Website hosting for contractors: I checked where 266 West Michigan trade sites live, and 30 WordPress sites were slow

Hosting is the part of a website nobody sees until it breaks. It is the computer that holds your pages and sends them to a customer’s phone. Usually it is one of two arrangements: a builder rents you the whole thing, or you, or whoever built the site, rent a server and put WordPress or a custom site on it. Which one the local competition uses, and with whom, is something I could read from public information, so I did.

How I measured it

The sites are the 266 trade homepages from my 23 September 2026 platform reading: the roofers, plumbers, heating and cooling companies, electricians, landscapers, painters and remodelers Brave Search lists in West Michigan towns, one row per website address. Dentists are left out, because they have their own page.

Between 3:07 and 3:13 a.m. Eastern on 10 October 2026 I did four things for each site. I asked public DNS which address the site’s name points to. I asked Team Cymru’s address lookup, a public service, which company’s network owns that address. I fetched the homepage once with curl, the way a browser would, and kept the response headers. And I timed how long the first byte took to arrive. That was one request per site with no retries, apart from the slow ones, which I come back to.

A network is not always the host, so each WordPress and custom site got its name from the first rule that fit:

1. A header with a host’s name in it. WP Engine’s servers answer with a header that says so; Flywheel, Pantheon and Kinsta each leave their own mark, and so do Automattic’s servers, the company that runs WordPress.com.
2. The network that answers, when it belongs to a hosting company such as Liquid Web, Hostinger or A2 Hosting.
3. Neither, because the address belongs to a big cloud or to Cloudflare. I say that and do not guess. Cloudflare usually sits in front of a host rather than being the host, and it hides which one.

The split: 94 on builders, 130 on WordPress

How the site is hosted (10 Oct 2026)	Sites	Of 266
Hosted builder or agency platform, hosting is in the plan	94	35%
WordPress, on a host someone chose and pays for	130	49%
Custom build, other system, franchise site or not identified	42	16%

It is the same split as the platform page, because the platform settles the hosting question. When a builder runs the site, the builder is the host.

Hosted builders: the hosting is in the price

For these 94 sites there was no separate hosting decision and there is no separate hosting bill. The builder’s plan is the hosting. That is easy to live with until you want to leave: a site that lives inside a builder or an agency’s system mostly cannot be moved to another host, so leaving means rebuilding it.

Builder or platform	Sites	Of 94
Wix	30	32%
Duda	16	17%
Squarespace	11	12%
GoDaddy Website Builder	10	11%
Hibu (agency platform)	7	7%
Webflow	4	4%
8 other builders and agency platforms	16	17%

If your site is one of these, the question to settle is what leaving would cost and who holds the login, not which server it runs on. If your web designer goes out of business lists the seven things to hold before you need them.

WordPress: 130 sites, 19 named hosts, and 48 answering from Cloudflare

WordPress is the one that needs a host. I could name the host of 82 of the 130 (63%).

Host of the WordPress site	Sites	Of 130
WP Engine	23	18%
Automattic’s network (the company behind WordPress.com)	9	7%
Hostinger	7	5%
Liquid Web	6	5%
Flywheel	6	5%
Unified Layer / Newfold Digital network	5	4%
GoDaddy	5	4%
A2 Hosting	5	4%
Namecheap	3	2%
Pantheon	2	2%
Aptum	2	2%
Kinsta	2	2%
7 other hosting companies, 1 site each	7	5%
Cloud server, host not named (Google Cloud 10, Amazon Web Services 6, DigitalOcean 4, Vultr 1, Microsoft Azure 1, Hetzner 1)	23	18%
Behind Cloudflare, host not visible	25	19%
All WordPress sites	130	100%

Two things stand out. First, there is no default. WP Engine is the largest name with 23, and then the list is long: 11 more companies with two sites or more and 7 with one. Second, Cloudflare hides a lot. 48 of the 130 WordPress sites answer from Cloudflare’s addresses. For 23 of them a header still names the real host (20 WP Engine, 2 Kinsta, 1 Unified Layer / Newfold), which is why a plain address lookup would have counted WP Engine’s sites as Cloudflare’s. For the other 25 I can see nothing, and the table says so.

94 sites had no hosting choice to make. The 130 that did are spread across at least 19 named companies.

How fast they answer

The number I timed is the first byte: how long from the request until the first piece of the page arrives, counting any redirect, the connection and the security handshake. web.dev, Google’s web guidance site, calls 0.8 seconds or less good and anything over 1.8 seconds poor (web.dev/articles/ttfb, read 10 October 2026). It is not how long the whole page takes to appear; it is the part that the host controls most.

A single reply can be slow because nobody had asked for that page in a while. So every site that took over 0.8 seconds on the first fetch, 69 of 262, I fetched three more times, 10 seconds apart. For 36 of them the median of the four fetches was 0.8 seconds or less: that is the cold-cache effect, where the first visitor waits and the next ones do not. The other 33 stayed slow. Here is the count in each group.

Group (10 Oct 2026)	Answered	First fetch over 0.8 s	Median of four over 0.8 s	Median of four over 1.8 s
Hosted builders and agency platforms	94	9	1	0
WordPress, host named by a header (WP Engine, Flywheel, Pantheon, Kinsta, Automattic)	42	13	1	0
WordPress, any other named hosting company	40	22	17	5
WordPress, on a cloud server with no host named	22	10	5	0
WordPress, behind Cloudflare with no host visible	24	7	7	3
Custom builds, other systems, unidentified	40	8	2	0
All sites that answered	262	69	33	8

The middle is the same everywhere. The median first byte was 0.50 seconds for hosted builders, 0.46 for WordPress and 0.49 for the rest, so a typical site is fine whatever it runs on. The difference is in the tail. Hosted builders had 1 slow site in 94, and the slowest took 1.6 seconds on the first fetch. WordPress had 30 in 128, and the slowest took 6.3 seconds. The quickest site in the reading, at 0.14 seconds, was WordPress as well. WordPress is where a good host and a poor one show up side by side.

Within WordPress, the sites whose response names WP Engine, Flywheel, Pantheon, Kinsta or Automattic were almost never slow (1 of 42), and those on the other named hosting companies often were (17 of 40). I am not going to rank hosting companies from groups of two to seven sites, and I do not know which plan any of these owners pays for: a cheap plan and a dear plan at the same company can behave very differently. The pattern is a reason to test, not a verdict on any brand.

One more reason not to wave the first-fetch figure away. A small local site is exactly the kind that has gone quiet when a customer arrives, so the first fetch is what that customer gets. 69 of 262 sites (26%) were over 0.8 seconds on it. For the speed of the whole page, which this reading does not measure, see whether speed matters for a local business site.

A Michigan footnote

9 of the 266 sites answer from Liquid Web’s network (6 of them WordPress). Liquid Web’s own page says Lansing, Michigan is its founding location and that it runs two data centers there (liquidweb.com/data-centers/lansing-web-hosting, read 10 October 2026). That does not put those 9 sites in Lansing: an address does not tell me which of a company’s data centers a site is in. It is only a note that one of the names in the list has a Michigan home.

What to ask about your own hosting

Question	Why it matters	What to insist on
Which company hosts the site, and whose name is the account in?	If it is your old web designer’s account, you cannot renew, move or fix anything without them.	An account in your name, or your business’s, with you as the owner.
Who can log in today?	The one person who knows the password is a single point of failure.	You and whoever maintains it, each with their own login.
How long does the first byte take?	In this reading 30 of 128 WordPress sites stayed over the 0.8 seconds web.dev calls good.	Time your homepage a few times at different hours: in a browser’s developer tools, Network, then Timing, “Waiting for server response”.
Is there a backup, and has anyone restored one?	A backup nobody has tried is a hope.	The date of the last backup, and a copy you can download.
When does it renew, and at what price?	A card that expires takes a site offline without anyone deciding to.	The renewal date in your calendar and a card you control.
What would moving it cost?	Builders mostly cannot move; WordPress usually can.	A written answer before you sign anything.

For what hosting costs once you add it up, see what a plumber’s website costs each month. For who should hold the keys, see who owns your website and domain.

What this reading does not prove

• The network is not the invoice. A hosting company can resell another’s servers, and Cloudflare can sit in front of anything. The names are who answers, not always who you would pay.
• One place, one night. The timings are from one Mac in Grand Rapids between 3:07 and 3:13 a.m. Eastern. At a busy hour, or from another city, they would differ.
• First byte, not the whole page. A fast first byte can still lead to a heavy page. This is the part the host controls most, not all of it.
• Small groups. The largest named host has 23 sites and most have under seven, so I do not rank companies.
• Plans differ. I cannot see whether a site is on a company’s cheapest plan or its dearest.
• Correlation, not cause. An owner on one of the five hosts named in that first group may also have had a better site built.
• Four sites did not answer. One returned 401, one 403, one timed out, and one served a certificate chain curl could not verify (browsers sometimes repair those, so I flag it rather than call it broken). The other 262 answered over HTTPS with a certificate that verified.

---

Common questions