Would it surprise you if searches on the Web make up around 10 percent of all pageviews on the Web, and indirectly led to more than 21 percent of the pages viewed online? It surprised a couple of researchers from Yahoo.
That’s the result of a study conducted by Ravi Kumar and Andrew Tomkins from a sample of over 50 million user pageviews that they collected during 8 days in March, 2009. The information was captured through the Yahoo toolbar from people who agreed to the collection of data for this kind of analysis. Additional information was added by looking at the search logs from Yahoo.
While the data is limited to users of the Yahoo toolbar who agreed to the use of the data, and doesn’t include mobile searches or searches that used AJAX to display results, it does capture how people browse the Web and search at a number of search engines as well as searches at sites like eBay and Amazon.
The study is described in a paper titled A Characterization of Online Search Behavior (pdf), and is being presented tomorrow at the WWW2010 Conference in a session dedicated to User Models on the Web.
While the information collected by the Yahoo researchers may not reflect all searchers, it does provide some interesting insights into how people browse and search on the Web, and gives us some interesting statistics to think about. The researchers also provide some interesting details on what kinds of things people tend to search for when they search.
The study begins by comparing all pageviews recorded by the toolbar, and seeing what percentages of those are from a main web search (from a search engine such as Google, Yahoo, MSN, Ask, and AOL), what percentages are multimedia searches (from places such as YouTube, Hulu, Flickr, and Picasa), and for site searches of items (on sites like Amazon, eBay, Craigslist, Imdb, Singlesnet, Careerbuilder, and Leboncoin).
They tell us that main web search accounted for 6.2 percent of all pageviews, multimedida searches take up 1.4 percent, and item searches another 1.4 percent. They followed browsing trails from these searches to the pages that searchers followed from the searches, and tell us that those pages browsed as a result of searches result in another 12.4 percent of pages visited on the Web.
The searches and the pages browsed as a result of the searches account for 21.4 precent overall of all pages visited during the 8 days included in this study, or about 1 in 5 pageviews.
Types of Searches
The Yahoo paper goes into a fair amount of detail on how they identified pages visited as being searches, and how they decided to investigate further to see what kinds of things people were searching on those search pageviews.
In investigating what kinds of things people were searching for, the researchers decided to take a random sampling of queries identified from the Yahoo web search query log.
They noticed that about 50 percent of queries refered directly to some kind of specific item or object, that 8.5 percent were about some broad topic or concept, and smaller percentages of searches included things like searches for URLs of pages and for navigational queries that didn’t include URLs but were aimed at bringing searchers to specific pages.
The study took a look at the kinds of specific items and objects made up about half the queries identified, and broke them down into the following categories, with percentages of queries for each:
- Event – 2.31
- Games – 1.15
- Notable person – 13.08
- Ordinary person – 4.42
- Specific product – 11.35
- General product – 8.56
- Places – 4.90
- Business Categories/Services – 7.69
- Health Issues – 1.35
- Real estate – 1.15
- Media title (names of movies, ED albums, etc.) – 10.10
- Organization (businesses, nonprofits, government) – 33.94
The study also looked at “checkout” pageviews to see how often people arrived at ecommerce checkout pages, and tell us that approximately 20 percent of people making a purchase online eventually arrived at that page directly or indirectly from a search pageview.
The conclusions from the paper include this statement:
We conclude that the impact of search on online activity is more significant than we had anticipated on undertaking this work, and that the impact of structured data on search, while already large, leaves significant headroom for future extensions of the search product.
Conclusion
It’s rare when one of the major search engines digs into the data it might collect from a source like their toolbar or log files, and shares the results of their analysis of the information they’ve collected. It would be great to see more studies like this one.
People do start off viewing pages on the Web in ways other than through searches, such as visiting sites that they’ve bookmarked, typing addresses straight into their browser address bars, following links to other pages, and accessing pages through other applications such as email or IM.
This study tells us that 8 days of Yahoo toolbar data, collecting the browsing activity of a large number of people, indicates that searches of the Web, multimedia, and items take up about 10 percent of all pages viewed online, and those search starting points lead to another 11 percent of pages viewed on the Web.
It’s possible that if you conducted this study with the Google toolbar, or the Bing Bar, or looked at data from places like internet service provider log files, you might see different numbers. But, it’s an interesting study, it gives us a sense of how important search is to people who use the Web.
Whilst the research carried out is indeed interesting, I agree that had the study been carried out with perhaps the Bing bar or more so the Google toolbar, the figures could have resulted in somewhat different figures.
I don’t know anyone that uses the Yahoo toolbar and I would have it as a guess that the majority of these users would be the less tech savvy crowd, probably most of which have installed the toolbar by accident when installing programs (as I often see this being included with program installations) and don’t know any different with regards to unchecking the checked by default inclusion.
I would be interested to see how much people from the search/seo industry make up searches on the web alone, heh.
I also believe that if this research is conducted with the Google toolbar, the results would be different because users of Google are different than the users of the Yahoo toolbar. I really think they should have categorized the users.
I also agree with other readers that yahoo toolbar gets installed by default with many software packages and less tech users leave those as it is.
Question remains same as how many SEO experts or Search Engine known person will like to use yahoo toolbar.
It’s great to get nuggets like this. They’re great info to have when trying to validate/justify SEO efforts.
I do agree that the distribution of searches are likely different for Yahoo vs Bing vs Google; as evidence would suggest in PPC behavior in each of those platforms. But in any case some data is better than none at all.
Hi Geoff,
It would be interesting to see similar kinds of research from Google and from Microsoft, and to see what kind of results they achieve. I hope that they would consider doing some research like this, and publishing their results.
I know I’ve installed (and immediately uninstalled) the Yahoo toolbar when downloading something else from Yahoo because I wasn’t paying attention, and didn’t uncheck it, but I do know someone who does have it installed.
With 50 million pageviews over 8 days, I would expect that some of the searches were probably from people in the search/seo industry, but that a good percentage weren’t.
Hi Andrew,
I agree that the audiences probably are somewhat different, and that we would likely see different numbers of Google toolbar users’ pageviews were analyzed. They did collect information about searches on Google, Yahoo, and MSN (it wasn’t Bing yet at that point) as well as at Ask and AOL, and it would have been interesting to see a breakdown based upon searches at the different search engines.
Hi Mike,
It would be interesting to find out how many actual users of the different toolbars there actually are and how they use them. I wonder if that is something that they search engines would be willing to share with us at some point in the future.
I suspect that more people who are involved in SEO are likely to use the Google toolbar than the Yahoo toolbar, if for no other reason than the PageRank display.
Hi Mark,
I agree – I think that the kind of statistics provided in the Yahoo paper can be helpful in showing site owners the value that improving their sites for search can bring to them.
The data is limited, but as you note, having some data is better than having none at all.
Hi Bill, yeah I must admit, I too, have ended up installing the Yahoo toolbar by accident before and immediately ended up uninstalling upon realising. I do believe that the Google toolbar allows installers of it to opt in to a service that allows Google to collect usage data, the Bing toolbar however, I have yet to even try so I may check this out and see what it’s like 🙂
That’s a relief. I might have a job for a while yet 🙂 I wonder what makes a person “notable”.
I liked the bit in the first section of the paper which indicated that search engines would be heading towards providing the information themselves. What is more fascinating is the referral forest. I could see where having placement on a good hub page or a page connected to a good hub page can be quite beneficial. I wonder if following user behavior in the referral forest would change how a search engine would provide information in its results for a query. I would have liked the study to delve more on the number of hops taken to reach checkout. What is user behavior in these hops? Can we create a standard model to match or improve upon? Where the hops needed, or is that little results section that you can find in Google searches listing prices for an item taking people past pages that the webmaster may have wanted them to hit for a bit of upselling or suggestive selling. Well, I guess that I would like to see more studies on the topic in general. 🙂
Yes, but if the toolbar doesn’t get installed with other applications automatically then who’s going to download it? They’ve gotta find a way to get it on people’s computers.
Hi Geoff,
There seems to be a little controversy over whether Google continues to collect information or not after those features that track are opted out of. Ben Edelman wrote about the problem in January, in an article titled Google Toolbar Tracks Browsing Even After Users Choose “Disable”. Google responded to the post (over at Search Engine Land), saying that the tracking stopped after the browser was restarted, but Edelman disagreed, and explained why. I’m not sure of the present status of data collection through Google’s toolbar at this time, and if it was an ongoing problem, if it has been resolved.
I may have to try out the Bing toolbar, myself. I know that Microsoft conducted at least one study with browsing and searching data that they collected from toolbar usage, though they stated in their whitepaper that it was done with full disclosure and explicit permission. I wrote a post about it, at Microsoft Tracking Search and Browsing Behavior to Find Authoritative Pages.
Hi Steve,
It was encouraging to see this report, and it’s conclusions on how prevalent search is on the Web.
The paper does actually provide a definition for what they call a notable person, as opposed to an ordinary person. A notable person is someone:
The difference between a notable person and an ordinary person is that an ordinary person is more likely to be “of interest” to someone who already knows him or her.
Hi Frank,
It is interesting that search engines might evolve towards understanding what they refer to as structured objects, and an object type search that collects information about specific people, places, and things and presents information about those with references to the sites where that information was found, so that instead of a list of pages about a particular topic, we might start seeing more of an information rich type of search result.
I thought it was interesting that they were tracking referral trails as well, to see where people traveled on the Web as a result of a search. I agree – would love to see more research and data along those lines as well.
Hi James,
I have seen people download and use the Yahoo toolbar. It does seem like it provides some useful shortcuts, especially if you use yahoo applications like Yahoo Mail and some helpful features (like a popup blocker).
I would love it if there were a way to find out how many people were using the Google toolbar, the Yahoo Toolbar, and the Bing bar.
This is indeed a very interesting research. Search is indeed very important to internet users. With toolbars like Google and Bing, it makes search more convenient. It is also interesting to find that most of searches are relating to organizations and people, where real estate and games are the least search areas.
Best regards.
Results like this would be naturally skewed. People with the Yahoo toolbar will more likely favour Yahoo search.
Interestingly though there are lot of people buying things online.
Hi Aldric,
I found their breakdonw of queries by search types pretty interesting too. It seems reasonable that searches related to organizations and people would be more popular, but nice to see a study that uses some data to support that reasoning.
Hi sedunia,
The data they provided regarding search didn’t give us any breakdown of how many people searched using different search engines, but I suspect that you’re right that people who have the Yahoo toolbar installed would probably be more likely to use Yahoo to search with. As I mentioned in the comments above, it would be great to see some similar research from Google and Bing, with results from their toolbars.
Cannot trust this data…think of yahoo toolbar, widgets etc
Hi SEO Gibraltar,
I understand some of your hestitation, but the researchers involved in this study, Ravi Kumar and Andrew Tomkins, have authored a good number of research papers and are fairly well known and respected in the search industry, and have frequently been cited by others from other search engines and in academic circles.
I believe this. People get online and go immediately to a search engine, type in what they are looking for, and get it.
This article shows me that search engines like Google, Yahoo and Bing are doing their job at creating a really good search experience for the end-user if 1 in 5 pageview are coming from search.
Hi Derek,
The number of pageviews identified as searches does seem to indicate that search engines are a large part of what is happening on the Web. I do think they are improving in the experience they provide for searchers as well.
I don’t really think it is surprising. Internet users most times resort to internet in order to find information. Finding this information means that they have to search for it. So, i guess this number pretty much identifies that need.
Interesting. I always thought that it would be higher than 1:5 because, at least in my opinion, search is the default starting point (and browser homepage for that matter) of web surfers.
Maybe it would be different if Google did this or publish their data about this. Anyhow, I think the study is valid. Some people may say that these people who had their yahoo toolbar on are not tech savvy. As marketers, we all have to empty our cups and step back and see how the average web user goes about through the internet.
Thanks Bill for bringing this to our attention.
Hi Spyros,
What makes that number interesting is that it also considers things like people viewing pages to check email, visit pages that they have bookmarked, follow links from email and IM and other web pages, and other ways that people might visit web pages. Many people do log on the web with a specific destination in mind.
Hi Jim,
Many of the pages I view are ones that I type into the address bar of my browser, or that I click on a bookmark to visit. I’m not sure that it’s safe to say that search is the default starting point for most web surfing sessions. It is intersting to have some data to look at when considering that, though.
It would definitely be interesting to see Google publish the results of a similar study.
This is indeed a very interesting research. Search is indeed very important to internet users. I also agree with other readers that yahoo toolbar gets installed by default with many software packages and less tech users leave those as it is.
Question remains same as how many SEO experts or Search Engine known person will like to use yahoo toolbar.I also believe that if this research is conducted with the Google toolbar, the results would be different because users of Google are different than the users of the Yahoo toolbar. I really think they should have categorized the users.
Hi Sam,
It’s Yahoo’s toolbar, but it’s also their study. I’d love it if Google published information along these lines as well.
i don’t know if these statistics are true but i expected categories like games for example to have bigger percentage than 1.15%. I mean, most of the teenagers search for pc – video games. Also, we shouldn’t forget that a large number of casual gamers use facebook as their gaming platform (etc farmvile and other stuff like that).
Hi Alex,
The study was taken from actual data collected by Yahoo, but it’s possible that the way they categorized certain searches might not have done the best job of capturing the intent behind a search.
As you note too, many people who might search for games may just go directly to a site like Facebook (that they might not search for). And while there may be very large numbers of teenagers searching for games, Search is a pretty popular activity on the Web, and there are a lot of people searching for other things as well.
i agree with you. Anyways, these studies show some leads to what people search. No study can be 100% accurate. After all, without studies like this one we wouldn’t make a clue of what is profitable or not in the end.
Hi Alex,
One of the things that I hate about reading studies like this one is that I don’t have access to the raw data. I’d really love a peek at what the actual results of the study were, and how they decided which types of pages fit into what categories, and so on. They do come up with some interesting conclusions – but I’d just love to see how they got to those conclusions.
Hi Bill,
Thanks for the great post. Has Google or Bing published any report like this? It would be interesting to compare all three reports by these leading search engines.
Hi Max,
I’ve been keeping an eye out for a report like this from Google or Bing, but really haven’t seen too much from them.
The closest I’ve seen from Google has been a study from them on how people might search differently on desktop computers compared to web enabled phones, and iPhones – Computers and iPhones and Mobile Phones, oh my! A logs-based comparison of search users on different devices (pdf)