Creating Successful Personalization
Updated · Apr 25, 2001
Personalization gets a lot of press. When done properly, it can be great. When done poorly, it can be devastating. Unfortunately, the difficult part is doing it at all.
There are two kinds of personalization: overt and covert. Overt personalization is the "Hello, Alexis. Welcome back" that you see when you return to a site at which you’ve created a membership. Covert personalization can be responsible for a site showing you a book about pool and yard care if you’ve ever (even once, a very long time ago) purchased a book about pool and yard care.
Tread Lightly with Personalization
Both kinds of personalization can get you into trouble with privacy advocates. In February, I wrote about the Spyware Control and Privacy Protection Act, legislation proposed by Senator John Edwards (D-NC) (see Technology Is the Answer). Even performing basic clickstream analysis, which is required to do covert personalization, can get you into trouble. Heck, the term "spyware" refers to Web server logging! Anyone who has Web server logging turned on is running spyware, according to this tech-savvy congressman.
However, merchants are wise to inform customers as to what is happening with their data in publicly posted privacy policies — the Federal Trade Commission has fined companies with misleading or inaccurate privacy policies. Yet, it’s hard to believe that a brick-and-mortar store is entitled to count me as I enter and leave, and monitor my activity via video cameras, but an online store isn’t allowed to track comparable customer behavior with appropriate technologies.
AMR Research Exposes the Myth of One-to-One Marketing
AMR Research published a brief in February 2001 entitled The Myth of One-to-One Marketing. In this interesting report, AMR Research suggests that sites use pattern strategy based on overt personalization rather than on covert personalization. Pattern strategy is AMR’s term for collaborative filtering – or aggregating customers based on behavior and/or other information. They lament the presumptive results that covert personalization yields.
The basic premise is that aggregated data based on information provided voluntarily by customers is more accurate than information derived from click-stream analysis. The report also acknowledges that many installations of personalization fail – meaning that sites purchase the software never to see it implemented. However, the report doesn’t address how much easier it is to implement collaborative filtering (pattern recognition) on overt data than on covert data.
Avoid the Data-Silo Problem
How personalization would work under the AMR Research scenario is that visitors and customers would be asked to provide personal information. The site then would aggregate this information on some basis, and would deliver personalization to visitors based on how they fit into the defined groups. Presumably, some degree of clickstream analysis is going to be required to add value to the data that the group voluntarily provides.
This report highlights the idea that data silos need to communicate with each other – clickstream data needs to be integrated with customer data and the entire data set needs to be evaluated for purposes of personalization – whether it’s called pattern recognition or collaborative filtering. Companies that are already collecting personal information voluntarily are halfway to providing effective personalization.
Alexis D. Gutzman is an E-commerce Technology Author and Consultant and author of The HTML 4 Bible, FrontPage 2000 Answers!, and ColdFusion 4 for Dummies. Her newest book, The E-commerce Arsenal: 12 Technologies You Need to Prevail in the Digital Arena is now available. She can be reached at [email protected]
Reprinted from ECommerce Guide