I suspect (not tested in this study) that the hyphenated version is seen as a possible fake if the unhyphenated version also exists. That is to say, face-book could be seen as fake because facebook exists, but a novel hyphenated phrase like vexing-sesquipedalian would not suffer the penalty because vexingsesquipedalian hasn't been used.
It would be tricky to test this, because you need to use a domain with hyphens where the original doesn't exist (and then you introduce new variables into the mix)
This was my first thought, as well. Also, with how often search engines re-factor their search algorithms, it makes you wonder if an article from 2011 is still even relevant? Anyone else have some insight maybe on that front (relevancy a few years later)?
It could be that the ranking prior to the swap is continuing to have some lasting effect on the sites after the swap. It would be worth doing the same test on new domains (with new content), to be sure there weren't any side effects from the swap itself.
There is a very real possibility that hyphenated domains have a negative impact on scoring with Google because historically hyphenated domains have been spammy (vs. un-hyphenated domains) - and Google has picked up on this correlation.
It might have been a GOOD (or indifferent) idea a decade ago. But it's a bad idea if you're registering a new domain. There's different rules for old sites and new sites, and if you're making a new one, play it safe: don't have a hyphen.
I think there's a little more to it than that. If the exception in question can be seen to have specific properties which are clearly the cause of its exceptionality, and if those properties can be seen to be rare, I think the exception supports the conclusion that the rule is a pretty good one, by showing what an unusual combination of circumstances is required for it to be violated.
That's pretty much what the saying is about. The existence of an exception proves the existence of a rule. The reliability of the rule can be estimated by considering the exceptionality of the exception.
My first thought was to the age of the domains ... all are the same though, I wonder what order they were crawled in?
I'm sure there are plenty of sites that contain hyphens, that should be favoured (and probably are with all the other signals) over their hyphenless cousins, still seems unfair if there is a penalty though!
I have an hyphenated domain name and it has a very decent page rank. What I have found is that for a particular keyword the sites with that keyword in the domain name will rank higher than ones without the keyword.
We recently had an internal debate about whether the anti-hyphen bias extends to the rest of the url. We haven't tested it though. I'd be curious if anyone has ...
Thanks for the link! Very helpful. However, my own searching suggests Google treats /something-about-something/ similar to /somethingaboutsomething/. So it isn't a perfect example. If the same content is on both url's, would one always win?
Our own debate was whether to hyphenate a two word phrase in the directory structure: /kitchensink/whatever/ vs /kitchen-sink/whatever/. So we had to do a similar debate as the domain hyphen analysis. Spam signal/Google treatment vs. User Convenience vs. Does it really matter.
The suspected reasons to why there's an anti-hyphen bias (primarily, fakers trying to capitalized on well-known non-hyphenated names) don't seem to apply for the rest of the url. i.e. a legit brand has no control over what a spammer does with an impostor domain, but the legit brand has complete control over what their own url structure looks like
Good point as to domain credibility. But I have to admit that I groan whenever I see a legit brand with a url that is: .com/our-red-widgets-are-perfect-for-your-classic-1957-red-chevy/
The sheer number of people doing it makes me wonder if at some point they'll implement it, if they haven't already.
Positions two and three swapped for me. Searching from .co.uk from Firefox private browsing. With this post gaining popularity though (it appears at position 5 for me), I doubt we can learn anything more from these sites.
It would be tricky to test this, because you need to use a domain with hyphens where the original doesn't exist (and then you introduce new variables into the mix)