Public audits, not case studies

Why this exists instead of case studies

Most agency sites prove themselves with a logo wall and three numbers. We do not have client results worth publishing yet, and inventing them is not an option on a site whose whole position is that it publishes its sources.

So this is the alternative: public, reproducible audits of real businesses, using only facts anyone can re-check with a terminal and a minute.

The standard

• Mechanical facts only. What a server returned to an ordinary HTTP request on a stated date. Nothing about how any assistant actually treats the business.
• No assistant testing. Measuring an assistant honestly needs a frozen prompt set sampled many times. A single run is noise, and we will not publish noise.
• Reproducible. Every line in every table can be re-run with curl. The method is published with the findings.
• Dated, not silently updated. Sites change. When one does, we note the date rather than rewriting history, because the fetch date is part of the claim.
• Corrections on request. If you run a business covered here and something is wrong, tell us. We re-fetch, correct it, and say that we corrected it.

Why we are not naming winners

Every business in these teardowns has a good site. They are fast, they look right, and they sell what they are meant to sell. Nothing in them is a failure by any conventional standard.

What changed is that a second kind of reader now matters, one that gets a single HTTP response, keeps a couple of hundred characters from around the heading, and never sees the photography. Almost nobody has written for that reader yet. That is why the same gaps turn up across independent businesses on the same platform, and why it is cheap to be early.