Web Design Research · By Anthony Limpert · November 8, 2026

The Best, the Good, the Bad and the Ugly of 95 Sites Web Designers Shared

Updated November 9, 2026 with every site people shared, 95 in all, 42 rebuilds and what they cost us, plus what we believe makes the best website.

On November 7 I put one question on my Facebook. What makes the best website design?

I asked designers to tag their company and main website. I said every site would get rebuilt. Then I did what we do with everything, which is measure it.

The post pulled 212 comments in about a day. Designers and agencies from the US, the UK, Pakistan, India, Nigeria, Spain and more jumped in, plus a stack of messages on Facebook, Instagram and LinkedIn. Every one of them got a reply. 29 people wrote out a real answer to the question. In all they shared 95 sites. We studied 27 of them in depth, ran the other 68 through the same phone test, and rebuilt one site per person, 42 in all.

This is what they said, what their sites actually do on a phone, the 42 rebuilds we sent back, and what those rebuilds cost us.

What 29 Designers Say Makes the Best Website

I read every written answer and tagged what each one talked about. One answer can hit several things.

Clean, uncluttered layout came up most, in 16 of 29 answers. Fits the brand followed with 15, then a clear next step with 13 and strong photos and visuals with 12.

The best website design isn't about looking pretty, it's about working. Alan Mariotti

Zia Ahmed went past the page and into what happens after the click. Tracking, forms that reach the right inbox, and a structure the owner can update without breaking it. Ceobaby Dadon said a great site should save the owner time, like a booking page that fills the calendar overnight.

Only 2 of 29 brought up ranking on Google. That one matters, and I will come back to it.

How We Measured the Sites

Every site got the same test. Google Lighthouse on a simulated phone, run twice where the site allowed it, plus accessibility, page weight, the first phone screen, the forms and the tap to call links. For several we also pulled what Google ranks them for through DataForSEO.

Then each site went into one of four groups.

• Best means a phone score of 90 or more, accessibility of 90 or more and nothing blocking leads or search.
• Good means a phone score from 75 to 89, or a fast site with one small flaw.
• Bad means a phone score under 75, or one problem that costs leads or search, like no title tag, no form, a speed claim the site does not meet, or a live site set to noindex.
• Ugly means a phone score under 60, or something that breaks the site for Google or for customers, like hacked pages or noindex on every page.

Ugly is about what the site does, not about taste. Some of the best looking sites in this group landed in Bad or Ugly because of one setting nobody checked.

The Best

Four sites scored 90 or more on a phone with nothing in the way of a lead.

What they share is restraint. Small pages, few requests, one job per screen. Elmir's tool page scored 100 on everything. Reply Rail puts tap to call and text right where a phone user needs them.

The Good

Seven sites are solid and close. Each has one thing worth fixing.

Marc Tecke's work is some of the strongest design in the whole thread. His host blocked Lighthouse, but our own phone lab test loaded his home page in about 2.3 seconds. The gap is search, not looks. He was not in the top 10 for his own name.

The Bad

Eight sites lose leads or search to one or two fixable problems.

Two patterns show up here. The first is a promise the site does not keep. One hero says 99 plus Lighthouse and scores 78 and 70. Another says 1.2 second loads and scores 77. The second is a missing next step, like a form 14,982 pixels down a phone screen, or no quote form anywhere.

The Ugly

Six sites have a problem that stops Google or a customer cold.

The heaviest home page weighed 14.6 MB. Its main picture took 18 seconds to show in one test and 64 in the other. One portfolio tells Google not to list any of its pages. One client site had its menu hijacked to casino pages. One AI studio ships its proof counters as 0 products and 0 countries until JavaScript runs. None of that shows up in a screenshot. All of it shows up in a test.

Two Sites We Could Not Score

Two hosts answered Lighthouse with an error, so we left their speed blank instead of guessing.

What They Said Versus What Their Sites Do

11 of the 29 answers said a great site loads fast. The median phone score across the 24 sites we could test was 77. Eight scored under 75. Six reached 90.

13 of the 29 said the next step has to be obvious. Three sites in this study put the only call link or form more than 5,000 pixels down a phone screen.

Only 2 of 29 talked about Google. Of the designer sites where we pulled rankings, none ranked in the top 20 for a search that would bring them a client. Pixril, BeniBrain and Dania360 showed 0 ranked searches in their home country. AIUNSEEN showed 1, at position 80.

That is the real finding. The design talk was good. Clear, honest, smart. The gap is between knowing it and shipping it, and the part almost nobody mentioned, search, is the part where the work stays invisible.

Every Site They Shared

A lot of people shared more than one site, some a whole client list. We built one demo per person, but we ran every site they sent through the phone test. That is 68 more sites from 24 people.

These were measured on speed, page weight, accessibility, layout shift, the page title and whether Google is allowed to index them. We did not check the forms or call links on these, so the quick grade is about the phone test only, and they sit apart from the 27 studied above. Client sites stay unnamed.

Ali Raza Janjua

20 sites, median phone 61.5

Bryn Walker

13 sites, median phone 63.5

Léyah Nycolé

5 sites, median phone 65

Palak Patel

5 sites, median phone 44

Russel Joy

5 sites, median phone 67.5

dev_ayobamiakande

2 sites, median phone 74

Affan Ahmed

1 site, median phone 83.5

Amine Boumahdi

1 site, median phone 83.5

Colby Black

1 site, median phone 70

Crea8 Agency

1 site, none could be scored

Dania360 (José)

1 site, median phone 74

Fatima Kaleem

1 site, none could be scored

Gulf Coast Technology Company

1 site, median phone 74.5

Junaid Raza Maher

1 site, median phone 72.5

Kody Kronberger

1 site, median phone 74.5

Lane A Houk

1 site, median phone 95.5

MaAz Ahmed

1 site, median phone 89

Moe Siddig

1 site, median phone 84

Monica Fenty

1 site, median phone 90

Mudassir Shahbaz

1 site, median phone 72.5

Muhammad Awais Ali Khan

1 site, median phone 95

Peter Campbell

1 site, median phone 98.5

Ryan Mitchell

1 site, median phone 57

Talha Mahmood

1 site, median phone 85

What We Believe Makes the Best Website

That is what the designers said and what their sites do. Here is what we build to at Clickflame, every site, ours and our clients'. Most of it is boring on purpose. It is the stuff that turns a visit into a call.

1. One Clear Next Step on the First Screen

   Call, text or a short quote form sits right there before any scrolling. On a phone a bottom bar keeps Call Us one tap away. In this study three sites put the only call link or form more than 5,000 pixels down.

2. A First Screen That Shows the Work

   Real video or the real product, big. One headline, one offer line, two buttons. No stat rows or tables in the hero.

3. Real Photos Over Stock

   Real jobs, real people, real places, and the photographer gets the credit. No AI people standing in for staff.

4. Built for the Phone First

   The phone gets its own layout, not a shrunk desktop, and it loads fast on a normal connection. The designer sites in this study had a median phone score of 77. Our rebuilds of them had 98.

5. Easy to Read

   Light pages, plain readable fonts, nothing under 14 pixels and strong contrast, so anyone can read it in the sun.

6. Proof You Can Check

   Reviews, licenses and awards link to the source that issued them. No invented numbers and no made up clients.

7. Built to Be Found

   A title and description that say what and where, with the phone number. Structured data, one address for the whole site, a map on every page, and pages that answer what people actually search.

8. Nothing That Gets in the Way

   No sliders, no sideways scrolling and no pop ups before the visitor has read a word.

9. Honest and Helpful First

   Say what you do in plain words, own your mistakes, and help the owner grow before you try to sell them anything.

Three of Ours, Measured the Same Way

One for design, one for traffic and one for local search. Same phone test as everyone above, checked November 8, 2026.

Best for Design

Best for Traffic

Best for Local SEO

The 42 Rebuilds

One demo per person, for the site with the most room to grow, sent back to the person who shared it. Elmir's was the only one we left alone. The screenshots below are the first screen of each live demo. Click any one to open it.

The median rebuild home page weighs about 350 KB.

Want In on the Next Round

The thread is still open and this page will grow. If you build websites and want your site measured and added, drop your company and main site in the comments on the original Facebook post. If you were measured here and want to add a comment or a correction, message me and I will put it on this page.

Lighthouse 12, simulated phone, two runs per site, run November 7 to 9, 2026. Rebuild numbers are the mean of two runs before and after. Rankings from DataForSEO on the same days. Written answers are quoted as sent, with spelling cleaned up. Client builds are named by the designer who shared them, not by the business. Every rebuild is a private demo set to noindex.

Anthony Limpert runs Clickflame, a web design and SEO company in Long Beach, and Supercanic. He builds the sites, measures them, and publishes the numbers here on Data for Trades.