When you run a large portal site that provides updated information regularly, one of the goals you have is keeping people on your site to see your content and advertising.
When you run a search engine, one of your goals is helping visitors find what they want quickly, showing them relevant advertising while using your service to find other pages.
Yahoo is in the interesting position of being both a portal and search engine, and that may provide some interesting challenges to what they have to offer.
A recent post at the Yahoo corporate blog (Yodel Anecdotal), Making new Yahoo! homepages your own, reflects many ways that Yahoo will likely use in the future, including a “My Favorites” section where you can place links to your favorite pages regardless of whether they are on Yahoo or not, several new applications, and new ways of looking at older applications such as your email. Local personalization will also be a feature of the home page, where you can see news and applications geared towards your location. The post tells us:
For example, the new homepage in India will include a Cricket app and a whole host of others that are India-centric, while the UK site will include apps such as underground alerts, news from the BBC, and more.
A new Yahoo patent granted this week describes some of the challenges that Yahoo faces as a portal site and hints at the possibility of even further personalization and information aimed at your location, such as News and applications that might be available depending upon which State you live in. The patent focuses on how Yahoo might change and update the news and information that it displays on its home page based upon interactions with its users.
In the “Background of the Invention” section of the patent filing are some interesting paragraphs on the problem that Yahoo is trying to address with the patent:
The hosting of web portals such as Yahoo.com is competitive. Central to such businesses is attracting many users to the portal and keeping such users interested in the portal content. Portal search engine quality is one method for attracting users. However, more is needed to keep users at the portal site instead of linking away from the portal to sites identified by search engine results. Although how advertising revenues are computed can vary, such revenues are generally dependent upon the number of viewers attracted to a portal and the length of time such users stay at the portal. Therefore, portals have worked to extend the amount of interesting content found directly at the portal to retain users at the portal.
One way to attract and keep users at a portal is to provide useful information, including, for example, breaking headline news, financial news stories, feature articles, stock quotes, and the like. Such information attracts users to the portal and helps retain such users at the portal for longer periods. However, users have unique interests. What one user finds to be noteworthy, another may find uninteresting. To attract a high volume of users, it is necessary to offer a wide array of topics on the portal website. This puts a strain on the total web page of real estate available at the portal. Of course, informational topics can be nested so that the portal main page includes a series of links to other pages. While this technique is useful, it is somewhat unsatisfactory because it requires the users to spend a considerable amount of time sorting through lists of topics to find a topic that interests them. Thus, when overused, links to other pages cause users to drop off to sites other than the portal, adversely affecting advertising and related revenue.
Thus, to remain competitive, web portals, as well as other entities interested in attracting large volumes of users to their websites, have devised methods to reduce website clutter while at the same time attempting to maximize user interest in the content that is offered on the web site. However, such attempts have been problematic. For example, user interests are constantly changing. One day, what is an interesting story to many users may suddenly become a story that enjoys very little user interest the next day. Furthermore, the wrong choices can be made. For instance, a website editor may select a news story that enjoys little to no popular interest.
The patent was originally filed in 2004, but the post from the Yahoo blog seems to validate some of the ideas contained within the granted patent. The patent is:
Systems and methods for adaptive scheduling of references to documents
Invented by Jeffrey E. F. Friedl and Kathleen L. Hartnack
Assigned to Yahoo
US Patent 7,539,674
Granted May 26, 2009
Filed July 22, 2004
It makes sense to try to gauge the interest of your visitors in news articles, feature stories, and other information that you might include upon a home page by carefully tracking how popular links to those features might be and replacing them or updating them when interest doesn’t seem very high or appears to have fallen off. Those articles or stories may be given more weight if they seem to be chosen by viewers and stay around for a while, or maybe given less weight and may be removed.
Teasers or a combination of links, snippets, and images may be used for some articles or stories, and there may be more than one teaser for the same story. The effectiveness of those teasers may be compared to each other, and the ones that appear more successful may be shown more frequently.
Different Classes of Viewers and Different Content
The Yodel Anecdotal blog post tells us that Yahoo is experimenting with providing different news and applications to viewers in France, India, and the UK.
The patent tells us that when visitors who have a Yahoo profile and are logged into Yahoo visit the portal, they may see the different information for people who aren’t logged in. Their selections of what they click upon from news articles and feature stories and applications may influence what other people who share similar profiles may see:
The characteristics in the user profiles of users requesting host web pages are used to stratify the user population into classes. Then, a different set of selection weights is refined for each class of users based on the popularity of documents within the class.
As I mentioned above, this might even be based upon which state a visitor lives in:
In one example, the characteristic that is used to stratify the user population into classes is geography. Each user is classified into a different geographical region (e.g., state) based on the correspondence address of the user. Then, each document is assigned a set of selection weights, one for each geographical region. In the case where the geographical region is a state, there will be selection weights for each document in the document database, one for each state.
Other information may also influence what “class” or “classes” that a user of the portal may falls into. The patent tells us that those classes may include the following:
- Date of birth
- Any interests or hobbies
- Primary language spoken by the user
- Political affiliation
- Education level
- Income level
- Academic background and/or;
- How often the user uses the portal
Is the Yahoo portal of the future one that might show very different information to visitors with very different backgrounds and profiles?
If you run a site that could be considered a portal, how frequently do you update information on your site’s front page? How much attention do you pay to how frequently different articles and features are selected on that page?
Would you consider personalizing your main page to show different stories and features based upon aggregated selections of people from different locations, or different genders, or different political backgrounds? Is it possible that at some point, this kind of personalization can go too far?
It will be interesting to see how far Yahoo might go with the personalization of its portal page. It makes sense to show news and application to viewers in India or the UK relevant to those locations. I wonder how I would feel viewing a Yahoo that might show me different news and applications depending upon whether I was male or female, Republican or Democrat, or some other demographic contained in my profile. I don’t think I would like that.