Search Engine Click-Through & Bounce Rates as Ranking Metrics

[Webnauts]

This thread is not for Neophyte SEOs. So think twice before posting, to avoid spoiling the thread with nonsense or irrelevant posts which will be removed without prior notice. Thanks.

I would like to start a discussion about two topics, which are fully related:

• a. Click-Through-Rate (CTR);
• b. Bounce Rate.

a. Click-Through-Rate (CTR)

Document Scoring Based on Query Analysis (US Patent Application 20120016874)
This patent application looks at the frequency of selection of specific pages for specific queries.
How frequently a certain page is selected in search results over one period of time might be compared to how frequently that page might be selected in search results over a later period.
If the page is selected less, it might be lowered in search results. If it is selected more frequently, it might be increased in rankings.

Source: http://appft.uspto.gov/netacgi/nph-P...DN/20120016874

--- and ---

b. Bounce Rate

Web Search Result Summarization: Title Selection Algorithms and User Satisfaction
Metrics such as click through rates on search results could be set by the search engines looking at their query logs to tell whether a page was clicked or not within their search results for certain queries, and which ones were clicked. Bounce Rate of being a ranking signal, it is if someone would click on your link in the search results and then they would return back and repeat their query and click on a link of an another web site. For sure there may be more bounce rate metrics search engines can use, like recording user behaviour through their personalized search, IP, cookies, etc.

Source: http://www.kanungo.com/pubs/cikm09-w...marization.pdf

---
I picked up the valuable information from the site of Bill Slawski (thanks Bill), and later studying the patents myself.

So should be dig into the above? I think this thread can turned out to a very fruitful discussion.

[bill slawski]

Great topics. Thanks for the mention, John.

I was really impressed with the approach in the kanungo paper, and I think it's important that while the paper was created at Yahoo, Kanungo is now at Bing, and chances are that Microsoft would incorporate an approach like the one described in the paper.

The Google patent application is one of a number of continuation patents sharing the same name, and each of them have different claims sections focusing upon different ways of analyzing queries and in a few cases how people might interact with search results, such as frequency of page selection, or how long people might spend on pages selected within search results.

The page selection approach isn't going to determine whether or not those pages rank highly for a result in terms of relevance/importance/popularity, but they do have the potential to boost pages or reduce them in rankings.

One approach that I've been trying to take involving click selection, is using the Google Webmaster Tools query information on impressions and clicks to identify pages with high impression levels, but click levels that aren't as high to target as pages where meta descriptions and/or titles might need to be changed to make them more engaging/persuasive to try to attract more clicks.

[Webnauts]

Great topics. Thanks for the mention, John.

I thank you Bill, and welcome to our quiet community. :)

I was really impressed with the approach in the kanungo paper, and I think it's important that while the paper was created at Yahoo, Kanungo is now at Bing, and chances are that Microsoft would incorporate an approach like the one described in the paper.

That is interesting! I did not know that Kanungo is now Bing.

The Google patent application is one of a number of continuation patents sharing the same name, and each of them have different claims sections focusing upon different ways of analyzing queries and in a few cases how people might interact with search results, such as frequency of page selection, or how long people might spend on pages selected within search results.

The page selection approach isn't going to determine whether or not those pages rank highly for a result in terms of relevance/importance/popularity, but they do have the potential to boost pages or reduce them in rankings.

One approach that I've been trying to take involving click selection, is using the Google Webmaster Tools query information on impressions and clicks to identify pages with high impression levels, but click levels that aren't as high to target as pages where meta descriptions and/or titles might need to be changed to make them more engaging/persuasive to try to attract more clicks.

On this topic I recently posted a real case at WPW forums about that: http://www.webproworld.com/webmaster...l=1#post601427

What do you think about that approach?

[Doc]

One approach that I've been trying to take involving click selection, is using the Google Webmaster Tools query information on impressions and clicks to identify pages with high impression levels, but click levels that aren't as high to target as pages where meta descriptions and/or titles might need to be changed to make them more engaging/persuasive to try to attract more clicks.

Hi, Bill- a pleasure to see you here.

High impressions/low CTR is certainly a good indicator of titles and snippets that need some luv. I just get frustrated at the sometimes whimsical selection of snippets. For instance, logged in or out, I can see different snippets for the same query. I don't understand why personalized search would affect snippet selection.

I realize that's slightly off the thread topic... it just makes it difficult to judge the effectiveness of the meta data. I assume it may skew it for the algo, too.

[Webnauts]

Doc, I do not think you are going off-topic at all. Rich snippets are also playing a significant role in the CTR factors. I will bring up some interesting stuff about those in the thread as it goes.

[sithburns]

I can see this helping seasonal ecommerce sites. For instance weight loss products get far more activity in January and February due to people making new years resolutions to lose weight. While the site may have gotten decent click through and decent bounce rate throughout the year, as people's interest and purchasing likeliness is heightened during this specific time period, the site could theoretically get a nice boost as their metrics notice an increase.

Or am I being too hopeful on the potential positive outcomes from this patent in particular?

[bill slawski]

I just get frustrated at the sometimes whimsical selection of snippets. For instance, logged in or out, I can see different snippets for the same query. I don't understand why personalized search would affect snippet selection.

Thanks, Doc. My pleasure.

In addition to showing potentially different snippets for personalized search results, Google is also presenting expanded snippets taken from text within different HTML elements on pages, when it might think those are more appropriate. In the past, Google would sometimes show snippets from content on pages, but those would be reasonably lengthed. Now, they are more frequently showing longer snippets. I've written a little about how they might be doing that, but I haven't seen anyone write about many of the characteristics of those longer snippets. It's definitely worth exploring, especially as John noted in the opening post that clickthroughs can influence placement.

[bill slawski]

Hi Sithburns,

I spent an hour or two a couple of days ago looking at impressions and clicks in Google Webmaster Tools on a fairly large site that's segregated into multiple parts that has a great amount of seasonal-based traffic, looking at decreases in rankings for Christmas related terms after the holiday, and increases in rankings for Valentines related terms now, and those increases and decreases look very much like they might be attributable to clicks and/or dwell time. They seem to fall in line with the document scoring patent filings.

[Webnauts]

Bill, did you think about the possibility that search engines could log a site's frequency of selection and bounce rates for example of Christmas 2012, and re-use that as an additional information to rank the site next year Christmas 2013?

[Blogvertize]

Hi Guys,

well as much as it pains me to agree with the likes of the Google God, I must admit I can see big advantages to the use of click through rate as a ranking factor.
On more than one occasion after building a new website & succeeding in getting it ranked pretty much at the top of page 1, I then discover a whole herd of wannabees that will use the exact same title as I had used on my sites!

One in particular I recall was a site based in Pakistan, the owner adopted my title which was at the time "My Blog Review" & ok he wasn't ranked above me, but nevertheless he was on the top half of page one with Google. On checking his site it was found that he was pushing those conterfeit Ugg Boots we had all heard about. Now when you only get one site copy your title it's no big deal, quality will eventually shine through, but I had the same thing happen with Diy Computer Repairs and to the extent that I eventually changed the title of my own site in order for it to stand out more. On this particular occasion the entire first page of Google was literally full of every conceivable variation on Diy Computer Repairs.

So yes I for one would happily accept more focus on titles & snippits and hopefully this would help weed out those people trying to ride the crest of somebody elses wave, so to speak !

Now on bounce rate I'm afraid my experience over the last 3 years has shown that bounce rate is really a worthless signal for the quality, usefullness & how engaging a web page might be, especially with the recent increase in the smartphone market!

Example 1. I spent an entire day last week bouncing from one site to another & back to Google to modify my search terms, but there was not a single website that I felt had let me down or that I found unhelpful or useless. What I was searching for was information on the "Time To First Byte", every site I visited had their own take on how best to tackle this problem and so as I bounced from one site to another I gradually scribbled down an A4 sheet of useful tips & info. Now I have no doubt at all that the bounce rate for these sites is probably somewhere in the region of 80%, just as is the case for My Blog Review (my own site) & this I think is the price you pay for running an information site. People hit my site looking for one particular topic, just as I was a few days ago with "Time To First Byte" & if they are anything like me then when they are locked into trying to sort out a problem, they become oblivious to all else, nevermind how good the articles are!

Example 2. Not everyones fav topic but let's just look at Internet Marketing for a moment. I have over the last 12 months built several sites to try my hand at Internet Marketing (having a crappy pension means I could use the extra bucks!). Now through experience with these sites I have found that the vast majority of my visitors are using such search terms as "Panasonic Lumix DMC-S3 14.1 MP Digital Camera with 4x..." without the quotes obviously. Now I don't believe these people have memorized that entire search term (call me Mr Skeptical if you must), I believe it is far more likely they have picked up this title from either an online magazine or perhaps a review site and are now in search of the cheapest price for this item. So how long do you think it would take to hit my site on that particular search, check the price & then click the buy button, which then sends them off to Amazon..? Well it takes roughly 7 seconds!!!

Now yes my bounce rate is terrible for this very reason & when you also factor in the mobile market, where people are much less inclined to read a full page of text, it is easy to see that bounce rate cannot be a true guide to a websites value to visitors. I'm sure my many customers would agree that they picked up some great bargains and therefore the site in question must have value to them.

So I would give a thumbs up to Titles & Click Through, but a big thumbs down on using Bounce Rate, as there are simply to many variables in play too make this a trusted metric.

oops sorry about the length !