We removed 1,006 names that were never studios

For a while, blogger.com was the 25th biggest web studio in the world according to our ranking.

The number was not invented. Thousands of websites really do carry Blogger's name in their footer. Our crawler found those signatures, counted them correctly and did exactly what it was designed to do.

The problem was one step later.

Blogger is not a studio.

That distinction led us to review a much larger part of the No Man Site register. By the end of it, 1,006 domains had lost their studio classification, our exclusion list had grown from 208 domains to 1,137, and we had manually checked thousands of records that automation alone could not safely decide.

The register became smaller.

That was the point.

Counting a signature is not the same as understanding it

A large part of No Man Site is built from evidence found on websites themselves.

When a studio places its name or link in the footer of a website it made, that signature gives us something unusually valuable: evidence of a relationship between a studio and a piece of work.

It is much better than asking agencies to submit portfolios and then taking everything on trust.

But platforms leave signatures too.

Website builders do it. Hosting companies do it. Theme vendors do it. Analytics services, marketplaces and publishing platforms can appear across thousands or millions of websites.

Blogger had accumulated enough such appearances to reach number 25. A shop selling website themes had reached number 44.

The counts were real.

The classification was wrong.

We found that one of the three mechanisms capable of assigning the studio type was relying on exclusions elsewhere in the pipeline instead of applying its own.

That has now changed.

Any rule capable of deciding that an organisation is a studio now carries the exclusions necessary to make that decision itself.

The list behind those checks grew from 208 addresses to 1,137.

Then we checked the people behind the numbers

Fixing the rule prevented the same mistake from repeating. It did not tell us how much damage the old rule had already done.

So we went through the register.

2,028 studios were present because we had first discovered their signatures in the footer of somebody else's website. Because that is also the exact route by which platforms can enter the system, every one of those records was reviewed.

Another 97,246 entries had come mainly from agency directories. We ran a broad filter across their names and descriptions looking for terms associated with hosting, themes, builders, marketplaces, stock imagery, analytics and similar services.

That produced 1,035 suspicious records.

Those were checked individually too.

By the end of the review, 1,006 domains were no longer classified as studios.

We did not delete the evidence we had collected about them. We corrected what that evidence meant.

The 226 we kept mattered just as much

A fully automated cleanup would have damaged the register in the opposite direction.

226 records flagged as suspicious remained classified as studios.

Some genuine agencies also sell hosting. That does not turn an agency into a hosting company.

Some work in industries that look suspicious to a keyword filter.

goodgiant.com, for example, works with casinos. That makes casinos part of its client market; it does not make the company itself a casino.

And keenstreet.com.au was caught by the word casino for an entirely different reason: Casino is a town in New South Wales.

A regular expression cannot understand that distinction.

A person can.

That is why our approach remains deliberately uneven: machines should do the enormous amount of work involved in finding evidence, grouping it and detecting anomalies. But where the evidence can reasonably mean two different things, the final judgement still belongs to a human.

A smaller register can be a better register

Directories are usually rewarded for growing.

More companies. More pages. More entries. Bigger numbers.

A register like ours has a different obligation.

If we cannot explain why something is classified as a studio, it should not be classified as one simply because removing it makes the database smaller.

This review removed 1,006 studio classifications.

It also left 226 suspicious ones in place because the evidence supported keeping them.

Both decisions matter.

No Man Site is not useful because it contains a lot of names.

It is useful only if those names mean what we say they mean.

The full methodology is described in How this register is built.