Should Search Engines Refer Searchers to Other Search Engines?

Do you have a favorite search engine? Is there a particular reason why you use the search engine that you do?

Do you use more than one search engine regularly? Have you switched from using one search engine regularly to another one?

If you ran a search engine, you would probably want to understand why people shift from one search engine to another, either temporarily or permanently. And you might be interested in seeing if you can identify why and when these shifts take place, and a way to predict when such a changeover might happen.

Microsoft has been exploring why people switch search engines, and have filed for a patent on predicting when someone might switch from one search engine to another. It seems like an odd subject for a patent application, and they even tell us that one behavior might indicate such a switch might be when someone submits a query for “Google” in Microsoft’s Live Search.

The patent filing describes studies that Microsoft has conducted where they collected information about searchers switching to different search engines, and provides some details on how the ability to make such a prediction can be used by a search engine in a number of ways…

The patent filing is:

Predicting and Using Search Engine Switching Behavior
Invented by Allison P. Heath, Ryen William White, Christopher J.C. Burges, Eric David Brill, and Robert L. Rounthwaite
Assigned to Microsoft
US Patent Application 20090112781
Published April 30, 2009
Filed December 18, 2007

Abstract

Aspects of the subject matter described herein relate to predicting and using search engine switching behavior. In aspects, switching components receive a representation of user interactions with at least one browser. The switching components derive information from the representation that is useful in predicting whether a user will switch search engines.

The derived information and information about a user’s current interaction with a browser is then used by a switch predictor to predict whether the user will switch search engines. This prediction may be used in a variety of ways examples of which are given herein.

They’ve also published at least a couple of white papers on the topic that provide some interesting statics. One of those papers explores ways to encourage searchers to use more than one search engine for their searches.

Predicting and Using Search Engine Switching

One of the main business models of today’s major commercial search engines is advertising shown with search results, and search providers obtain revenue that can be tied to how many people use their search engines. When someone switches from one search engine to another, the new search engine gains income while the old search engine losses.

During a three month long study tied to this patent filing, fifty percent of the users studied stayed with the same search engine for all of their searches while the other half switched to other search engines at some point during that time.

People who used more than one search engine tended to perform most of their searches on one preferred search engine.

The inventors of the patent filing define a “switch” as being defined by one of the following behaviors:

  1. Performing a query with a different search engine than the previous query;
  2. Navigating to a homepage of a different search engine;
  3. Querying for a different search engine name, such as submitting a query for “Google” to Live Search.

They also tell us that sometimes a searcher will mistakenly type a query into a toolbar plugin for a different search engine, and those searches may not be indicative of a desire to switch search engines.

One question that this patent filing asks is, “How can a search engine dissuade users from switching to a different search engine?” But, the researchers involved in these studies also seem to be interested in enabling switching in some instances.

Search Engine Switching Behavior

The inventors listed in this patent application have conducted a few studies on switching behavior by searchers. Information about those studies can be found in a couple of white papers from Microsoft:

Defection Detection: Predicting Search Engine Switching (pdf)

Abstract

Searchers have a choice about which Web search engine they use when looking for information online. If they are unsuccessful on one engine, users may switch to a different engine to continue their search. By predicting when switches are likely to occur, the search experience can be modified to retain searchers or ensure a quality experience for incoming searchers.

In this poster, we present research on a technique for predicting search engine switches. Our findings show that prediction is possible at a reasonable level of accuracy, particularly when personalization or user grouping is employed. These findings have implications for the design of applications to support more effective online searching.

Enhancing Web Search by Promoting Multiple Search Engine Use (pdf)

Abstract

Any given Web search engine may provide higher quality results than others for certain queries. Therefore, it is in users’ best interest to utilize multiple search engines. In this paper, we propose and evaluate a framework that maximizes users’ search effectiveness by directing them to the engine that yields the best results for the current query.

In contrast to prior work on meta-search, we do not advocate for replacement of multiple engines with an aggregate one, but rather facilitate simultaneous use of individual engines. We describe a machine learning approach to supporting switching between search engines and demonstrate its viability at tolerable interruption levels. Our findings have implications for fluid competition between search engines.

When do people tend to switch to a different search engine?

The patent filing tells us that some patterns in behavoir seemed to indicate that a switch was about to happen. These include things such as:

  • Increased query length,
  • Viewing multiple search engine result pages, and;
  • Revisitation of previously-viewed pages, amongst others.

Some interesting statistics from the second of those white papers tells us that:

Our analysis showed that 36.4% of searchers used more than one search engine in the duration of the logs. The findings also showed that 6.8% of all sessions and 12.0% of sessions containing more than one query involved a switch between two or more search engines.

The patent application also informs us of at least three classes of search engine switching behavior:

1. Erratic switching: Users switch between Web search engines for almost every search and may use multiple engines concurrently. Such switches may be linked to a desire for coverage, the use of multiple Web browser tabs, and the use of applications that automatically submit queries to multiple engines simultaneously.

2. Short-term or “bursty” switching: Users switch engines for individual search sessions of groups of sessions, but generally return to their preferred engine for most of their searching. Switches of this nature may occur because a user feels that a particular search engine is better suited for the current task. For example, the recent improvements to the Image Search feature on Live Search may have encouraged Google or Yahoo! users to switch to Live Search for image-related queries.

3. Long-term switching or “defection”: Users switch from one search engine to another and rarely return to the original engine. This appears to represent a change in their search engine preference. Such switches–sometimes referred to as “defections”–have profound business importance as this represents a lost customer and a potential erosion of query share.

The paper on Promoting Multiple Search Engine Use provides slightly different names for each of these behaviors, but the descriptions are very similar. It also gives us some statistics on how frequently each of these behavoirs were seen in the Microsoft data behind that paper, with 33.4% of searchers in the study switching between search engines for every search, 13.2% switching over to a different search engine for a specific search session, and 7.6% of those searchers defecting to another search engine and not returning to their original preferred search engine.

Why Predict Switching?

Overall, the main reason seems to be to enable a search engine to understand why searchers might shift from one search engine to another, and improve in areas that might cause defections.

This might involve evaluating a search engines performance for a specific query or for all queries.

It might take into account the interfaces that are shown to searchers in search results, and the ranking algorithms used for queries.

It might help lead to personalized prediction models that can be developed for queries.

Conclusion

It seems unusual that the topic of understanding why a searcher might switch search engines has been developed as a patent application, but it makes for an interesting read, especially if you might be interested in some of the business concerns of how a search engine operates, and how people use search engines.

As a site owner or a searcher, it makes sense to think about the role of search engines on the Web, and how they influence which pages that people might see when they search.

The paper on Promoting Multiple Search Engine Use suggests that it might sometimes make sense for a search engine to “automatically determine when it is in users’ interest to try another search engine.” Would you use Windows Live Search more if it sometimes told you for some queries that the results at Google or Yahoo or Ask.com were also worth looking at, and provided links to those searches? In the conclusion to that paper, the authors describe the benefit of making it easier for searchers to switch for particular queries:

We proposed a machine learning-based approach for supporting switching that estimates in real time whether more accurate results exist on alternate search engines. Estimation is based on features of the query, the result set, and the titles, snippets, and URLs of the top-ranked search results.

An empirical analysis of classification performance demonstrates that it is accurate at predicting when users would benefit from switching between engines at low recall levels. The promotion of multiple search engine use through application components such as that described has the potential to improve the retrieval experience for users of all search engines.

Share

28 thoughts on “Should Search Engines Refer Searchers to Other Search Engines?”

  1. Interesting topic and would be very profitable for Microsoft too in the market but I don’t think it would be easy to challenge google in search engine market

  2. I can see why a new search engine or one of the search engine “also-rans” would want to research peoples’ use of popular search engines like Google or even niche search engines like ZabaSearch.

    New or unpopular search engines will need all the competitive advantages they can get to compete in the online search race.

  3. An interesting article … I wouldn’t really use any search engine apart from Google to perform a general search. However if I was referred to the other search engine results from Google I would probably feel more inclined to look through them.

    Justin – Big Click Studios

  4. I understand why Microsoft is doing something like this with competition such as Google would make a good enough reason and search is big money in the PPC area. The only reason I switch search engine is because I understand that the algorithm is different and I am able to find things I like better with the different search engines out there. Thanks for the information – interesting.

  5. It looks to me like Microsoft want to understand what might make people switch to them or away from them to Google, as without understanding this it’s difficult for them to put together a strategy to challenge Google in the search market.

    But interesting article and it is promising for Microsoft that “36.4% of searchers used more than one search engine in the duration of the logs.”

  6. Hi Miami,

    Taking on Google in the search engine market is quite a challenge these days. Understanding when and why people might leave your web site to go to another web site, either temporarily or permanently is something all site owners should be concerned about regardless of whether they are a search engine or an ecommerce site or a blog. It makes sense for Microsoft to be able to study this kind of switching behavior. Does it make sense for them to let their users know when another search engine might be showing better results for a specific query?

  7. Hi People Finder,

    Yes, it does make a lot of sense for Microsoft to pursue this line of inquiry. Businesses on the Web should be conducting ongoing competitive analysis of others in their market space, whether search engine or ecommerce site or blog.

  8. Hi Jason,

    It is an interesting idea, making it easier for searchers to switch temporarily when they aren’t satisfied with the search results that they see. I do wonder if doing that would make it easier for many searchers to switch permanently though…

  9. Hi Justin,

    Thanks. A number of years ago, Google did point searchers to results for the same query on other search engines. Matt Cutts recently wrote about that in a post on his blog in: Hacking Google: Retro Links Revives Old Google Feature, which mentions a couple of scripts that allow you to use that feature today. I like looking at the same query in different search engines to get a sense of the differences in results from those search engines, and also to find things that I might have otherwise missed.

  10. I only ever switch search engines if I genuinely believe that I can get a better result elsewhere – there have been occasions where I’ve though Google have been poor. Microsoft don’t need to do much investigation into why Google is the main search engine… Google is so integrated into modern life and even pop culture that it’s impossible to expect users to go anywhere else first.

  11. Well, if they want to be fair to searchers then i don’t see a reason why won’t they be pointing them to competitors. But as no one really is fair in the game that is really hardly to expect.

  12. We used to just use “Dogpile” to get results from other sites.. which included Alta Vista back then. MS will do anything right now to get any searcher to use them – they constitute around 5% of the market. I would not want to see Yahoo or MSN results on Google but I do like seeing Flickr images in the Google image index…

  13. Hi Adam,

    If I want to find out as much about a topic as I can, I’ll often use a number of different search engines for my queries, regardless of how good the search results might have been in one of them. As for Google’s popularity, even Google had to begin somewhere. :)

  14. Hi Web Design Beach,

    I guess the reasoning is that if they can’t give you the right pages for your search all the time, they want to be known for being able to point you to better results when and where they might be available. Will they? I don’t know. It would be interesting to see happen.

  15. Hi Mal,

    Interesting point. There’s a long section of the Microsoft patent filing that discusses meta search engines, and some of the problems that those face.

    If Google told you on a small percentage of their search results that the results at Yahoo or Microsoft for a specific query were pretty good (by a link to those results), and they truly were, would that cause you to use Yahoo or Microsoft Live more, and consider leaving Google, or would it make you feel even better about Google?

    I like that Google is showing flickr images as well.

  16. As a fan of Dogpile and there connection with causes, I’d have to say search engines using multiple search engines is a GREAT thing. I like the mixed results and the fact that I’m not giving to any one behemoth at a time.

    BTW – Dogpile.com just launched a new site called http://www.DoGreatGood.com – and their donating a portion of there revenue to Petfinder and the ASPCA.

    Great that theres a way we can help dog adoption programs at no cost.

    Love the site – keep it up!!!

  17. Hi Frank,

    I used to use meta search engines about 9-10 years ago, but I don’t really find myself using them that much these days. Nice to see sites like “Do Great Good.” Thanks for sharing a link to it.

  18. It is really a difficult to work on such measures. Internet surfers are so difficult to predict. I am glad that we have such models and patterns which can describe internet surfers behavior. To be very true, I still give some importance to Yahoo and MSN, because I am receiving traffic from their as well.

  19. Hi Steve,

    I’m in agreement with you – it is difficult to predict how people use the Web and search engines. Looking over the analytics files for a website can give you an idea of how people might use that site, but when you have many millions of visitors to your web site everyday like the major commercial search engines do, it might be easy to get lost in all of the data that you have available to you. When those search engines also can collect information about how people browse the web through things like their toolbars, it because an even more staggering task.

    It makes sense to pay attention to Yahoo and Microsoft as well as Google, and to even find ways to help your audience find you outside of the channels of the search engines – websites stand a better chance of survival if they don’t rely too much on any one source of traffic to their pages. :)

  20. Hi Jojo,

    Looks interesting. Thanks for pointing that one out. I’m going to have to spend some more time there to get a better sense of what they are doing with the search results.

  21. Hi Nathan,

    There are meta search engines that have been on the Web for a long time that will return results from multiple search engines.

    I’d rather search at one, and then look at results from another afterwards, personally.

  22. Hi Bill, I’m kinda late on that one; as Nathan and a few other guys here have said, if the users see a benefit to bring trafic from their current search engine to another one, they might come back and thus increase their loyalty.

    It’s a long term vision that might payoff in the future with aggregating multiple data.

    Interesting and I’ve bookmarked it for later use.

  23. Hi Jeremy.

    I’m not sure that the inventors of this patent were too concerned about webmasters who might have their sites listed on the search engines, but rather pursued this research to learn more about how searchers used their search engine, and how they might improve what they offer so that it’s more likely that those searchers will be satisfied with the search results that they see. Studying when people abandon your website is an activity that any webmaster, even the people who offer us search engines, should be engaged in.

  24. Google has always been my favorite search engine. At the beginning I used various different search engines, but I found Google always provided me with best results. However, I wouldn’t mind if Google would point me to any other search engine if this search engine would provide better results for my query. On the other hand, I can’t really imagine that Google would ever do something like this.

  25. Hi Robert,

    I’m not sure how likely it is that Google would point us to Bing or Yahoo or another search engine either. The study/patent is from Microsoft, and I think one of the purposes behind it is for them to study why people might switch, and what they might be able to do better to keep that from happening. :)

Comments are closed.