Linking Real Estate Content With User Context: Matching Content With A Known ContextMarch 30, 2007
Part 1 of this series focused on defining basic concepts and associating context with content in the form of metadata. Part 2 examined user context and a few ways to determine what a user is looking for. This third part of the series will examine how we can match user desires in the form of a user context and prioritize delivery of content that matches the requested context.
In each of the previous articles, we’ve looked at a single step. Let’s look at the entire process to understand how our previous learning fits in and how we should proceed:
You should instantly recognize steps #1 and #2 from the previous readings. At this point, you should know what contexts are important to your real estate site. You should also be able to assign metadata to your content elements in your database and have some idea about the concept of user context. The content itself can be anything (micromarketed ads, articles, links, videos, anything) We’ll move from concept to implementation apply that knowledge below.
Implementation begins with an action to initiate the filtering process or a requirement for contextual information. Some of these will be obvious…perhaps others are not.
– Contextual Choices Integrated Into Various Site Interfaces: The website can have contextual choices directly integrated into the interface via checkboxes, radio buttons, or dropdown menus. Some logical areas would be the search mechanism and the interface for user preferences.
– Context Derived From Searching Within The Site Or From External Sites: Searches within a site will tell you quite a bit about a user if you pay attention. If a user searches for “staging” using your site’s search, there is likely some sort of “seller” context that can be derived (or guessed at) from that action. But those are not the only searches that you can leverage for context. Google and many other search engines provide sites with the terms that visitors used to find their site. You can determine context from at least some of these terms. Perhaps there are others but other than us, the mioaklandcounty blog is the only real estate blog that I know of that is currently using this approach (do a search for “oakland county blog” – choose Maureen’s site and look at the top portion of the site to suggest related articles based on your search terms). We take things a step further by actually delivering a different set of ads, articles, videos, and content if we are able to discern a likely context from Google searches.
– Context Derived From User Actions: If you track all of the user actions during their time on your time, you’ll find likely clues as to their context. These clues can be categorized as “direct” and “indirect” indicators.
Direct indicators are choices that users make that are immediate and persistent. On our site, language is one of these direct indicators. People may experiment a bit with the various language selections but generally stay withn their familiar language context. There are many others.
Indirect indicators take time to establish as likely contextual indicators. Just because a new visitor clicked on a mortgage article does not necessarily make him or her a contextual “buyer”. However, if they are consistently clicking on buyer focused articles/ads, then after critical, it may be safe to think of them as being “buyer” likely. Think of this approach like counting cards at a blackjack table. Counting every card is beyond the intellectual capacity of most of us. However, keeping track of face cards and 10’s with respect to non-face cards is a much easier endeavor…and in doing so, there may be hands where you *know* that the probabilities of one type of card or another is higher. You could still easily lose but its worth the attempt. It’s the same concept with indirect contextual analysis of user actions….every once in a while, the likelihood of a given concept just seems higher. We actually use a very simple system to track a user’s previous habits that is very similar to the card counting analogy above.
– Use of a Default Context For a Site: If you are one of those people like us that have a number of domains in use, you can easily assign a default context for sites. For instance, we manage beta sites that have contexts such as “spanish speakers selling properties in Country Y”, a sister site that is “english speakers looking for properties in Country Y”, and another site that has the default context of “spanish speakers looking for properties in Country Y”. Our search data shows that users generally find the site that has the correct context for them based on their search patterns. Makes context determination REALLY easy when visitors just land on the site that they are looking for.
With any of these approaches, the final part of step #3 is apply the user context to the contextual metadata in order to have a filtered group of suggested content. This activity can occur at any one of several points:
– after processing of internal search results
– as a page request is made from a major search engine but before the page is rendered
– as a page request is made for the main page with no other indicators and before the page is rendered
Some “Don’t Forget”s
Some other points that are important for implemention:
– “ALL” can be an important context as well…dont forget that there may be content required for everyone. This context will be reflected in your SQL “WHERE” clause with an OR. (SQL example: “SELECT this FROM that WHERE context=’buyer’ OR context=”ALL'”)
– Date is important as an ordering mechanism because you’ll often want to present the freshest content. However, there may be times when other factors affecting relevancy trump the date. For instance, an alternative might be to assign a hypohetical priority to content such that date isnt the primary ordering mechanism. This will normally be reflected in your SQL “ORDER BY” clause. (SQL example: “SELECT this FROM that WHERE context=’buyer’ OR context=’ALL’ ORDER BY priority, date”) there are many ways to expand on this line of thinking….
Thats it! Now we have a set of disparate data that is related only by their context that is ready for use within a user interface.We’ll look at how to apply this data in the next article.