How does the content recommendation engine work, and how do I adjust the factors that drive it?
If your site makes use of the content recommendation engine, this article outlines how it works.
Not sure if your site uses a content recommendation engine? Do you have an element on either your homepage or your website's right or left rails that is titled "Recommended" or "You should read ..."? If you aren't sure, talk to your Solutions Manager. Here's an example from ePublishing client WattAgNet.com. To see recommendations, you must have an account on the website.
How it works:
At its simplest: The content recommendation engine tells you as logged-in user which articles users like you have read on the website that you have not yet seen. The engine recommends articles you haven't read yet, but might be interested in because of shared characteristics with similar users.
Here are the factors driving that decision:
- Has the logged-in user already read the article?
If so, he won't be offered the article as a suggestion.
- Are there one or more matching user demographic points for users based on their selections made when they registered or updated their account profiles?
See the example below featuring Derek and Elise for more details on this.
- Other factors: How many users you have on the system, how active they are once they are logged-in, and whether they fill in the user demographic fields on the Registration form / Account Profile form. These all factor into the relevancy of the content recommendation engine's offerings. Clearly the more logged-in users, the more traffic, and the more pageviews, the more useful the results will be.
We have two users:
- Derek, who identified as a CFO. His industry is regional banking.
- Elise, who identified as a CFO. Her industry is insurance.
Both Derek and Elise are active users of the site. Elise logs-in on Monday and reads six articles. Derek logs-in on Tuesday. His You might be interested in ... section of the homepage or in his right rail shows three of the articles Elise read the day before. Derek reads those three articles. When he returns to the site on Wednesday and looks at the You might be interested in ... area, those three articles will no longer be shown. But he will see other articles that Elise, the CFO, has read. Or he will see articles other users read that are in his industry of regional banking. If there are other users who share Derek's job title and industry, their reading habits will be offered as recommendations first.
How to control the content recommendation engine:
You can control the key characteristics that drive the content recommendation engine with the following actions. Assuming you have admin permissions, go to the SysAdmin menu bar and select User Demographics > List / Edit
Select the "Use for content recommendations" taxonomy (1) and then the "Get associated master details" (2).
The resulting list includes the custom user demographic questions that drive the content recommendations.
And now you may be realizing you either only have one type of user data driving recommendations, or you want to change what is being used.
To change or update the data used for content recommendations, assuming you have permissions, go to the SysAdmin menu bar and select User Demographics > List / Edit.
Select the Custom Input (or parent of "Use for content recommendations") taxonomy (1) and then the "Get associated master details" (2).
Click the name of one you know isn't already associated with the content recommendations.
To remove a user demographic used for content recommendation:
In the User Demographic Manager, select "Use for content recommendation" and click "Get associated master details."
The resulting list includes the custom user demographic questions that are driving the content recommendations. Click the name of the one you want to remove.
Scroll down to the Classification taxonomy and de-select the "Use for content recommendations" taxonomy.
You may have to wait at least 15 minutes to start seeing recommendations populate based on your new selections.