Wrong language bug spreads from WMT to Analytics

[Sgt_Kickaxe]

I tried to look up a few keywords in analytics a moment ago and could not. Reason: I don't speak most of the languages being reported (kudos on showing me russian and arabic characers properly though). My page titles are all being reported as mixed languages and in some cases multiple languages per title.

What's going on? This was happening in Google webmaster tools earlier this week.

If you have tools that export this data, make sure the data is not corrupted.

[g1smd]

It's still happening in WMT: initially French, Spanish, German, and some eastern European language, and now Arabic too.

[Sgt_Kickaxe]

WMT and Google Analytics and ... ?

*fix it!*

[lucy24]

I would love to know what the variable is. Your IP? Some specific other g### service that you're using, or have used in the past? Your exact configuration of privacy settings? Your UA? Some combination of these?

[g1smd]

Some wag thought that it would be good to randomise it knowing that lots of people would talk about it. :)

So far it's working.

[jecasc]

Don't worry guys. This is all part of Googles ongoing effort to predict user intent on past behaviour and then provide the according services.

Based on your past user behaviour Google has learned that your subconciousness had decided to take language courses and the user interface has been changed accordingly to support your learning process. Unfortunatly the changes kicked in prematurely before your subconciousness told you conciousness to sign up for the course.

But hey that is nothing new - Google presents it's users websites they do not (yet know they) want to visit all the time. So why are you surprised. Google knows what you want - before you do.

[lucy24]

Some wag thought that it would be good to randomise it knowing that lots of people would talk about it.

I thought about that. Seriously. In case anyone hasn't noticed the new WMT, or hasn't been checking every five minutes, they will now :)

Based on your past user behaviour Google has learned that your subconciousness had decided to take language courses

This would be a terrific explanation-- except that my accept-language headers are as long as your arm, suggesting that they're only throwing languages at people who are not interested. Does anyone hereabouts have their first-choice language set to anything other than English? I think everyone in this thread is a native speaker.

Doesn't seem to be browser-based. I get the same page in all five. (MSIE predictably doesn't work. It claims to want more cookies, but it's lying in its teeth. I checked. Didn't try Lynx.) How 'bout: Some particular category has to have a non-zero value? Or: they goofed at a fundamental level, so if any category has a value greater than 255 the whole thing goes haywire.

Then again, they may be holding some special bug in reserve just for me ;)

[glitterball]

Maybe worth a try, but in the past I've noticed Google language bugs where the user gets redirected through a localised tld (e.g. Google.co.uk gets redirected through google.com). Thus preventing changes from "sticking".

[Sgt_Kickaxe]

It's now spread from GWT to Analytics and into SEARCH RESULTS!

I noticed gibberish like "یاÙ" appearing in my reports so I ran it through Google translate and it auto-detects french(I know french, that's not french). I then tried running some of the weird stuff as a search query and some reputable sites display gibberish in the sers. I spotted a NY Times article and several youtube videos in the sers that, when visited, had no trace of the junk.

Could it have something to do with the new "translated foreign content" option in the sidebar? It looks like Google has translated all webpages into other languages for that option which is something that could easily cause problems like this. (if the translated content link isn't new I apologize, I never saw it until today)

[lucy24]

I noticed gibberish like "یا�" appearing in my reports

I suppose nobody will ever know what you really saw ;) ("Does the color green look the same to me as to you?") It comes through for me as six bytes getting interpreted as two UTF-8 characters, but it isn't UTF-8. Or any other common encoding I can think of. Something language-specific?

But language and text encoding are completely different things.

:: detour ::

Hmmm. The g### search pages (both vanilla and advanced) include a utf-8 charset declaration... but the SERP itself doesn't. That's just asking for trouble.