WordPress host TTFB benchmark — 11 hosts, 30 days of data (April 2026)

How the test runs

A bash cron fires curl five times against each host's public WordPress page, takes the median time_starttransfer, and writes a JSONL row. Two origins are running the same script: a Hostinger VPS in São Paulo and a home server in Ontario. Both sets get rolled into one 7-day median + p90 CSV that the dashboard reads.

Caveat I want stated upfront: marketing-page TTFB is a proxy for customer-site TTFB, not the same thing. Hosts optimise their own marketing pages aggressively. Treat these numbers as the floor of what each host can do, not what your site will do under load.

The 30-day numbers

Sorted fastest to slowest, median across all samples in the window. Numbers in milliseconds.

Live version refreshes every 24h at /2026-wp-hosting-ttfb-benchmark. CSV is at the same URL with a .csv suffix.

Rocket.net is not a fair fight

Rocket.net runs Cloudflare Enterprise on every plan. That single line is doing most of the work in the 86 ms median. Their origin servers are competent, but the reason they sit at the top of the chart is the edge cache. The other premium hosts will sell you Cloudflare Enterprise as an add-on. Rocket includes it by default.

For a $30/mo entry plan that's a stupidly good deal. The catch is what happens when you outgrow the plan, scaling cost climbs faster than at WPE or Kinsta. I'm not at that threshold yet on any of the 5 stores I moved over in Q1, so I can't give you the real answer on how it pans out at 100K+ visits. Ask me in six months.

Pressable was the surprise

I'd written off Pressable as the boring Automattic-owned cousin of WP Engine. The 163 ms median put it ahead of WPE in this window, and it's been stable, not a one-week fluke. The p90 is genuinely tight too.

I have not migrated any client to Pressable. I'm probably going to test one in May. If the consistency holds at the application layer (TTFB on a real WP-Admin login, not a static marketing page), it could be a serious WPE alternative for the agency plans where I currently spend $290/mo.

SiteGround keeps sliding

301 ms median is not bad in absolute terms. It is bad relative to where SiteGround was in 2023, when the same kind of test would have put them in the 180-220 ms band. I migrated 22 client sites off SiteGround in Q1 for pricing reasons ( writeup here ), and the performance trajectory adds another reason to be cautious.

I'm not declaring SiteGround dead. Their support is still genuinely good and they haven't fallen off the chart entirely. But the gap between SiteGround and the Cloudways-DO tier has narrowed to 77 ms, and Cloudways is half the renewal price. That math is hard to defend on an agency book.

Bluehost and HostGator: still rough

Both still in the 400-700 ms band on a marketing page. That's before any plugin bloat, any unoptimised theme, any of the things real client sites pile on top. Customer sites on this tier of Newfold-portfolio hosting routinely show 1.5-3 second TTFB in the field. I have not used either for client work in years and this benchmark does not change my mind.

GoDaddy is better than I expected

374 ms median for GoDaddy Managed WP is a real improvement on where they were five years ago. I'd still pass on it because of the reseller economics, but if a client showed up already on GoDaddy and wasn't bleeding revenue, I'd no longer feel the need to rip them off it on principle.

What I'm actually going to do with this data

Three concrete shifts in how I quote new agency clients.

New WooCommerce builds default to Rocket.net unless the client has a specific reason to be elsewhere. The 86 ms TTFB compounds across every page load on a checkout flow. I'm not going to argue against that on price.

Pressable enters my recommended set in May, pending the real-app-layer test. If it holds up, agency clients with content-heavy sites get Pressable as a cheaper-than-WPE alternative.

I'm pulling SiteGround off my "good enough" list for new builds. Existing clients who still want to renew can keep their accounts. New ones don't go there. The trajectory is the wrong direction.

Why I bothered building this

Review Signal used to publish a credible WordPress hosting benchmark. They stopped in 2020. Hostingstep does an annual roundup but they don't publish raw data, you take their word for the numbers. Every other "best WordPress host" article you'll find online is built on affiliate-payout order, not measurement.

I'm running this because I needed it for my own client recommendations and nobody was publishing it. The CSV is open under CC-BY 4.0. If you want to fork the script, the cron config and bash source are linked from the dashboard. If you have a host I haven't covered and want it added, the contact form on the about page works.

Honest limitations of this benchmark

I want to flag the things this dataset does not show.

It's marketing-page TTFB, not customer-site TTFB. As I said up top, hosts optimise their own pages. Your WordPress install, with whatever combination of plugins and themes you run, will be slower.

It's two origins. São Paulo and Ontario are not the world. A reader in Singapore will see different numbers, especially for hosts with weaker CDN coverage. I'd love to add an Asia-Pacific origin, that's on the to-do list.

It's TTFB only. It does not measure cache miss behaviour, sustained throughput under load, database query performance, or any of the things that actually break a site at 4pm on Black Friday. Those need different tests, and I have not run them.

Read the dashboard alongside the TCO calculator and your own load profile. No single metric picks a host for you.