SEO Articles
Distribution of Clicks on Google’s SERPs
October 26th, 2006
What is the distribution of clicks on a search engine results page? What percentage of clicks gets each search result according to its rank? How much more users’ attention gets the first listing compared to the second? Or how often do users click the listing below the page fold? The way users interact with SERPs is one of the most frequently discussed topics in the SEO community and is also a very important field of study for the search engine specialists. To answer the above questions researchers employ the so-called eye tracking experiments.
Eye-Tracking Studies
The objective of eye tracking studies is gaining insight into how users browse the presented abstracts and select links to click. The results of eye tracking research provide Internet marketers with information on clickthrough rates, thus allowing them to make correct predictions on traffic changes as their rankings are gained or lost. For SE engineers the results provide a basis for improving the interfaces of search engines and metrics to evaluate the relevancy of the presented search results.
To detect users’ interaction patterns the eye tracking experiment observes a number of indicators of ocular behavior using a CCD (charged couple device) camera similar to the appliance used to read bar codes. The indices of ocular behavior include eye fixations, saccades, scan paths and pupil dilation. Eye fixations are defined as a stable gaze lasting for 200-300 milliseconds representing visual attention to a specific area of a SERP. Pupil dilations or pupil diameter changes represent a measurement of interest in a particular listing. This variable is especially important as it helps interpreting an implicit user feedback to the relevancy of the presented search results.
Cornell University Eye-Tracking Analysis of SE Users’ Behavior
One of the most recent eye tracking studies was performed at Cornell University by Laura A. Granka, Thorsten Joachims and Geri Cay ([1]). They used a sample of undergraduate students instructed to perform search in Google for 397 queries o topics covering movies, travel, music, politics, local and trivia. This study has produced the following results.
Fig 1. Google SEPR Click and Attention distribution ‘heat-map’
Study Results: Clicks and Attention Distribution
As you can see from the graph below and a SERP ‘heat-map’ based on it, the first two listings capture over a half of the user’s attention in terms of time of the eye fixation. Whereas the attention is shared almost equally, the difference in number of click between the first two listings is much more surprising: over four times! After the second listing the eye fixation drops sharply. Search results number 6 to 10 receive roughly equal attention. Here an interesting thing is that the 7th listing gets less attention than the succeeding 8th – apparently here we can observe the effect of the page fold. The 7th listing is just below the screen edge and is often skipped as users scroll the page down to the bottom (during the study the 7th listing was clicked only once). On the graph you can also see the 11th listing from the second page of the search results. It gets only about 1 percent of clicks and user attention – 2.5 times less than the lowest ranked result on the page one.
Fig 2. Time spent on viewing each results compared to the number of clicks. Source [1]
Often people consider getting to the ‘top-ten’ of Google as a measurement of the SEO success. Evidently this is a rather rough approximation. The ‘top-ten’ itself is a very diverse group with the number of clicks increasing almost logarithmically as your rank grows. For instance, the first five positions get over 88% of the traffic, and the first three – 79%.
SERP Browsing Patterns
Another important result of this study is the discovery of the browsing pattern: the way people read a SEPR. To assess the performance of the search algorithm it is vital to know how users evaluate the presented abstracts before clicking one of them. For example, if a user clicks the third listing, did he look the abstracts above and below it? The following figure shows how many results above and below of the selected listing are scanned on average.
Fig.3 Number of results scanned above and below the selected abstract. Source [1]
The effect of the page fold is clearly demonstrated here as well. While the first 5 listings are clicked after browsing through 1 to 2.68 listings above and below, the 7th listing is clicked after the entire page is examined! The listings below the page fold (8-10) are clicked after the first five or four listings are scanned. You can also see that the number of listings scanned above the clicked result is much bigger than the number of listings below. This indicates that users browse the list from top to bottom.
To Sum Up
While the study deals only with the first page of the organic search results, it can be assumed that similar results can be produced for other pages and perhaps even for the list of the paid ads in the right sidebar.
In addition to the academic researches there is a number of companies producing eye-tracking studies for the commercial use. The most notable of them are Eyetools.com and Poynterextra (http://www.poynterextra.org/EYETRACK2004/index.htm)
References:
1. Laura A. Granka, Thorsten Joachims, Geri Gay. ‘Eye-tracking analysis of user behavior in WWW search’, SIGIR, 2004. Available at http://www.cs.cornell.edu/People/tj/publications/granka_etal_04a.pdf Retrieved on 26.10.06
Did you like it? Was it useful? Bookmark or share this post:
Subscribe for the Feed
Digg This!
Technorati
Del.icio.us
Furl
Blinklist
Ma.gnolia
Yahoo! My Web
45 Responses to “Distribution of Clicks on Google’s SERPs”
October 29th, 2006 at 4:54 pm
Interesting. I have some doubts about the validity of the study though. The testers were apparently provided with predetermined search terms. They had no intention of their own to find anything in particular. So they basically skimmed the results top to bottom with less interest the further down they went. That’s not searching!
When I search, my eyes rest on those entries that might match my intentions because I read not just the title but also some of the digest to see whether the link is worth following. These people had no intentions so all links were equally irrelevant for them.
October 29th, 2006 at 5:01 pm
The validity of the research suffers because of the relatively small sample - only 26 participants. The patterns of search were quite natural. All the participants claimed they regularly use Google, and it was their primary search engine. Probably the search terms were predefined, but I don’t see how this could damage the validity of the study. When I am looking for a subject I usually use no more than 3 keyphrases myself, as I am (as any average web user) able to come up with a keyword/phrase relevant to my search subject in the early stages of the search.
October 30th, 2006 at 2:24 am
The point is that in a real life situation, you have something in mind that you’re looking for. Then you translate that into keywords but the keywords do not express the whole context of what you want to find. So when the result page comes up, what you do is to match the entries to your original intent in its entire context. When someone else tells you to enter particular keywords, you don’t have the same situation. You don’t have the keywords on the one hand and your desire to find something that belongs to a context on the other. You have just the keywords. And that, I suspect, leads to a very different way of reading the result page.
What they should have done is ask the testers to use google to answer questions like “Which world power occupied the land that is today called Iraq during World War I?”. The testers would surely have received results about Iraq during WWI mixed with entries about the current Iraq war. But they would’ve filtered out the ones that were clearly not about historical matters. They would’ve preferred the ones from sources like Wikipedia to news items from TV stations because they knew that Wikipedia entries are more systematic and might mention occupying forces. They would’ve focused on entries that mentioned other powers like the British, the Turkish, etc.
What they would certainly not have done is to scan the page indiscriminately from top to bottom with decreasing attention.
October 31st, 2006 at 4:21 am
[…] Here is the original Cornell paper via SeoResearcher. […]
November 1st, 2006 at 11:16 am
[…] Here is the original Cornell paper via SeoResearcher. […]
November 8th, 2006 at 11:32 pm
[…] From Seoresearcher, an interesting and informative SERPS click tracking study. […]
November 16th, 2006 at 5:02 pm
[…] za Distribution of Clicks on Google’s SERPs Narzdzia SEO i SEM SEO GoogleNarzdzia SEO i SEM SEO Google […]
November 22nd, 2006 at 5:41 am
[…] Læs mere om rapporten her […]
December 4th, 2006 at 12:33 pm
[…] However, since I’m pretty sure that you’re not trying to rank for “yellow puma shoes” and your goals might be a little different than mine, the best advice I can give is: run a search for what you want to rank for, compare what you see, and think about what you would most likely click on as a human being if you had the attention span of 200-300 milliseconds. […]
December 15th, 2006 at 5:09 pm
[…] Wiadomość o badaniu i rysunki zostały wzięte ze strony : Distribution of Clicks on Google’s SERPs. Całość badania można przestudiować w materiale: http://www.cs.cornell.edu/People/tj/publications/granka_etal_04a.pdf. […]
December 17th, 2006 at 3:00 am
[…] What is the distribution of clicks on a search engine results page? What percentage of clicks gets each search result according to its rank? How much more users? attention gets the first listing compared to the second? Or how often do users click the listing below […]Read full entry […]
December 30th, 2006 at 6:20 am
[…] To put it bluntly, if you are not on the first page then you are nowhere. To make matters worse, the difference between being in the first position of a search engine results page (SERP) and last place is huge. A recent eye-tracking study by the Cornell University showed that the first result on a google results page recieves 56.36% of clicks compared to last result which only recieves only 2.55% of clicks. […]
January 2nd, 2007 at 2:23 pm
How did you come up with the percentage of clicks you used in Fig 1? When I reviewed the PDF in footnote 1, I did not see any specific data about how many people clicked each link.
Did you just estimate the actual number of clicks based on the gray bars in figure 2? If so, I think I have reverse-engineered the numbers you used:
Rank Clicks % of Clicks
1 155 56.36%
2 37 13.45%
3 27 9.82%
4 11 4.00%
5 13 4.73%
6 9 3.27%
7 1 0.36%
8 8 2.91%
9 4 1.45%
10 7 2.55%
11 3 1.09%
Total 275 100.00%
Interestingly, these numbers indicate that there are only 275 clicks in these results, although the study says there were 397 queries. Does that mean the other 122 queries did not result in any clicks? Or were they clicks on results 12 or higher, so they didn’t show up in the chart?
I’m curious to know more about what happens on the 2nd and 3rd pages of search results. This data implies (by the inclusion of result 11 in fig. 2) that many fewer people saw or clicked on that second page of results. Does that mean that results 12 and up had even fewer clicks than 11? Does the distribution of clicks on page 2 mirror that of page 1, or is it a flatter curve?
I’d love to hear about any follow-up research.
-James Grove
January 2nd, 2007 at 3:45 pm
Dear James,
yes you’re right, I estimated the percentages from fig. 2 exactly like you did. I was myself surprised with the missing clicks and it was possible that some of those clicks were made on the 2nd and 3rd pages but were not reported in the research results. Also some of the clicks probably were just not counted.
Obviously the authors of the research could not calculate a valid distribution of clicks for the 2nd and 3rd pages because of the low number of clicks done in those areas during the experiment. A much bigger number of participants would be required for that.
I believe there are more eye-tracking studies including those ones covering more than the 1st page of search results. I used this one because it was ‘approved’ by Google - I found it among the ‘Papers by Googlers’ at Google Labs: labs.google.com/papers.html
February 4th, 2007 at 5:16 pm
[…] The blog Online Marketing Research posted an article called “Distribution of Clicks on Google’s SERPs,” which discusses eye tracking behavior on Google’s search engine results pages. There is an interesting visualization of a “heat map” that shows how users view these search results relative to their position on the page. This is a convincing argument to sell a company search engine optimization services in order to get them up into the “red zone.” There are also some other great visualizations of user behaviors, as well. […]
February 4th, 2007 at 6:28 pm
[…] The blog Online Marketing Research posted an article called “Distribution of Clicks on Google’s SERPs,” which discusses eye tracking behavior on Google’s search engine results pages. There is an interesting visualization of a “heat map” that shows how users view these search results relative to their position on the page. This is a convincing argument to sell a company search engine optimization services in order to get them up into the “red zone.” There are also some other great visualizations of user behaviors, as well. […]
March 21st, 2007 at 9:11 am
[…] Hela studien (en av många) av detta fenomen kan du läsa här: Distribution of Clicks on Google’s SERPs […]
April 11th, 2007 at 12:24 pm
I\’m not impressed.
Sorry.
You all need to think about this….
You aren\’t keeping in mind that the students knew they were on some sort of research study. I don\’t know about you, but I would probably click on the first link also.
When looking for stuff, I myself have clicked on the first link many times. So what? I have also clicked on the second, third, fourth…. We all have. And we have because we are looking for better and more selections. That business that is ranked at the top or near the top may have a lousy site. They may cost too much. On and on….
Of all those test subjects who clicked on the first link, well, they aren\’t go any further because they aren\’t looking for anything. They are just on a test research.
Women, especially like to shop around. Not all, of course.
I have also read this \”research\”:
Ranking Number 1 receives 42.1 percent of click throughs.
Ranking Number 2 receives 11.9 percent of click throughs.
Ranking Number 3 receives 8.5 percent of click throughs.
Ranking Number 4 receives 6.1 percent of click throughs.
Ranking Number 5 receives 4.9 percent of click throughs.
Ranking Number 6 receives 4.1 percent of click throughs.
Ranking Number 7 receives 3.4 percent of click throughs.
Ranking Number 8 receives 3.0 percent of click throughs.
Ranking Number 9 receives 2.8 percent of click throughs.
Ranking Number 10 receives 3.0 percent of click throughs.
I\’m sorry, but that isn\’t even close to being believable, and not all is being told.
I have clicked on several links throughout page one. Sometimes on page two. And so have you! I have asked a lot of people if they stopped after clicking on the first one. The answer, every single time was \”no.\”
Look at all the porn sites out there. You think all those guys are going to stop at the first link and not click on the others? Nope. How \’bout all those \”funny news\” and \”funny pcitures\” sites? Nope. What about singles looking for dating and sex advice? Not even close.
And how \’bout during the holidays, and fathers and mothers day? You think many of those shoppers are going to click on JUST one link?
Sorry, but those figures aren\’t right. Not even close.
So there is much more to that \”research\” than meets the eye.
May 5th, 2007 at 9:08 am
[…] Oczywiście, najlepiej znaleźć się w pierwszej trójce wyników. Badania dowodzą, że prawie 80% kliknięć zbierają pierwsze trzy pozycje (Źródło: http://www.seoresearcher.com/distribution-of-clicks-on-googles-serps-and-eye-tracking-analysis.htm). Oznacza to największą oglądalność, a więc i sukces dla właścicieli, bowiem gdy treść tych stron zadowoli czytelników, nie szukają już dalej. […]
May 11th, 2007 at 5:18 pm
[…] ” Distribution of Clicks on Google’s SERPs What percentage of clicks get each listing on the search result page? … SEO Comics) ” Blog Archive ” Ituloy Ang Sulong - SEO SERPs tracking study Says: … […]
May 22nd, 2007 at 5:35 am
[…] Google SEPR Click and Attention distribution ‘heat-map’ (via seoresearcher.com) […]
May 22nd, 2007 at 8:35 am
[…] Die Studie ist nicht ganz neu, aber die übersichtliche Heat-Map vom SEOresearcher kannte ich noch nicht: […]
May 23rd, 2007 at 2:28 am
As others have suggested the original Cornell study is deeply flawed.
A more insightful study analyzing the AOL research data accidentally released last year is based on 36,389,567 search queries http://www.seo-blog.com/serps-position-and-clickthroughs.php
Of course clickthrough rates are search term dependent but this gives a very good indication of the ‘average’ clickthrough rates and I even constructed a tool based on this data http://www.seo-blog.com/position-and-clickthrough-tool.php
- Michael
May 24th, 2007 at 8:39 am
I find the research results very interesting, althought there were apparent problems, such as the small participant group size.
June 19th, 2007 at 7:57 am
Są frazy gdzie pierwsze kilka pierwszych miejsc zajmuja portale i platformy, wynik zaraz po nich (czyli na dole 1 strony) jest niekorzystny, lepiej byc 11 niż przykładowo 8mym. (1 na drugiej stronie)
July 20th, 2007 at 4:08 am
[…] Informacje o badaniach zostały zaczerpnięte ze strony Distribution of Clicks on Google’s SERPs. Polecam lekturę tego dokumentu, ponieważ można zapoznać się z komentarzami na temat wiarygodności przeprowadzonych testów. Dwa aspekty rzucają cień na wyniki: mała liczba zbadanych użytkowników (26 osób) oraz narzucenie testowych szukanych fraz. […]
July 27th, 2007 at 3:31 am
[…] Informacje o badaniach zostały zaczerpnąłem ze strony Distribution of Clicks on Google’s SERPs. Polecam lekturę tego dokumentu, ponieważ można zapoznać się z komentarzami na temat wiarygodności przeprowadzonych testów. Dwa aspekty rzucają cień na wyniki: mała liczba zbadanych użytkowników (26 osób) oraz narzucenie testowych szukanych fraz. […]
July 28th, 2007 at 2:02 pm
yes you’re right, I estimated the percentages from fig. 2 exactly like you did. I was myself surprised with the missing clicks and it was possible that some of those clicks were made on the 2nd and 3rd pages but were not reported in the research results. Also some of the clicks probably were just not counted.
August 3rd, 2007 at 5:16 am
It`s a great article ! I`m shocked about the 7th place thing ! It`s really useful. Can add a link to this article on my website? Greetings
August 3rd, 2007 at 8:29 am
[…] In the field of Search Engine Optimization (SEO) they are important to us because you ideally would like to rank in the first 10 results or first page of the SERPs. That’s because the first page of results is where most users find their information. According to a recent study by Cornell University the first two listings on the first page of the SERPs capture over 50% of the user’s attention. […]
August 18th, 2007 at 10:37 pm
[…] Google Sonuclarinda Kim Neyi Tikliyor Ilginc bir calisma http://www.seoresearcher.com/distrib…g-analysis.htm […]
August 19th, 2007 at 6:04 am
This is great working. 7th more low than 8th. Interesting.
October 24th, 2007 at 7:57 am
[…] Oczywiście, najlepiej znaleźć się w pierwszej trójce wyników. Badania dowodzą, że prawie 80% kliknięć zbierają pierwsze trzy pozycje (Źródło: http://www.seoresearcher.com/distribution-of-clicks-on-googles-serps-and-eye-tracking-analysis.htm). Oznacza to największą oglądalność, a więc i sukces dla właścicieli, bowiem gdy treść tych stron zadowoli czytelników, nie szukają już dalej. […]
November 24th, 2007 at 6:26 pm
[…] Source: searchresearcher.com You can clearly see that, although generally higher rankings get more attention and clicks, there is a blind spot at the 7th placement. It is the most unfortunate SERP in the top ten. With click through rate of 0.36 it is far worst than any other placement. If you rank 7th you are doomed and do anything to get out from this click desert, even if you can only go lower in ranking. […]
December 13th, 2007 at 12:10 am
This is interesting. Click throughs are not distributed hierarchically. If the above study is correct then I’d rather stay on 11th position than on 7th & 10th. But of course it would have been more accurate if done on large number of respondents.
December 13th, 2007 at 12:46 am
I mean 7th position only.
January 16th, 2008 at 1:45 pm
Very interesting points, I had a webpage on 5th place and I had 100 clicks per month, now I reached 1st and I have 300 clicks on average.
February 28th, 2008 at 5:06 pm
I agree that this survey can be thrown out since the search keywords are predetermined. This creates a situation where the user is more apt to click the first listing since they are really not “trying” to find anything. I assume that if the users was told to research subjects and then answer questions on the subjects, this would change the order percentage clicked greatly.
March 4th, 2008 at 3:27 pm
Although, the article is not actual the topic will be actual always.
Thank you for your great study with much new iadeas to improve the own online-strategy-concept.
Ralph
March 31st, 2008 at 8:33 am
Seo - Top Ten…
In der gesamten Seo-Blog-Szene und darüber hinaus ist eine Differenzierung zu verzeichnen. Da fällt es auf, dass die Qualität manchmal den Bach hinunter geht. In manchem Blog findet man Beiträge, die eine Länge von einigen weni…
April 9th, 2008 at 2:15 pm
This is an incredible study. More reason to go for no. 1.
May 1st, 2008 at 1:14 pm
Very interesting. Obviously results will vary between markets and search queries. They tested 397 different queries among an assortment of topics. People looking for information will likely focus on organic listings, where a higher percentage will look at paid search if looking for a product. The graph does not account for paid search nor does it mention how it affected results. Nonetheless, I think we can all agree, the higher, the better.
May 2nd, 2008 at 7:25 pm
Hello, do you have any such research on Yahoo and MSN? I think they as important as Google. Thanks.
May 3rd, 2008 at 3:59 am
@SEO Come: Yes, Yahoo is interesting for me, too. The actual problem is the bad image of Yahoo and the extremly position of Goolge for SEO.
Ralph
May 3rd, 2008 at 12:19 pm
Intersting. I never imagine if the last position (10) in google SERP page better than position number 7-9.
Thanks for the info