marketing copy. For a one-off check, a tool like KeyCDN's performance test or webpagetest.org gives a directionally honest signal.
What price tells you that TTFB does not
TTFB is hard to compare. Published price is not. The intro-vs-renewal gap is the single most honest hosting comparison signal: Kinsta does not have one (intro $35 = renewal $35 on the Single tier); Hetzner does not have one (€3.79 today and €3.79 next year on CX22); SiteGround, Bluehost, and DreamHost shared tiers all show steep intro-vs-renewal divergence (Bluehost Basic intro $2.95 → renewal $11.99, SiteGround StartUp intro $2.99 → renewal $17.99, DreamHost Shared Starter intro $2.59 → renewal $10.99). Whether you pay the intro or the renewal is what you should care about, not a non-reproducible TTFB number.
Sources
Kinsta SLA: kinsta.com/legal/sla. Kinsta features: kinsta.com/features. SiteGround Google Cloud: siteground.com/google-cloud-hosting.htm. DreamHost 100% uptime: dreamhost.com/legal/100-uptime-guarantee. Vendor pricing pages: kinsta.com/plans, hetzner.com/cloud, siteground.com/web-hosting.htm, bluehost.com/pricing, dreamhost.com/hosting. All URLs accessed 2026-04-30.