We graded our own websites. The two that sell came last.

We scored the four sites we run on two 50-point scales. The two Forge sites, the ones that bring in work, scored lowest on both. Here is why, and what we fixed first.

On October 31, 2026 we scored the four websites we run: this one, forge-automations.com, and two side brands. Each got two scores out of 50. Craft asks whether the design is good. Integrity asks whether the site is honest, fast and free of the tells that make a page look like a template.

The scores

The two side brands cleared the bar on both scales. The two Forge sites missed it on both. Those are the sites that are supposed to win us clients.

Why the gap

The copy on the Forge sites scored fine. The difference was how each site started. The two side brands were designed against a reference: a short list of real sites that do the job well, with a reason for every borrowed move. The two Forge sites were built straight from a starter layout and patched as we went. The top of this page held together. The bottom half turned into the same four price cards and question accordion you see everywhere.

What we fixed first

• The home page video had no poster frame, so the first thing a visitor saw was a grey box until the video loaded. We added the first frame as a still image.
• White text on our orange button measured 3.6 to 1, under the 4.5 to 1 minimum for readable text. We darkened it until the text measured 4.8 to 1.
• Text in 52 places was set at 9 or 10 pixels. We raised it to a size people can read on a phone.
• The four price cards became rows, one per job, so the bottom of the page follows the same idea as the top.

None of these took long. The first three were invisible to us because we had stopped looking at our own site as a stranger would.

The lesson.Every site we build now starts from a reference and gets both scores before and after. Fixing a site that started from a template cost us a second round of work. Starting from a reference would have cost one.

What it means for you

We do not build websites for clients. The habit carries over to the work we do sell: measure the thing before you change it, write the number down, and measure it again after. If a workflow install cannot show you a before and after, you have no way to know it worked.