Is Google Weighting Clicks On Serp Entries More Highly In Its Algorithms? Options

[bwelford]

I am getting the distinct feeling based on traffic to my websites that clicks on particular entries in SERPs are being weighted more strongly in the Google algorithms. In other words, suppose you had a web page sitting at #3 position for a given keyword search. If this web page gets more clicks by searchers than the #1 or #2, then this will eventually move your web page to the #1 or #2 position.

If this is so, the importance of the title that is shown and the content of the snippet selected by Google becomes of critical importance.

Has anyone else noticed this?

[ccera]

Barry,
I have been told that this is the case (but not by someone who was in a position to know or had other than anecdotal evidence).
If it's any help, it seems to work that way for me. My sites are all still small enough for me to watch the search phrases pretty carefully, and I can say definitely that traffic for a particular keyword or phrase seems to breed exponentially. It seems that clicks move the page up in the results, bringing more clicks. And this is for sites with virtually no page rank and few IBLs to help them.
CiCi

[bragadocchio]

I'm not sure that the answer is quite as easy as a yes or no.

Is the information that you're seeing about results placements from your searches on the search engine, or from an analysis of log files?

It's possibly not just a matter of clicks upon links, but possibly looking at considerably more user behavior related to those results.

Which page is clicked upon in a result set?
How long does someone stay at that result before choosing another?
How far do they scroll down the page, if they visit?
Whether they return to the search results and refine their query?
If they refine their query, in what manner?
Do they then end up coming back to that page, whether through repeating the search or not?

More complex, might there be some profile type building going on for the Website itself, or for the query, or for the searcher? Is Google collecting and aggregating some kind of user/query/result profile that may rerank results for you, whether logged in, or identified in some other manner, such as a cookie?

I do think that the role of a title/snippet/URL combination (caption) does become more important, but even more so if the page that a searcher lands upon matches what's promised in that caption.

[fisicx]

Here's a thought.

Because the #1 slot is often a sponsored link, are people looking further down the page for organic listing. Are they becoming ad blind in that the top slot is no longer regarded as a 'real' result?

[Respree]

I'd be lying if I said I wasn't thinking the same, Barry. Google seems to like the concept of popularity. Most often, we think of link popularity.

But what about click popularity (yes, I'm making up these names -- somebody has to)?

If you were at #3 and got twice as many clicks as positions #1 or #2, if I were the Google algorthm, I would consider that one metric in determining how popular a page is.

There's two problems, as I see it.

Number 1. I don't know how many clicks the other guys are getting.
Number 2. I have no way of proving or disproving the 'hairs on the back of my neck' theory.

About all one can do is to write a good snippet and title, and hope that your appeals to people better your competitor's.

My two cents.

This post has been edited by Respree: Oct 18 2007, 05:33 PM

[iamlost]

I'm not sure that the answer is quite as easy as a yes or no.

That is a definite fence straddle

I agree with Bill that the 'click' itself is only part of the equation. For instance my analysis shows a significant divergence between a click that 'remains on site' and a click 'that bounces'. As it can be difficult to adjust for varience and volume I only have reasonable confidence with contrasting the 'extremes'.

Furthermore, the divergence shows sooner and greater with long-tail terms. This would seem to indicate 'clicks' as a lower weighted indicator coming into its own when more powerful ones do not supply sufficient differentiation.

Number 1. I don't know how many clicks the other guys are getting.
Number 2. I have no way of proving or disproving the 'hairs on the back of my neck' theory.

1. But you can track your own percentages of 'click type', i.e. stay or bounce back, and SERP. Often long-tail term SERPs tend to remain static so if one term with lots of 'long stay' clicks starts heading up while others with a lower percenatge do not it becomes a reasonable hypothesis. Subject to the first contrary instance of course.
2. Those twitchy hairs are my crystal ball.

About all one can do is to write a good snippet and title, and hope that your appeals to people better your competitor's.

I strongly agree that title and description are much heavier rank hitters than click behaviour - at the moment. I do believe that they are connected. If PPC ad copy is crucial just think how critical the equivalent organic SE marketing. The traffic volume will be driven by the title and snippet; the traffic behaviour will be determined by how well the page fulfills their promise.

[bragadocchio]

I enjoyed these papers (and patent application), and though they don't discuss whether click throughs influence rankings, they do talk about what might make a good snippet:

Microsoft on snippets:

The Influence of Caption Features on Clickthrough Patterns in Web Search (pdf)

Web search engines present lists of captions, comprising title, snippet, and URL, to help users decide which search results to visit. Understanding the influence of features of these captions on Web search behavior may help validate algorithms and guidelines for their improved generation. In this paper we develop a methodology to use clickthrough logs from a commercial search engine to study user behavior when interacting with search result captions. The findings of our study suggest that relatively simple caption features such as the presence of all terms query terms, the readability of the snippet, and the length of the URL shown in the caption, can significantly influence users’ Web search behavior.

Joint paper on Snippets from Yahoo and A9:

Summary Attributes and Perceived Search Quality

We conducted a series of experiments in which surveyed web search users answered questions about the quality of search results on the basis of the result summaries. Summaries shown to different groups of users were editorially constructed so that they differed in only one attribute, such as length. Some attributes had no effect on users' quality judgments, while in other cases, changing an attribute had a "halo effect" which caused seemingly unrelated dimensions of result quality to be rated higher by users.

This patent application from Google has me wondering whether Google is satisfied with the amount of information that snippets present to searchers:

Expanded snippets

[0006] Oftentimes, the search results include three pieces of information, such as a title, a snippet, and a link. The title identifies the corresponding web page. The snippet includes a small portion of the web page that often contains one or more of the search terms of the search query. Typically, the snippet includes no more than a sentence worth of text and might include one or more partial sentences. The link includes the address of the web page.

[0007] Users often use the snippets in determining which search result(s) to select. Because of the short length of the snippet, however, the snippet may not provide enough information for a user to make a meaningful decision regarding which search result(s) to select.

[iamlost]

Thanks for the links...once again a Slawski Headache.

[0049] The expanded snippet may permit the user to access the entire search result document. As shown in FIGS. 7-9, the expanded snippet might include a "view entire document" link that, when selected, may cause the corresponding search result document to be retrieved.

Note: I was unable to view the images to possibly clarify the following concern:

I see four methods of 'retrieving' the document, one that definitely infringes copyright, one that probably infringes copyright, and two that do not.
1. Definite infringement: from a domain that has envoked 'nocache' (1) the page is pulled from the interior indexing cache and displayed or (2) the page is pulled directly from the domain and displayed 'off' the domain, i.e. as per a Google framed image result page.

2. Probable infringement: the snippet 'cached copy' is overdue for an infringement case - 'nocache' not being a requirement of copyright.

3. There is some form of opt-in program or proprietary meta or robots.txt where explicit permision is given.

4. The link takes the visitor to the page on its domain just as if the URL link were clicked.

Given the framed page version (would love to see a copyright challenge here too) used - and sadly generally accepted - for images I would expect a framed version of the document to be the Google preference.

Off Topic

I will continue to use and enforce nocache and to block my images. Except for a few specially selected images and videos to pretty up universal clutter search.

[bragadocchio]

There are some issues with that Google patent application - the main one is, "when does an "expanded snippet" cross the line from fair use to copyright infringement?" It really doesn't explore that.

The other two papers do provide some interesting suggestions on snippet.

[cvos]

Google emphatically denies they use analytics and Search Engine Results Pages click behavior to determine placement, but I don’t believe them for a second.

Google gigabytes worth of historical data on your website and they have a serious data fetish. These guys think that with enough information any problem is solvable.

PageRank is slowly becoming unsustainable through link manipulation, as nearly every HTML jockey knows that links and anchor text affect your ranking. Google needs a PageRank replacement, and I believe that user behavior tracking and data mining are their solution.

See all the ways Google can track you and gather data about your surfing habits:
* Google Toolbar, installed in Firefox browsers by default
* Google Analytics e
* Personal Search History
* Google Desktop
* Adwords - permanently storing your conversion data
* Adsense - all the sites that run contextual ads give google all your data just like Analytics
* Checkout - same as adsense

I'm sure I left out a few of the methods that Google uses to siphon off your statistics to feed the 'bots.

[redcardinal]

Difficult if not impossible to prove this hypothesis. And given that members here are more likely to be actively promoting their own content could it not be that your content that previously ranked number 3 has risen to number one due to factors other than click through?

The one thing I find most difficult with analysing SERPs etc is isolating factors - holding everything else constant is impossible. Not disputing the hypothesis, just saying it's virtually impossible to prove.

[earlpearl]

What an interesting topic. I'm sure that patent research may touch on some of the issues.

How could we know user behavior. We can't see if a visitor clicks on our site off a search query and then clicks on several other sites in the same query. Do they return to our site? Do they traverse through it, either the first or second time?

The only public view on this type of behavior of which I'm aware are the captured data from the AOL leak from 2006. From my limited views of that and some research on topics I saw a good amount of user traveling through a number of sites within a query.....but I did very limited observing.

I've got to read the snippet data. That is so vital. My understanding is that the engines are pulling snippets from meta tags. Is that accurate?

A really great question.

Dave

[bragadocchio]

I've got to read the snippet data. That is so vital. My understanding is that the engines are pulling snippets from meta tags. Is that accurate?

It's one option, but not the only one. A snippet may be pulled from DMOZ in some instances, if a site is listed there.

It's also possible for a page to be relevant for a term that doesn't appear in the meta description, in which case, the search engine will try to provide text from the page itself that shows the term or terms being searched for so that the searcher has a preview of what the page is about.

Think of it this way, the search engine's more concerned with what the searcher sees in the search results than that they are delivering traffic to your site. If the meta description includes the keywords from the search, it might deliver that. If the DMOZ description seems more relevant for the query used to find the page, the searcher might be shown the DMOZ title for the page and description. If none of those are a good fit, it will look to the text on the page.

It might consider a number of factors when looking at on page text, including where the mention of the query terms are located on different parts of the page, whether or not they can deliver a full sentence or not. It's possible that a readability test is done on different sentences that might contain the keyword phrase that the searcher used as a query. Google's patent application on snippets for product search reviews includes the possibility of a grammar check and a readability test of candidate snippets. Those may be possibilities for snippets in web search results, too.

[bragadocchio]

Google emphatically denies they use analytics and Search Engine Results Pages click behavior to determine placement, but I don’t believe them for a second.

Right. Instead of using click throughs, analytics, and other data to directly determine place, use it instead to determing which algorithms are the best ones to use, or to build a statistical model that can gather data about all kinds of activities much beyond mere clicks. Use it to build profiles for pages, for searchers, and for specific queries, so that instead of applying a cookie cutter approach that fails quickly, you have a much more intelligent one that succeeds.

For example, I search for java, and I choose the coffee result. You search for java, and you choose the island result. Someone else searches for java, and picks a programming page. A search engine simply using clickthroughs to rerank results means that the results are really likely to fail, as an average of different user intents conspires against them.

Show me results for other folks like me who tend to pick the same pages when doing the same or similar queries. Show you the results chosen by people like you who tend to pick the same pages when doing the same or similar queries. Same with the programmer I referred to. Assign us profiles (or a few profiles) so that it is anonymous, and so that our interests aren't too pigeon-holed into too small groupings, and create profiles for web pages (and sites), and for queries, and use those profiles to try to match people and web sites and queries that might be similar to folks and pages and queries that they have more data about (from analytics and personal search profiles and other places.)

Clicks and analytics don't determine placement. It's not that simple.

[earlpearl]

I had to look back through notes. I had been unhappy with dmoz language populating my snippets.

the following describes meta tag descriptions that will eliminate the dmoz snippet:

<META NAME="ROBOTS" CONTENT="NOODP"> is a directive for all web crawling bots, while <META NAME="GOOGLEBOT" CONTENT="NOODP"> is a directive for Googlebot only.

If the search query references a page with which doesn't adequately respond to the search via the meta tag references, the snippet will reference language within the page that seems to correspond to the search query.

[amanda123]

Well i think Because the #1 slot is often a sponsored link, people don't like looking further down the page for organic listing. That might be the reason...