Amazon.com filed a patent application on Aug. 10 for a process that would allow Amazon customers to use the retail Web site to gather information on other customers, including "birthday, interests, occupation, education level, income level, location, race, ethnicity, religion and sexual orientation."
That information would be obtained by Amazon through several means, including "from a user profile for the recipient, from past ordering patterns of the recipient or from publicly accessible databases" and Amazon would then make recommendations based on "appropriate recipients for the items (e.g., religion and race)," the patent filing said.
Amazon officials have been careful to stress that the company has no current plans to deploy such a system.
Its also fair to point out that large companies typically file a huge number of such patent applications with the U.S. Patent and Trademark Office, to protect their intellectual property and, more importantly, to preserve their future options. Its not unusual for a large company to have drawers full of patent applications that it will never use.
That said, the Amazon filing is not likely to sit well with privacy advocates.
One such advocate is Liz McIntyre, co-author of a book arguing against RFID usage ("Spychips: How Major Corporations and Government Plan to Track Your Every Move with RFID") that is being sold on Amazon.
"Consumers should always think twice before purchasing anything with an identifiable form of payment or offering up any personal information," McIntyre said during a phone interview with eWEEK. "This Amazon patent application illustrates why consumers need to think very carefully about how they may be leaking details today that could come back to bite them tomorrow."
McIntyre went so far as to say that she and her co-author are considering yanking their book link that goes to Amazon and "searching for a more privacy-friendly online bookseller, if Amazon cannot guarantee the sanctity of customer information."
"I am relieved that Amazon is not using this system, but I am concerned that the data it collects could be put to new uses in the future," she said. "After reading this patent application, my co-author and I are rethinking the link to Amazon on our Web site and considering ways to warn our membership about Amazons intentions and practices."
What the Amazon application discusses is something called gift clustering, described as being used when one Amazon customer is trying to buy a gift for another Amazon customer. Clustering refers to gifts that have been grouped for people with certain preferences. The filing gives an example: "If gift cluster 122 with the moniker Dog Owners is selected, then that gift cluster will become the current gift cluster and the Dog Owners moniker will replace the moniker Moms Birthday Present." (Sentimentality is not a strong suit of Amazon Patent writers, apparently.)
On one level, the associated purchased technique is similar to what Amazon already does with its product recommendation system, or what Tivo does by suggesting television shows based on a consumers viewing patterns and history.
The flaw with Amazons methodology is that it cant distinguish whether a consumer is buying a product for personal use or for someone else. That flaw is not a big problem today because most consumers can filter out obviously inappropriate recommendations.
But it might be a more significant problem when Amazon recommends gifts for other people than when it recommends a product for the consumers own use. A third party might not know that its the intended gift recipients niece who loves ballet and baking. Without that knowledge, an analysis of the purchase trends might be misleading or error-rich.
The patent filing gets quite specific about the methodology of the data collection, saying, "The system can in some embodiments automatically search for appropriate user-defined gift clusters. For example, even if a customer does not know demographic information or interests of a possible recipient, the system may be able to access such information (e.g., from a user profile for the recipient, from past ordering patterns of the recipient, or from publicly accessible databases). … If so, the system could receive an indication of a recipient, access relevant identifying or categorization information about the recipient, and automatically search for gift clusters that match the accessed information."
The filing offers an example: "Some or all individual items may have categorization information associated with them (e.g., a toy with a suggested age range or a gender-specific health product) and, if so, the categorization information for the items in a gift cluster could be combined to create an aggregate categorization for the gift cluster.
"Alternately, the system could track information over time about the recipients of a gift cluster (e.g., from information specified during searches or from user profiles for recipients), and could aggregate the information about the various past recipients in order to determine a categorization for the gift cluster to assist in identifying future recipients."
Retail Center Editor Evan Schuman can be reached at Evan_Schuman@ziffdavis.com.