Roger keeps telling me my canonical pages are duplicates

[icecarats]

I've got a site that's brand spanking new that I'm trying to get the error count down to zero on, and I'm basically there except for this odd problem. Roger got into the site like a naughty puppy a bit too early, before I'd put the canonical tags in, so there were a couple thousand 'duplicate content' errors. I put canonicals in (programmatically, so they appear on every page) and waited a week and sure enough 99% of them went away.

However, there's about 50 that are still lingering, and I'm not sure why they're being detected as such. It's an ecommerce site, and the duplicates are being detected on the product page, but why these 50? (there's hundreds of other products that aren't being detected). The URLs that are 'duplicates' look like this according to the crawl report:

http://www.site.com/Product-1.aspx

http://www.site.com/product-1.aspx

And so on. Canonicals are in place, and have been for weeks, and as I said there's hundreds of other pages just like this not having this problem, so I'm finding it odd that these ones won't go away.

All I can think of is that Roger is somehow caching stuff from previous crawls? According to the crawl report these duplicates were discovered '1 day ago' but that simply doesn't make sense. It's not a matter of messing up one or two pages on my part either; we made this site to be dynamically generated, and all of the SEO stuff (canonical, etc.) is applied to every single page regardless of what's on it.

If anyone can give some insight I'd appreciate it!

[JoelDay]

Hi Valery,

I took a peek at your campaign and it looks like those few remaining duplicate pages are in fact different, but very minor differences. Basically there's pages for different sizes of things.

While being different, they vary in such minute ways that Roger see's them as duplicates.

I Hope that answers the question.

Thanks,

Joel.

[icecarats]

Joel -

Thanks a lot for looking into that. The pages are very similar, so I'm not surprised they're being duplicate triggered; but what does surprise me is that they are apparently being considered duplicate to a canonical version of themselves? When I click on the duplicate list I'm expecting to see:

Product1.aspx

Product1-Blue.aspx

Product1-Red.aspx

But instead I'm seeing:

Product1.aspx

product1.aspx

product1.ASPX

And so on. The first scenario to me implies that the 3 pages are duplicate to each other, whereas the second is saying that there's either a canonical problem or I literally have different-case versions of those files.

[JoelDay]

Hey Valery,

I see those on closer inspection. I know it looks weird, but that's accurate. Your server must be UNIX or Linux so they will actually treat case as a different word.

For example: banana.com/pancakes.html would be treated differently than banana.com/PanCakes.html.

So if you have any pages generated dynamically or otherwise that differ only in case, then they will be tagged as duplicate.

In your CSV file you can see the duplicates being caused by case. I'd also be happy to help provide a few specific examples but would want to generate a ticket for you so we don't divulge any private information.

Cheers,

Joel.

[icecarats]

I think I get your point; the canonical is pointing to where the juice should go, but the URLs are still functionally different things. I'm guessing some sort of URL rewrite is in order, and to standardize how I do in-text links on the site (with user-editable content this part could be a pain).

[ThompsonPaul]

Valery, you're definitely going to want to normalize your URLs to lowercase. This is a quirk of IIS that it actually respects case in URLs and will consider different case URLs as different pages.

In addition to the search engine problems it creates, it's also a major problem for usabilty - yours and your users. For example, a user who is trying to type in a direct URL can get a 404 error depending on what case they use.

More importantly, your Google Analytics will report on each of those version as separate pages, unless you write a normalizing filter into your GA profiles. Better to do that normalization for the actual site, not just your analytics

While rel=canonical can resolve a number of issues, I've always found it vastly better to correct the actual problem at its root, rather than rely on canonicalization as a catch-all. Anecdotally, I've found correcting issues like this with rewrites seems to allow affected pages to rank better than when just corrected with canonicalization. WIsh I could find time to do an actual case-study on that

Managing rewrites on IIS servers will require a plugin like asapi-rewrite as IIS doesn't handle it natively.

P.S. IIS will also allow and respect spaces in URLs. Users in Internet Explorer will see them as normal with spaces but browsers like Firefox will insert the html entity for a space (%20) into each necessary spot in the URL. This is again a mess for usability, so much better to force rewrite of all URLs to replace spaces with dashes when creating new pages. Many CMSs have plugins for this or you can also use sitewide rewrites to do it after the fact.

[icecarats]

ThompsonPaul -

Thanks for that info, it pretty much nails exactly what I had discovered independently. This is an IIS7/Win2k8R2 install so luckily the rewriting is a bit easier than in previous iterations. The whole platform is hand coded by us (after the 10th ecommerce site or so you can generally do them in your sleep) so I don't have to worry about CMS implementation and the like, and luckily we already knew that about the spaces so they simply aren't allowed in the filenames. I'm in the middle of making a regex right now that is going to down-case anything in an href="" or src="" tag that will hopefully handle everything on the site side user-created or not. Will consider what to do in regards to external links a bit down the road I think.