Many results for one query - domain clustering returns?

[Saffron]

Mod's note: Members are again reporting seeing many results for one query in Google serps, enough so that I've decided to split this discussion off from our September 2014 Serps...

The following 4 messages were cut out of thread at: http://www.webmasterworld.com/google/4699490.htm [webmasterworld.com]

---

Has anybody else noticed domain clustering? I keep seeing the same site listed multiple times in the top 10 search results. Two, three or four listings all for the same site.
.

[edited by: Robert_Charlton at 4:57 am (utc) on Sep 5, 2014]

[Robert Charlton]

I keep seeing the same site listed multiple times in the top 10 search results.

I've seen a return of domain clustering, but only for very long tail queries. It's likely either that Google is testing something new in the algorithm, or has decided to refine results on extremely rare queries. Can you characterize what you're seeing, and what type of query, without going into specifics.

One way to get a sense of how rare the target phrases are (and, by implication, how rarely searched they might be), would be to put the entire query in quotes.

In some cases, if the query is, say, a phrase plus a placename, a more fruitful approach would be to put the phrase in quotes and keep the placename separate, as in this syntax...

["red widgets" placename]

You can also use keyword suggestion tools, like SEMrush or SpyFu to get some idea of search frequency.

[edited by: Robert_Charlton at 3:51 am (utc) on Sep 5, 2014]

[Saffron]

I'm searching for simple phrases 'broken widget', not long tail queries, but more specific. The phrases are very, very common.

Okay, another example (and this is absolutely nothing to do with what I'm searching for, if it's too specific please delete). Say something like 'oil leak in car' or 'engine not starting'...again, examples, this is not what I'm searching for myself. But all common, short phrases.

The same site over and over again. It's unbelievable how much the dial seems to have been turned up on this one site. I get that they are number one, but it really hurts other sites when one site gets so many listings.

Wow, the site I'm referring to has quadrupled in SE traffic since last October according to semrush. That's huge.

[Robert Charlton]

[Oil leak in car] is four words, conceivably long tail.

If a simple two-word phrase, why don't you put it in quotes and search and then let us know how many pages Google returns for the query.

The other possibility is that Google might now be rewriting these queries to a more generic form (note Hummingbird and recent AdWords "fuzzing"), and there might be sites dominating this generic form in the serps so strongly that Google is using host or domain clustering (not host crowding) to expose the top pages and get a more exact sense of which ones users prefer.

Google did this a while back with the "m*ac and ch*ese recipes" which we discussed at length, and ultimately these serps did return to normal in a matter of months. See our discussion here....

Many results from one site - Host Crowding vs Brand Authority
June, 2012 - Aug, 2012
http://www.webmasterworld.com/google/4464096.htm [webmasterworld.com]

[Saffron]

Most of them were two words, but that kind of general search term (it's hard to explain). Very, very commonly searched topics.

I put a two word phrase into Google in quotes and there were 30 pages of results. With this particular site having FOUR listings on page one.

I'll go read the other thread now.

[Robert Charlton]

Possibly also that Google has changed the nature of exact results. This is complete conjecture... it's something I haven't checked, so I don't know for sure.... But it's something I thought might happen after the announcement discussed in this thread regarding disappearance of exact match in AdWords....

Google AdWords Going Close Variant Keyword Matching: Exact Match No Longer Possible
http://www.webmasterworld.com/google_adwords/4696348.htm[/4696348.htm [webmasterworld.com]

[Robert Charlton]

PS: What was your default setting per page, and was the query in quotes? Try ten results per page. Also, 30 serps pages of results is not a great number. That might just be 3000 results returned.

What is the number of results that Google says it's returning?

[Saffron]

Oh sorry, I was just clicking on the bottom links 1, 2, 3 etc. Just searched again and at the top of the Google results it says there's 5,240,000 results for the phrase.

For that particular phrase, it would be hard to get variants. Easy to spell, you could change the words around from say 'oil change' to 'change oil' and a few longer phrases such as 'how to change oil', but all pretty similar.

My page shows 10 results per page.

[Robert Charlton]

Saffron, are you using "double" quotation marks in your searches, as opposed to 'single' quotation marks? It can make a huge difference.

[Saffron]

Just single, I'll try double.

1) No quote marks - 5,220,000 results - that site gets 4 listings on page 1.

2) Single quote marks - 5,240,000 results - same site gets 4 listings 1.

3) Double quote marks - 57,700 results - same site gets 2 listings on page 1.

1 and 2 were pretty much the same results.

[Wilburforce]

My own site has been returning two pages in adjacent positions (Home page and Blue Widgets page) for Blue Widgets for a while now, and - this has happened before - whenever I see this I invariably see domain clustering discussions here.

This has always resolved in a matter of weeks, sometimes with one page remaining, sometimes the other, and always sticking with the selected page until the next time this comes around.

I think it unlikely that Google is looking at that specific query. My suspicion is that there is a "multiple results from same site" filter that usually excludes all but a single page for any query, and that periodically - for whatever reason - Google turns the filter off.

FWIW, the same two pages have both been on Bing page 1 for years.

[Saffron]

I noticed it quite a lot last year, particularly after the January 2013 update, along with lots of 'general' sites such as wikihow. Then things improved in September, but the past few days it's reverted back for me.

Having said that, I've only searched within my niche, I haven't had time to do searches on other stuff. Maybe it's just the searches I'm doing.

[Wilburforce]

Maybe it's just the searches I'm doing.

No, it is more widespread than that. My impression (as previously posted) is that it is a global filter.

Generally, one would expect e.g. a clutch repair specialist's website to have more than one page (the Home page, the Clutch Repair page, the Maserati Clutch repair page...) with similar ranking factors and authority for Clutch Repair, so it seems more likely to me that a filter is usually applied than that "extra" pages are being included now.

[Robert Charlton]

Note that I've added a question mark to the title of this thread, as I'm not convinced by examples I've seen.

I asked Saffron to send me info on the phrase she used as a test and what she saw in terms of clustering. She's posted above as I requested, without mentioning the phrase. I'll exemplify it and call "widget testing".

She and I are seeing more or less the same number of results returned, but I (in the US) am not seeing any site getting more than two organic listings for the search she tried. She reports four. I believe she's in Australia.

I'm seeing each of the two sites she mentioned getting only two organic results on page one. One of those is also running AdWords, so there are three listings for that site on the page. Serps of course can vary by location, time of day, personalization, etc... but two results are essentially the old standard maximum.

Two, by the way, is what "host crowding" actually is... ie, Google had decided a while back to limit the number of organic rankings to two per hostname. So when we see four or five results per page (on a default of ten listings per page), that is not "host crowding". That is the opposite of host crowding... "host uncrowding" you might call it. I think that Saffron's "domain clustering" term to describe it is a good one.

As we can see from Saffron's examples, Google's syntax is for double quotes. Single 'quote' marks don't affect whether Google sees a phrase as an exact quote, whereas double "quotes" do.

Worth noting, as it's interesting how subjective we can get about searches in our own niches... Saffron's example "widget testing" is not remotely comparable to "oil change'.

I'm seeing "widget testing" in quotes returning 57,600, essentially the same as Saffron's 57,700. It's not what I'd call long tail, but it's not very competitive and therefore not very frequently searched. I can imagine a site with a lot of authority dominating a serp for such a query until Google gets enough searcher data to decide which of the displayed pages should continue to be returned.

Without quotes, though, the number of results up around 6-million... so yes, the quotes do make a big difference. With quotes, Google returns (and reports) the number of pages that match the exact phrase. Without quotes, it returns and reports the number of pages that contain both words in the phrase. For me, quoted searches are a very rough indicator of competition.

More precise for SEO is the allintitle: operator. For this query, the syntax is...

allintitle: widget testing

...and that search returns 8,200 results, suggesting the number of pages specifically targeting those words by using them in the title.

By comparison, "oil change", which Saffron mentioned as comparable, though it may superficially resemble "widget testing" (ie, it's two words), is wa-a-ay different... "Oil change" is much more commonly used... returning about 5,400,000 results when searched in quotes. When searched with allintitle: without quotes, it returns 1,370,000 results... so it's significantly more common and more competitive.

A long tail search I've been watching, in comparison, is four words, which, with no quotation marks, returning 51,000 results. One site is holding the top 4 positions for that search, and has been for about six weeks, and it's the site in its niche I'd expect to do so.

I've been using it in a way as a niche barometer... and though it's a very plausible query, there's a question of whether the query is ever searched. It shows zero searches on SEMrush, which I would expect to be low. I've been intermittently watching the serps without clicking on them, assuming that Google is displaying those four results because they're the top results (in terms of relevance and domain authority) for the query.

As I've observed over time with other domain clusters that dominate searches, little by little, as Google assembles user data, the number of pages returned for a site is reduced. The reduction first occurs on page one, and then migrates deeper and deeper as enough searchers dig down and data accumulates.

I should add regarding the "m*c and ch*ese recipes" query noted above that the number of pages returned for the top domains has gotten honed down over time, until most of the top sites display only two pages. Very recently, on page one, I've noticed one site move up to three pages again.

I do think there's something new afoot and there's some testing going on, but I'm certainly not sure about it. Changes in the way queries are parsed, if that's what's going on, may result in a new round of sampling. Since statistical data takes time to accumulate, frequently searched queries are going to settle down much faster than longer tail will .

Re Wilburforce's observations, I do believe that Google has chosen times to return all sorts of extra pages... sitelinks, mega sitelinks, etc, to test various aspects of the algo. These are my theories, anyway.

This is more than I intended to write on this... but there are frequent questions popping up about "host crowding" (which is the wrong term), and I thought it might be worth discussion.

[mrengine]

Without quotes, search for: <materialmodifier material perforated widgets>

Listings 4, 5, 6, 7, and 8 all belong to the same company when searching from my USA based IP.

I don't produce or sell <perforated widgets>, but I could post many other examples if needed.

Regardless of how competitive a query is, crowding limits competition and consumer choice.

The question is not has crowding come back, but why. I suspect when Google restricts diversity in the search results, more businesses are inclined to increase their adwords spending.


Mod's note: Exemplified specifics.
No actual queries, please, unless obviously general examples, which these were not.
.

[edited by: Robert_Charlton at 5:04 pm (utc) on Sep 5, 2014]

[Catalyst]

I've been seeing it a lot in the local 7 pack in US so assumed it was Pigeon related.

Typically in the past a company would only get one listing in the pack no matter how many listings they had. (For instance a practice with 7 Drs.) Now I'm seeing pack monopoly and that's never happened before.

Linda Buquet

[bwnbwn]

highly competitive phrase

<snip>

...the 1st 4 are from same site in both searches.

.

[edited by: Robert_Charlton at 9:55 pm (utc) on Sep 5, 2014]
[edit reason] no specific searches, please [/edit]

[Saffron]

When I use quotes, I see FIVE results for the same domain

When I don't use quotes, I THREE results for the same domain.

[Planet13]

Seeing the same thing as bwnbwn.

Oddly enough, sites like wikipedia, about, and amazon are NOT ranking in the top three.

[Robert Charlton]

OK, I'm seeing this at many levels of competition.

(Please, for the protection of all members here, let's avoid posting specific searches.)

The results are likely to persist longer, though, in less frequently searched queries. What seems to be the common factor is that the ranking domains carry high authority for their niche, and the ranking pages on a given domain can appear to be pretty much the same.

I see it as a data sorting problem. If a site has a large number of pages that Google thinks make searchers happier than competing pages, does Google still only display two pages, or does it display a larger sample until it's evaluated searcher behavior within these pages displayed competitively?

Matt Cutts discussed some of the considerations in the video that Brett mentions in his Host Crowding vs Brand Authority post noted above....

How does Google decide when to display multiple results from the same website?
Matt Cutts - June 11, 2012
trt 5:40
http://www.youtube.com/watch?v=AGpEdyIcZcU [youtube.com]

[Wilburforce]

An interesting video, but principally it discusses the objectives, not the mechanics.

From what MC says there are clearly algorithm elements that both filter and cluster multiple same-site results.

Filtering is both on page-count (the higher the number, the less likely it is that any further pages will be displayed), and on authority and relevance (the higher the authority and/or relevance, the more likely it is that more than a single page will be displayed).

As MC says "we've done a bunch of different experiments" we can probably safely assume periodical testing of any of these factors, together with layout (e.g. indentation), so what we are seeing in the current results is probaby not of any particular long-term significance.

[blend27]

I just searched for a popular niche term that one of my friend`s site use to rank at #1 for the past decade. It was a completely healthy niche up until a few month ago. 2 words search.

ebay came at #1.
facebook shared about 4 of the rest of 94.

THE REST was Amazon.

I am not kidding here. It is mind blowing.

No Quotes, not signed in, no cookies.

I have never seen something like this before. Ever.

---added--
One thing I'd like to point out here is that most of pages that are ranked at Amazon contain low quality product images. These images are wholesaler images, Something that retailer would get in return for buying a small amount of widgets. Almost like OCR is a factor on rankings. Several resellers thru amazon using the same JPEGS. Sites that use their own imaginary are no where to be found in GOOG SERP, well maybe at 400 +.

MODS, I've been at this forum for almost a decade. Feel free to delete the next line.

Query term is Multic*l*r Amb*r.

.
Mod's note: Obscured several letters so words wouldn't distort other search results.
.

[edited by: Robert_Charlton at 6:05 am (utc) on Sep 9, 2014]

[Leosghost]

@blend27

Searching same "term" ..
Seeing exactly* the same amazon thing as you from France....*using 10 results per page..
#1 result ebay..then amazon all the way ( #2 to #8 )

then a block of 4 images..of which again one is from amazon ..lastly one result #9 for etsy and one # 10 for facebook..

Other sites only show up when going to page two..

Have seen this before..about 2 years ago...running searches on many different terms KW1 and KW1+KW2
informational and commercial ..

..At that time I was checking in 100 results per page SERP..
YouTube , ehow , ebay and amazon were at least 90 "slots" out of every 100 of each SERP..

Informational searches got more YouTube into the mix ( sometimes all the first 30 to 40 places )..commercial searches gave less YouTube and more ebay or amazon ..

This was happening only if all personalisation was "turned off"..as if these were the true "underlying results" that were then to be filtered and made more "varied" by personalisation..

[Robert Charlton]

Multic*l*r Amb*r

blend27 - OK, this can be our "m*c and ch*ese recipes" example for this discussion, a term which we can monitor for a while.

There is a difference, though, which has prompted me to obscure several vowels... in a way that I'm thinking anybody here should be able to guess... but which Google most likely won't be able to parse, or at least not for purposes of indexing and ranking. I've replaced some vowels by asterisks.

My concern is that this phrase is relatively rare compared to the recipe search we were looking at in 2012, and I don't want our use of it to distort results.

Here's the competition level, with no searches indicated on SEMrush....
- all the words (default) - 2,180,000 results
- quoted search - 45,200 results
- allintitle - 6,320

Regarding the domain clustering you're seeing... for this unquoted search, at the default 10 results per page, I'm not seeing what you and Leosghost are seeing. What I'm seeing is not extremely unusual...

- eBay #1
- Amazon #2, #3, and #4
- images following #4
- Facebook #7 and #8

The next Amazon result doesn't show up until the top of page 3. I should add too that the quality of images I'm seeing aren't all that terrible, and there's even a magnifying window you can move over the image. (The merchandise itself isn't all that inspiring, but that's something else). The pages are all fairly heavily reviewed, and the 4 Amazon pages I looked at are all selling different items from obviously the same manufacturer.

What you and Leosghost are describing, though, was typical of what I saw on some wholesaler type results that appeared on Amazon during the 2012 host authority days, but I should add that using 100 results per page is a clustering mechanism in itself, and greatly distorts what most users are likely to see.

Eventually, at 10 results per page, the pages from a single domain did get pared back to 2 per page, as Google made choices among the surfaced pages (starting, as I mentioned, with the first page, and gradually working back to deeper results).

I am guessing that we may be distorting the results on this particular query just by visiting the pages, but I'm not sure of that. On the query I describe that I've been monitoring, I'm careful not to click through. I'm wanting to see indications of traffic that's not related to me.

Regarding these results, I'm assuming that there's some kind of new algo factor that's needing to be calibrated, and I'm wondering what it is. Another possibility could be that this type of sorting needs to be run periodically on different types of pages.

[Saffron]

I'm seeing the same thing.

Edit:

1) Ebay
2) Amazon
3) Amazon
4) Amazon
5) Amazon
6) Amazon
7) Amazon
8) Amazon
Images
9) Etsy
10) Facebook

Wow, so much diversity ;)

[Robert Charlton]

Hmmm... In the past, when Google frequently used extra results in the serps to test (and by extra I include things like Sitelinks, inline Sitelinks, mega site links, etc) geo localization played a big part.

The results were in fact "hyperlocal". I remember on one search that I could change the results by changing Location to different communities going down the freeway. Results would change almost from exit to exit.

I just tried changing my location from the SF Bay Area to New York City, and my results on this search don't change.

[Wilburforce]

I'm seeing the same thing.

Me too.

[GreyBeard123]

I see...
1) ebay
2) amazon
Images
3) amazon
4) facebook
5) facebook
7) facebook
7) .org site
8-10) .com sites

[Saffron]

When I did the search, G automatically changed it to the US spelling of the second word. When I then told it to use my spelling (how we spell it in Australia/UK) I did notice better results.

[blend27]

@Leosghost,

That is exactly what I see here. I am in US now. I dont allow cookies from Goog or its properties being set. So as one would think there is no "Personalization" in a way.


And then...

The images that are displayed in regular SERP are Spot On. The Ads that are displayed on top and right side of the SERP are spot on as well - 100%. The SERP it self is nothing but rubbish Amazon(97%), Ebay, Facebook.

I did some additional searches by replacing the first word in a query with other popular c*l*rs. Pretty much same thing.

I know this niche very well, I know the ecom and educational(some very well known edu pages) sites on the subject that should rank for this query. There is no SEO trickery on them.

Very Sad.

As far as the original query term, just to make it clear: it is specific to this niche. Millions of people search for it when looking for these widgets.