|Google Updates and SERP Changes - May 2012|
< continued from [webmasterworld.com...] >
Google *insisted* on presenting me with nothing but results for:
"TRADE_PERSON KnoxVILLE" .... WTF!
I've tried the same search 12 hours later and now i do get what i was actually trying to search for, "TRADE_PERSON Knoxfielld". Obviously different servers, constant flux...
Zombie traffic anyone? ....well heres one of the main reasons. Google trying unsuccessfully to second guess its users and present them with wrong results. User ("used to" getting the best results from Google) automatically click the link but oopes...it is not what he/she was looking for. Click back.
On both occasions Adwords ads for "TRADE_PERSON Knoxfielld" was spot on. Coincidence?... ya right!...
I wish they'll get it over and done with already and move to "paid only" results. Maybe the broadening of exact match and phrase match (Adwords) are the first signs of them heading down that route...
TRADE_PERSON = insert "doctor", "painter"...etc
[edited by: tedster at 3:43 am (utc) on May 9, 2012]
Off topic for a serps thread, but commenting on points made in the opening post of this thread...
Obviously meant to be an example... but I wonder whether you've intentionally used the apparent misspelling "Knoxfielld" to make a point, or if that's the place you actually were looking for.
I ask because most of the really bad results I find on either Bing or Google are bad because there are hardly any pages satisfying the query. Google uses what's statistically the closest match, and it relies on searcher behavior to provide corrective feedback. Bing appears to do the same, with a different set of matching criteria. This isn't much different from what goes on in the real world. If your name happens to be John Joans, you probably get used to it. ;)
Taking the "Knoxfielld" example literally (probably not how it was intended, but it might serve to illustrate the issue)... when I search for just the word Knoxfielld, both Google and Bing try to force Knoxfield on me. After going through all the PITA process of correcting the spelling, I get more or less 7 results on Google, two of which are from this thread. On Bing, which also auto-corrects spellings, I get 2 results, and this thread is apparently not in the Bing index.
If I search for Knoxvile, both Google and Bing return results for Knoxville. If I don't flush my cookies on the Knoxvile search, I sometimes eventually get Knoxvile by default on Google.
Bing has begun to do something interesting... it combines results for both the correct spelling and the misspelling, so for the [knoxvile] search it returns this message above the organic results...
|Including results for knoxville. |
Do you want results only for knoxvile?
I've yet to see these combined results rank the unusual spelling anywhere near the top, so you generally need to request the correction. Bing clearly is thinking through the psychology of user satisfaction better than Google is, by "including" rather than excluding the user request. Effectively, the process is the same.
I don't want to pollute the serps by suggesting "knoxvile" two-word phrases, which do exist, except to say that on searches where knoxvile is a place, I'm seeing both Bing and Google return Knoxville, not Knoxvile ads. The statistics are going to vary from phrase to phrase, so it's impossible to generalize from either example who's buying ads and what they've requested that the match type be. Note that that's a big difference between ads and organics... you can specify targeting and misspellings in AdWords that are independent of the organic search algo.
There are lots of areas where there are multiple uses of a word, and the unusual uses need to be targeted by ads. That's one of the things I look for when evaluating target keywords.
I don't think, though, that Google is purposely returning spelling errors in organic to be able to sell more literal results in the ads, which was, I'm assuming, the concern of the opening post.
|I don't think, though, that Google is purposely returning spelling errors in organic to be able to sell more literal results in the ads, which was, I'm assuming, the concern of the opening post. |
The typo was made in error. Google should have returned
"Do you mean TRADE_PERSON knoxfiled?" (name of a known local town near by).
But NO, Google insisted on presenting nothing but results for "knoxville" with the "did you mean" link nowhere to be seen. I actually clicked a link in error thinking i am going to a result about "knoxfiled". When i tried the same search phrase 12 hours later it worked as it should. Different servers, constant flux...what i saw, may not be what you are seeing now.
This crap probably happening trillions of times per day. No way they don't see it... they've been dumbing down the SERPs for almost 2 years now. I no longer give them any benefit of the doubt.
It walks like a duck and sounds like a duck...
Another day another change. To my eyes google is still testing. The #2 spot is where they put suspect sites and then they get demoted down the page over the course of a week.
Is anyone else seeing huge amounts of traffic on new keywords recently that aren't very relevant in lieu of earlier keywords that weren't very relevant? Again almost like a new test. Totally random example of two different meanings lets say your a professional dog walking service. All you do is go to apartments to walk dogs. Suddenly last week google starts sending you hundreds of visitors a day for "walking dog" and those people bounce at 100% because they were really looking for information on the proper way to walk a dog, not a service of dog walking. This is not a real example and I didn't test it but it wouldn't surprise me if in penguin it's an issue. Is this partially google suggest going haywire? On this set of terms only similar businesses rank and the keyword tool lists few searches each month yet we are getting hundreds a day on these two words?
Overall I don't think the update is done or a new version started rolling in the last 36 hours
|Solid, highly relevant, plural two word terms have remained unaffected, but strangely we lost many of the exact same terms in their singular versions on which we used to rank equally well. Anyone else observe this? |
Yes! There are 2-3 word terms that I consistently ranked #1 both singular and plural for years. Starting Friday, April 27, the plural (unfortunately the less popular in this case) remains #1 but the singular is bouncing around 5/6/7. Curiously, synonyms still rank #1 both singular and plural.
Some of the sites that moved above were previously in the 2/3/4 spots, while others are scrapers that hadn't even been first page before this. This is only impacting a few search terms, but perhaps it's no coincidence that these are also my highest volume queries. In each case, the term appears (in singular or plural form) in the page title, H1 tag, and in the page content. This is entirely appropriate given the page context, but perhaps this has triggered a spam/over-optimization penalty that for some reason targets only the singular term.
web_speed... fascinating. Some brief notes for now, as it's late, and I don't want to take this discussion too far off the topic of serp updates. I think you are onto something worth exploring, and I do have some thoughts about where to discuss it if I can catch up with my own work. This is potentially a very long discussion.
But, let me emphasize that I don't think it's Google being duplicitous. Google simply doesn't have the data, and from the checking I've done, I'd bet the data is rare to non-existent. Eg, neither knoxfield nor knoxfiled shows up in a database of the top 25,000 places in the US as sorted by population, and I couldn't determine population for them anywhere on the web.
|But NO, Google insisted on presenting nothing but results for "knoxville" with the "did you mean" link nowhere to be seen. |
That's how all the engines I tried react to queries that include these placenames, and I tried a bunch. To oversimplify what I saw, just about every search involving either "knoxfiled" or "knoxfield" reverted to "knoxville".
Google occasionally returned knoxfield, depending on what the TRADE_PERSON or widget search for was, but on many of the searches I tried, there was no "did you mean" link because there were no matches for the queries.
It's possible that Google, as you say, is switching databases and they may have data satisfying the queries somewhere, but I have a very strong feeling that there isn't much. I have seen, during an update process, scanty data disappear for a while and then come back. As I understand it, that's the way a very large set of interconnected database systems has got to work. There's necessarily a degree of latency, and the tradeoff for speed is temporarily dropped data.
Also, on all the engines I tried, and I tried a bunch, every search for knoxfiled tn reverts to knoxville tn.
For "knoxfield tn" in quotes, Google showed only 2 results (and one of those is like a wildcard site and nothing's there)... and for [knoxfiled tn], without quotes, Google showed only 8 results. None of the other engines showed any results at all.
Also, when I tried to set location, which is a fairly dependable way of observing US geo-effects on Google, knoxfiled, tn and knoxfield, tn both tripped a "location not recognized" message in red. Of those we're discussing, Knoxville was the only location recognized in the location setting. Conceivably, Google is working off a variant of a placename database that's commonly used, and those names aren't there, and there are some good reasons why they might not be.
Rough observations about the ads... they appear to be tripped by a combination of keywords and IP location... I'm not an AdWorda person, but I know you can target AdWords locally by IP... so for certain trade_person queries, if I used my own default location, I'd get what appeared to be national trade_person directories targeting, say, a zip code radius... just a guess. So, if Knoxfile is near you, possibly you'd get a local ad even though there are no listings in organic. How large is Knoxfile, and is it an incorporated town, a suburb, or an unincorporated region?
I see that scottsonline is posting on this thread. He and I went through a discussion of a similar rare query search, involving also Google Suggest, autocomplete, spelling correction and query rewriting, and results not appearing that for content we saw was indexed. This was back in June and July 2010, as Caffeine was with us. See these threads...
A Red "Did you mean: ____" Shows in Google Drop Down Suggestion
2:Google Updates and SERP Changes - June 2010
- - Clickable link would break in WebmasterWorld redirect script in this url:
Among other things, the discussions make clear why we don't want to use too many specifics here, as we end up greatly distorting the results.
seems like sites with keywords in the domains are really popping high over past few weeks - many of which are pure crap
@ Robert Charlton
I am sorry, i should have mentioned that i am in Australia.
Knoxfiled is in Victoria Australia.
A search from my IP usually yields the proper results 1st go. It didn't at the times i mentioned above which leads me to believe that it is probably a serious geo targeting issue.... or is it? (as mentioned, exact match ads were spot on for plumber, dentist etc in knoxfiled... evident by their display phone numbers).
[edited by: Web_speed at 8:49 am (utc) on May 2, 2012]
they also have been breaking up domain names in to keywords in suggest as well
I also launched a site 5 days ago and the domain name/keywords have it ranking number 1 - without me being signed in - signed in I also have individual posts of the blog
@ Robert Charlton - this kind of location information confusion has been going on for some time now.
I have one site where four locations are confused in the same way as the OP.
In other words, when I check rankings for location 1 my location 2 pages are returned. When I check rankings for location 3 my location 4 pages are returned.
(Note that the pages are not thin location pages with few content changes - they are distinct locations with different information where the inhabitants would string you up if you confused the two and the only similarity is in their names - as per OP.)
As a test, I moved location 1 to a new domain. Now location 2 shows correctly on the old domain (as it always did) and location 1 shows correctly on the new domain.
I also moved both location 3 and location 4 to the new domain (they are actually in the same folder and closer geographically). Both are still confused in the same way on the new domain.
My theory is that this is an inadequacy of Google's semantic machine learning when it comes to place names.
NB. Google KWT exact match:
Location 1 : 14800
Location 2 : 74000
Location 3 : 33100
Location 4 : 14800
web_speed - Thanks for the correction.
I can't really venture an opinion for Australia, in part because the specifics of local geo-targeting are different in every country, and I don't have that data. But you get the principle. In the US, there are a lot of place names which seem relevant to the people there that don't make a blip in the population data. I'm guessing that may be true of Australia as well.
Here are some stats I'd noted earlier for placename search only, without modifiers or country ID...
|word occurrence [knoxville] in Ggl |
word occurrence [knoxfield] in Ggl
1,580,000 (many not in US)
word occurrence [knoxfiled] in Ggl
I was seeing non-US results in the above, some of which must have been Australia.
|above which leads me to believe that it is probably a serious geo targeting issue.... or is it? |
I can't quite tell, because both of your examples confuse me, as I'll describe....
...but that is a Knoxfield page you are returning. There is no Knoxfiled in Wikipedia. That was, in fact, one of the searches I'd tried. Apparently, Wikipedia internal search is independent of Google search. I also had tried searching Wikipedia with Google, and didn't get Knoxfiled that way either....
|There were no results matching the query. |
The page "Knoxfiled" does not exist
So, maybe Knoxfiled is misfiled ;) ...or maybe it's also a rare search. I had to use quotes on the placename in the query below to keep "knoxfiled" from getting rewritten, but the query only returns 647 results....
I'm sure, though, that smallcityname, countryname is not a common way of describing the place, and you may have more commonly-used searches you can check with.
I'm confused by both your examples in terms of what you actually entered and what was returned. Possibly, if you've made typos posting here, which is how I'm interpreting what I'm seeing in your posts, then Google is perhaps statistically justified in its rewrites of our queries... that they may be helpful more often than not. I hate them when I happen to be write, but Instant is such a PITA that I find myself making many more typos with it than I do with regular typing, and I wonder whether those too are somehow a factor in the algo.
I'm going to bed, but please clarify exactly what placenames you'd typed into the Google search box. My assumption is you're saying you'd typed in "knoxfiled", but that got rewritten to something else, and that your examples here just happen to have their own typos.
I'm also guessing that your knoxfiled searches are rare enough to trip the query rewrite... but there might also have been wires crossed within the multiple database structure that Google uses. In systems like theirs, I assume that they have databases that do nothing but manage the order of changes within other databases.
|My theory is that this is an inadequacy of Google's semantic machine learning when it comes to place names. |
stever - That's also an interesting angle, as actual place name locations are in many cases dependent on IP location, and maybe that's hard to sort out, or maybe it's not wired in adequately as yet. I think that the question of geo-intention also must factor into evaluating local queries.
|[knoxfiled victoria] |
|I think that the question of geo-intention also must factor into evaluating local queries. |
Possibly, although my locations are in the same country and their names are not duplicated (as far as I am aware) in any other country.
Location 1 and location 2 are in different states and are both locally important administrative centers. Location 3 and location 4 are relatively closer (7 miles) but each have populations in the thousands, are well-known individual tourist destinations and are not contiguous.
|i should have mentioned that i am in Australia. |
A little bell went off. The discussion we're having here connects, I believe, with several other discussions. See backdraft7's report in this thread about sales from Australia at particular times, which is curious, he says, because his "site is set to target US traffic only".
Zombie Traffic and Traffic Shaping - Analysis
web_speed - I'm wondering whether you can correlate times you're seeing anomalies with times backdraft7 is seeing traffic jumps. Even if you can't, I'm guessing there might be some kind of a connection.
|probably a serious geo targeting issue.... or is it? |
I think it may be more a case of word association/search volume factors in the algo auto-suggest area.
1. We target [placename] keyword1 keyword2, but the results shown are for [placename] keyword1 keyword2 KEYWORD3.
This is as if Google is saying, nope, you don't actually mean that, so I'm adding KEYWORD3 for you. It could be a result of search volume.
To prove it we recently added pages targeted for [placename] keyword1 keyword2 KEYWORD3, containing a small amount of information and with a link to our [placename] keyword1 keyword2 targeted page, on the basis of if you can't beat them join them.
Where these have been indexed these pages now rank above our pages for [placename] keyword1 keyword2.
2. We have a [placename] similar to a mode of transport. The SERP's get completely confused showing a mix of what I am searching for and mode of transport websites.
We also see that it may be Google is now ignoring all titles and the description meta tag. Which explains why you can have a page with just one mention of the keywords on the page ranking high, its like they have gone straight to page content.
Example 1: I found we were ranking for some [placename] searches because Google had picked up the word Home from the Home Page menu item, so we ranked for [placename] home !
Example 2: A transport related site is ranking well for [placename] keyword1 keyword2 ( non transport related keywords ) Google adds KEYWORD3, and the only mention on the page on that site is a graphic that shows a bubble when moused over. The bubble has keyword1 keyword2 KEYWORD3 in it.
Added: So in the case of Knoxfiled, it may be Google is saying nope, you mean Knoxfield as everyone searches for that and we think you have misspelt and no we are not giving you Did you mean.. as Google knows better than you ( which of course is not true )
Well, I knew it couldn't last....I had four days of what I'd consider "normal" traffic quality and conversions. Now, it's back to foreign traffic and plenty of US traffic, but zip on the conversions.
The past 4 days had steady sales all day, at regular intervals (as in past years). Somehow or another they are turning the quality spigot on and off. I suspect this happens when the new data sort begins another iteration. It's similar to the propagation delay when changing an IP address, two or more different locations are seeing two or more different results.
Overall search results are still bad with outdated news and thin blogs with plenty of Google ads above the fold taking the lead. It's no wonder paying customers can't find us. They fix one thing and seem to break another.
The "Ever-Churn" continues.
I'm seeing a constant rotation of sites rising to the top, then a day or so later falling to 6-10 or worse around a core of stable sites, then coming up again.
Is this something new or just sites testing some thresholds, perhaps quality thresholds? Any views ?
|web_speed - I'm wondering whether you can correlate times you're seeing anomalies with times backdraft7 is seeing traffic jumps. Even if you can't, I'm guessing there might be some kind of a connection. |
While i can not recreate the problem now (i bumped into it only once), it makes perfect sense as being the most likely cause to traffic problems at backdraft7 end.
I myself see wild foreign traffic fluctuation over my own network of ecom sites with zero conversation. Logs full of lost one hit wounder. Zombie bandwidth eaters...
It is no doubt a combination of Google suggest, wrong (and quite misleading at times) typos suggest plus geo targeting gone highwire.
As mentioned, i no longer give them any benefit of the doubt. While we web businesses continue to lose money and pushed to the brink, they Google, continue to consistently excel and double (even triple) their profits every quarter.
Mod's note: The balance of this post edited for excessive editiorializing and violiation of WebmasterWorld terms of service. See IMPORTANT - The Focus of This Forum [webmasterworld.com], which is pinned to the top of the Google SEO Forum index page.
[edited by: Web_speed at 11:52 pm (utc) on May 2, 2012]
[edited by: Robert_Charlton at 2:16 am (utc) on May 3, 2012]
The bot activity is back. My Google bot traffic was out of normal range from 3/24 to 4/27. Then it fell back into normal range. Today Iím back to 2x normal. What's up? SEO session starts and sales are within normal range. Are we going to see more yoyo this week?
|Their in-depth investigation is long overdue. Not to mention a long overdue major web publishers action against this monopoly. |
The one thing I have noticed about the internet is it does create monopolies, which in it's self is surprising given the wider choice the internet offers however
Wikipedia has a monopoly position on Encyclopedia
Ebay on auctions
Amazon on books
Youtube on video
Flikr on photos
Skype on phonecalls
Paypal on transactions (via email address)
and Google on search
And their are many more monopoly positions online. Crucially search is the only one to effect everyone online.
So should Google be more consistent ?
I think yes they need to be consistent about the penalties handed out and more open about the cause of such penalties. When people see the overall quality of the SERPS reduce then its only natural that speculation arises that changes are motivated by the need to increase adwords revenue.
Most of the monopolies you listed above have been created by Googles rank placement to begin with. Wikipedia, ebay and Amazon are major examples.
They'd all be lost a long time ago if not for google's rank treatment and exposure.
Check out ebay forums for so many "no sales" posts popping up lately... sure looks like someone has turned off the tap on ebay's traffic lately, too.
[edited by: Web_speed at 12:11 am (utc) on May 3, 2012]
After lunch today time on site jumped and traffic returned to normal patterns for the first time in days. Ironically just as backdraft reported it going the other way.
Clearly they are still testing.
Definitely testing. Traffic is down today compared to the past couple of weeks. I saw peak traffic today at 10am. It is normally 7pm. This happened last Thursday as well, before they rolled out some weird change. It looks like short-tail keyword traffic is down since that time.
I agree with you about Amazon and Wikipedia getting favorable results but eBay have not had such an easy ride with Google. I think all three though have developed communities and would work with or without Google
Wow, I'm seeing BIG changes again this morning. SERPs look like they did around 2-3 weeks ago. Loads of same domain results on page 1 again.
It looked as though they were heading in the right direction for a few days there, but seems as though someone hit "restart".
Getting 50% foreign traffic today on my .co.uk sites. These cycles seem to be happening very regularly now. Every Thursday - I wonder if it's some kind of A/B test that gets switched around every week. Either way, results are poor again right now.
Tuesday was the worst and lowest traffic the site has had in 4 months.
Wednesday was the best/highest the site had had in 6 months.
Today is the best start to a day since launch.
Bizzare. Will no doubt be deindexed tomorow by googe to keep this randomness up. Sitting tight until whatever is ging on settles.
Results seem to have settled a bit for me.
The data set size appears to also be fairly constant.
What's weird is the positions look great, traffic is normal... conversions are the worst I've ever seen. Visitor to sale ratio for today is now well over 100/1. This is normally 15/1.
Perhaps it's market forces and not Google at the moment.
Also it's Thursday night, so perhaps another big shift is on it's way. We will see...
A horrible mess in Google.co.uk today, lots of spammy scrapers and Google's AI certainly does not recognise deliberate keyword phrases created for one thing only such as "buy" "product" "county" and repeated in various orders.
Honestly, why does none of this garbage appear for my widget Bing searches?
April 27th, Google pushed out a Panda refresh. I confirmed it with Google and posted on SELand.
I see new Pagerank numbers in my toolbar. Of course this is the least thing to worry about comparing to problems most of us having cause of Penguin nowdays.
We removed over half of the URLs on our site and 301'd them on Thursday the 26th. These URLs had mostly thin and/or very old content.
Our Google organic traffic started climbing at 3:00pm on Friday the 27th, and has been between 30% and 40% up every day since. I was assuming the lift was related to the Penguin update on the 24th (assumed it just took a few days to fully roll out). No totally ruling that out, but now knowing that Panda 3.6 hit on the 27th, it seems that may be the cause. The only thing I wondering is what % of the changes we made on the 26th Google would recognize by the 27th?