Yahoo on Search Advertising and Behavioral Targeting

The advertisements from search engines that we see accompanying search results, or on portal pages, or as part of a content network, are often related to the query used in our searches or the content of the pages that we are viewing.

Would ads which we view that are more targeted towards our interests be more effective?

If so, how would a search engine know what we are interested in seeing, and would they understand what our intent might be in looking at different pages – whether we are just gathering information, or if we are serious about buying something?

Building Profiles

Search engines can collect a lot of information from us, as we go about the internet and they follow the trail of information that we leave behind.

It’s possible for them to record our travels, and attempt to analyze our activities to try to determine our behavioral patterns and interests, with the intention of targeting us with advertisements that provide us with a “more meaningful and rich experience.”

If there’s a hint that we might be interested in certain products and services, they may try to show us advertising targeted towards our interests.

A series of patent applications from Yahoo describe a behavior targeting system that attempts to determine user profiles from the online activities of people so that it can get a sense of which ads to show to those folks.

Profiles and Different Marketing Objectives

What makes it interesting is that these user profiles can be aimed at different target marketing objectives, such as:

  • Direct response advertising,
  • Brand awareness advertising,
  • Purchase intention activities, and;
  • Intra-company business unit marketing.

The patent applications describe a marketing funnel that identifies marketing objectives in three stages.

1) At the top of the funnel, an advertiser may desire to acquire brand awareness for the advertiser’s brand. Typically, for this type of marketing, the advertiser’s goals are to promote a brand for a product by associating one or more positive images with the brand.

2) In a second stage of the funnel, a user may desire to gather information for product consideration. To address this cycle of purchasers, advertisers may use direct response advertisements. There may be many different objectives associated with direct response advertising:

  • Acquisition — to get consumers to become a customer or visitor of the product/service.
  • Retention — to maintain existing customers or visitors to the web site.
  • Engagement — to elicit more activity in existing customers.
  • Monetization — to increase profitability of the customer via active purchase activities, such as cross-selling, as well as passive activities, such as consuming banner ads.

3) The last, and most focused part of that funnel, is where the user is actively shopping, and intends to buy something.

I wrote about a presentation by Yahoo’s Usama Fayyad from this summer where he talked about some of the experiments that they were conducting following users across this sales funnel.

Their case studies showed that people who saw banner ads on one of Yahoo’s portals were more likely to click upon an ad in search results for that brand or product than people who hadn’t see those banner ads.

Yahoo Patent Filings on Behavioral Targeted for Ads

Evidently, Yahoo was doing more than taking notes during those case studies, and they filed a number of patent applications on their attempts to match users to profiles to determine which advertisements to show to them:

Here’s the abstract from the first of those listed:

A behavioral targeting system determines user profiles from online activity. The system includes a plurality of models that define parameters for determining a user profile score. Event information, which comprises on-line activity of the user, is received at an entity.

To generate a user profile score, a model is selected. The model comprises recency, intensity and frequency dimension parameters.

The behavioral targeting system generates a user profile score for a target objective, such as brand advertising or direct response advertising.

The parameters from the model are applied to generate the user profile score in a category. The behavioral targeting system has application for use in ad serving to on-line users.

Some of how it works:

1) When the system receives a user event or after a predetermined amount of time for accumulating events, the system classifies the event into a category that might match up with advertising categories.

2) A user behavior model may be selected based on the user/marketing objective.

3) The behavioral targeting system processes one or more events to generate a raw user interest score for a category, and may look at how recently events have occurred, and how interested the user may appear to be in the subject category. Frequency may also be looked at to see how frequently that user event happens.

Those events can include advertisement clicks, search queries, search clicks, sponsored listing clicks, page views, and advertisement views, they may also include any type of online navigational interaction or search related events. Here are some examples:

  • Page view event — Someone might visit a page on a portal about music, by clicking on a link for the music category page. A page view event is recorded for the user’s view of that music category page.
  • Advertisement view event — Someone at a portal may click upon a banner advertisement on the home page of the portal. Vieing the advertising page in response to the click would constitute an advertisement view event. If they look at multiple pages on the advertiser’s site, those could be counted as multiple page from the user.
  • Advertisement click event — Similar to a “view” event, this counts the clicks when a banner ad is clicked, and when links on the advertiser’s site are clicked.
  • Search query event — occurs when someone submits search terms to a web based search engine. Searching for the query “Deep Sea Fishing”, and receiving a set of results is a search query event that can be recorded with the search terms “Deep Sea Fishing.” If a user clicks on one of the links, a search click event occurs.
  • Sponsored listing advertisement — refers to advertisements that are displayed in response to a user’s search criteria. A sponsored listing click event occurs when a user clicks on a sponsored listing advertisement displayed for the user.

These kinds of events may be associated with different categories in a hierarchical taxonomy, so that a profile can be created showing the things that a specific user might be interested in seeing.

That profile can be used to serve banner ads to someone using a portal, but could also be used to “match sponsored listings to user behavioral profiles and to webpage content.”

Deciding Upon Which Ads to Show, Based upon User Profiles

The patent applications listed above focus upon the creation of user profiles, but they don’t go into detail on how advertisments might be chosen based upon those profiles. Another patent application, from early October explores that territory:

Method and apparatus for selecting advertisements to serve using user profiles, performance scores, and advertisement revenue information


A method and apparatus for selecting additional content to display to a user when the user requests base content is provided.

A user profile of the user having user interest scores of categories or keywords is received, each user interest score reflecting the degree of interest the user has in the category or keyword.

Performance scores reflecting the probability that a user having particular user interest scores will select additional content associated with particular categories or keywords is also received. In addition, revenue amounts associated with each category or keyword of the user profile is received.

The user interest scores, performance scores, and revenue amounts are used to produce an expected revenue amount for each category or keyword in the user profile.

Additional content to be sent to the user is then selected using the determined expected revenue amounts.

I’m not completely convinced that I really want to see ads that are “targeted” to my interests, and it will be interesting to see how advertisers and consumers react to a behavioral targeting system like this. Do targeted ads provide a “more meaningful experience” for people who use the Web?

29 thoughts on “Yahoo on Search Advertising and Behavioral Targeting”

  1. Hi Bill, One thing I have been wondering about in regards to your research is – How much and how many of these patents actually make it out of the idea phase and into the search enviroment?

    I hear a lot of people regurgitating things they have read in patents, or more likely things other people have read in patents and then read thair summary as if they are fully functional right now. The most popular ones are that Google uses text surrounding the link and that Google has a history of all the link on the internet.

    Also, in reagrds to your question:

    Do targeted ads provide a “more meaningful experience” for people who use the Web?

    Personally I don’t click on adverts and only click on PPC if I cant find anything on the organic list. I think Yahoo would probably be better off spending their resources on improving their organic index. Although I guees they make their money from PPC, so they are more intrested in investing in that.

  2. Hi David,

    How much and how many of these patents actually make it out of the idea phase and into the search enviroment?

    It’s really all over the place.

    I see some granted patents and patent applications where it’s pretty obvious that a search engine has been doing what is in the patent for a few years – examples include Google’s Definitions and autolinks and Microsoft’s commercial intent sliders.

    There are others where the patent filings spell out some processes that sound so fully formed that you could swear that they have fleshed it out completely, and it’s something that you just haven’t seen yet. Sometimes that’s true, like Yahoo’s patent application for Y!Q, and sometimes you have to wait 6 months, like in the case of Google’s Web History patent applications.

    Sometimes the patent applications are really timely, and there’s something like an Official Google Blog post or Google Lat/Long blog post that announces it within a week of the patent application being published. I believe the one that talked about being able to customize map routes was timed like that.

    There are times when it’s almost impossible to tell that the patent or patent application is something that has been implemented or not because it involves a mechanical or algorithmic process that isn’t very transparent.

    Sometimes those seem to make a lot of sense given the way some things work, like Google’s supplemental index and Google’s granted patents on extended databases. Sometimes it’s harder to tell if a certain algorithm has been put in place, like the phrase-based indexing patent applications, thought there are some behaviors on part of the search engine that lean towards an adoption of at least some of the parts of the inventions described.

    Sometimes things described in patent filings are implemented in parts, like the kiosks at the gas stations from last week. Will we see similar kiosks in shopping malls and resort centers like the patent application describes? I’m more of a believer now that the gas station kiosks have appeared.

    With the Yahoo patent application described above in this post, if I hadn’t watched the Usama Fayyad presentation I mention above, I might have wondered if the process described in this patent application would be something that we would ever see from Yahoo. After watching the presentation, my wonder is not if, but rather when. The whole “understanding the intent of a consumer during different parts of a marketing funnel” seems like something Yahoo has given a lot of thought towards.

    Regardless, there are a lot of patent applications that explore topics that will likely never be used. Sometimes the difficult part isn’t coming up with the patent, but instead implementing it.

    Then again, sometimes the patent application is the tough part. I remember reading the blog of one of the co-inventors of one patent application, and he wrote that my description of the patent filing used more words than what they wrote as a team, if you didn’t count the code that they put together for the process described in the patent.

    The most popular ones are that Google uses text surrounding the link and that Google has a history of all the link on the internet.

    Yes, and that text around a link is something that Google has been saying that they may use ever since the first Brin/Page paper on PageRank, and the first couple of Stanford patents on PageRank. I’ve seen a lot of other primary resource materials (from search engineers) that leave that open as an option. Its been repeated enough by folks who work on search engines that it is probably a good thing to test. 🙂

    It’s difficult to tell if Google has a history of all the links on the internet, but it would be a nice thing to have. The concept of “link lists” that are used by the search engines appears in a lot of patents, and Google hasn’t been shy in suggesting that they are keeping track of changes in linking, in the use of anchor text from links, in the amounts of links to and from pages, and even sometimes in the text that surrounds links.

    Of course, just because something appears in patent applications and whitepapers doesn’t mean that it is being used, or ever will. But if the stuff keeps on reappearing, and it seems that a good number of folks from search engines have given it serious consideration, then it wouldn’t be something that I would dismiss lightly.

    I agree that Yahoo would be well served by working on their organic index, but like you say, it is important to pay attention to the part of the process that brings in money, too.

  3. That is good information on what Yahoo is doing. I just hope that Yahoo will speed up in this area and roll it out to the marketeer quickly…I think that Yahoo is alittle behind in the game of on site advertisement…


  4. Thanks for the in-depth summary of your knowledge. It maybe worth posting such a summary up as a sub-page. I did look at your about page to see if I could get an overview of what your thoughts are.

    Moving further comments over to the next article…

  5. Thanks, Anthony.

    It will be interesting to see how quickly Yahoo might pursue this.


    Appreciate the suggestion. I may put a summary like that on a sub page, or my about page. It sounds like a good idea. Much obliged.

  6. Pingback: This Week In SEO - 11/16/07 - TheVanBlog
  7. Personally I rather use organic list then click on adverts.
    Investing to improve their organic index would be the best, but since they make money from ppc, they will be more interested in to investing in that

  8. Thanks.

    I tend not to look at the paid results in search listings much either.

    It does seem to be a conflict though. Is it true that the better the organic search listings, the less likely that people will click upon ads?

  9. Hi Garth,

    Tough question to answer, I think because you may be asking me if Yahoo might be leaning towards incorporating more user behavior signals into its search results. I do think that is a possibility.

    Natural search does play a role in the behavior targeting model I’ve described above in a slightly different way – in the searcher’s information gathering stage, where a potential buyer may be looking to find out more about a product.

    It doesn’t hurt for an advertiser that uses banner advertising to show up in search results for brand name keywords, and Yahoo may make it more likely for that to happen. See my post Redefining Navigational Queries to Find Perfect Sites.

    So, someone who sees a banner ad and may search for trademarked terms in it, but is looking for an organic result because they aren’t quite ready to buy, may be more likely to see the web site for the brand in question.

    If they later decide to buy, it’s possible that they might be more likely to click upon an ad showing in search results. Natural search can play a role in this sales funnel that goes within this behavioral targeting system.

  10. Hi Bill

    Do you see behavioural targeting being adopted by search engines for their natural search?

  11. I also believe that Search Engines are becomming accustomed to how and what we search for. However, i think this cannot affect the keywords the user is actually looking for, only contribute if the keywords they are searching for has any resemblence to previous search behavior.

    For example if your tap burst and needed a plumber – which is a term probably you don’t search for that often (unless you have a really dodgy tap) your previous results on music searches and perhaps purchasing a TV should not really influence your search on your plumber.

    The question is, will the search engines be smart enough to know when you are generally surfing for the ussual or searching for a one off cause?

  12. Hi Claire Web,

    You asked one of the really big questions in search these days:

    The question is, will the search engines be smart enough to know when you are generally surfing for the ussual or searching for a one off cause?

    It’s possible that search engines are evolving to become less ways to match keywords, and more ways to provide recommendations on pages, based upon things such as your previous searches, searches of people who may be similar to you in some way (they might like some of the same topics or pages or live in the same area), or have similar browsing and bookmarking histories.

    Search engines also seem to be trying to understand more when a search is informational in nature, transactional, or navigational.

    If you see a set of product search results at the top of your search results, for instance, a search engine might think that your intent is to buy something. If you see a map, and a list of service provides, like plumbers, they might think that your query is aimed at helping you find a plumber. If the first result is for a business that might share a trademarked term, or service mark with part or all of your query, and the top result includes a set of sitelinks, the search engine may have thought that your search was navigational one in nature.

    Hi Dazzlecat,

    I’m not sure that I saw the BBC reference, though I have had the good fortune of having one of the technology bloggers from over there link to a couple of posts from here.

  13. I don’t see the public reacting well which this behavioral targeting idea.
    Hows the search engine suppose to know when a user is performing a one off search and how is it going to work on public computers. I don’t see this going any further than the patent.

  14. Hi Sam,

    Good points. I’ve seen a few odd ads on “one off” searches, and public computers can be an issue.

    If you haven’t had a chance to watch the presentation from Usama Fayyad, it’s worth listening to how he describes some aspects of behavioral targeting.

  15. I’m not sure that it would work either, I feel that the majority of people would become frustrated by it. There will be times when people will make one off searches or share their computers. No reason why they could include it as a option however.

  16. Hi SJL,

    If you haven’t had a chance to watch it, the yahoo experiments described in the Usama Fayyad presentation are worth taking a look at (once you get past the first eight minutes introducing Usama Fayyad).

    A few one off searches or shared computers may not make too much of a difference. It may take a number of related events (as described above) before a profile is created for a specific user, and advertisements related to the profile are shown to that user.

  17. Im a bit confused after reading all of this, William could you provide a link to those videos, seems they might shed some light on the situation.

  18. That was a really interesting article. I don’t really know how I feel about people gathering information about me via my searches but it could help the marketing world.

  19. I think this is taking it too far, it’s a bit intrusive in my opinion. It makes sense that the revenues after this type of tracking would be crazy since they can show the type of ads that we often click but it takes away the user friendliness that they encourage the webmasters to maintain so often. We’ll see how this turns out.

  20. Hi Daniel,

    I agree that it has the potential to be pretty intrusive displaying ads based upon the interests of visitors rather than the query used, or the content that surrounds an ad spot.

    Would it be less user friendly, if the ads shown related to the content found on the page, the profile of the visitor, or a combination of both? I’m not sure.

  21. Very interesting stuff. I believe it would be difficult for Yahoo to properly measure the effectiveness and fully implement this behavioral targeting technology. Say Yahoo’s search results are 30% worse than Google’s, then I can’t possibly see how Yahoo could display the right ads when they’ve been sending you to dud pages in the first place.

    It is a lot more complicated than that, but you really would need a good foundation (search engine) to roll this sort of technology.

    I’ve tested many keyword phrases on both search engines simultaneously, and many times I wondered how Yahoo ended up with some of the results I got.

  22. Hi jakubsnm,

    I don’t know if talk of Yahoo’s search technology is moot or not at this point, with the possibility of a merging of that technology with Microsoft’s. It’s possible that the deal between the companies might not go through, so we probably have to wait and see.

    I don’t know if you had a chance to watch the presentation from Usama Fayyad that I linked to in my post, but it’s definitely worth a watch.

  23. This is definitely helpful. Google may be the number one SE but I do consider Yahoo because of the tremendous traffic it has been giving my site and exposure here has been of great help. Thanks for sharing this info here.

Comments are closed.