A couple years ago, I wrote a series on metadata. For those unfamiliar with the term, metadata is data about data. A sweater’s metadata, for example, includes its color, available sizes, fabric, and style. A computer’s metadata includes information about the components inside, speed, amount of memory, and size of its hard drive.
Metadata can be used to simplify product categories by creating search mechanisms that allow the user to refine a search by attributes while staying in the same category. Instead of having a separate category for 24k gold necklaces that are 13 in. long, for instance, simply have a “necklaces” category and let the user filter the products by metal type, karat, and length.
Metadata can also be used for personalization by correlating product attributes together in an effort to effectively cross- and up-sell products that are similar or complementary to the product the user’s viewing.
At the time, I wrote about the advanced personalization that can be performed with metadata. But there were no off-the-shelf tools to make this process easy. We’d just finished building our own tool for a multichannel retail client. Now that’s changed. I recently saw a sneak preview of Intelligent Cross-Sell, a new product from CNET Channel, a subsidiary of CNET.
Most people know CNET as a source for product reviews and detailed product specs. CNET Channel is its own business that feeds product data to such retailers as CDW, Amazon.com, and Best Buy. Because it has a vast collection of metadata on every electronics product, importing information about products from CNET Channel is much cheaper for these companies than creating it from scratch.
CNET’s subsidiary is going beyond merely providing the data. Recognizing the power inherent in metadata, it’s creating applications that use it. Most electronics companies already use its product comparison feature. It lists the metadata of various products next to each other. Intelligent Cross-Sell is a huge step toward effective one-to-one personalization using metadata.
In essence, it’s a large, rules-based engine whose sole purpose is to recommend products based on complex metadata filters. An electronics Web site might want to recommend products that would be good cross-sells for a computer, for example. Although most retailers use traditional technologies for collaborative filtering, some are now realizing the potential of using metadata for these recommendations. In the collaborative filtering world, recommendations are based on what other people bought with the computer. This may or may not include products that really are a good fit.
By harnessing metadata, retailers can present products that really fit with the current product. They can present exactly the correct memory that’s compatible with the computer because the system knows how much memory the computer has on board, how many expansion slots it has for more memory, and what size chips it takes. Further, the system knows which memory chips and brands are compatible with each motherboard and what their speeds are. So the system can generate intelligent recommendations. Moreover, the system can explain to the user why the product’s a perfect match. The rules-based engine can ensure certain brands are preferred over others if needed or any other business rule that goes beyond the actual data.
Though the CNET Channel product is currently filled with data about electronics, its system allows users to import other data. A company called Muze provides retailers with metadata about books, music, and movies. Muze would be wise to look into this platform for current customers.
Having spent several years as director of personalization at barnesandnoble.com, I can immediately think of easy applications. When my team created the “People who bought this product also bought…” functionality, we very carefully analyzed the importance of metadata over various book categories. The publisher is very important to techies buying books on programming, for example. The author is very important to people buying fiction. Genre is very important to people buying nonfiction. This method of weighing metadata enabled us to create lists of cross-sells that were much more targeted based on the kind of book in question than collaborative filtering or simple statistical analysis could provide.
If we could have imported that information into CNET Channel’s rules-based engine, I suspect the process would have been greatly simplified. We could have also created internal lists of New York Times Best Sellers, Oprah’s book lists, award winners, books currently in the news, and the like and weighted recommendations based on that data. We could have added business rules on top of that to harness even finer metadata (such as not recommending books that are too heavy and would therefore increase shipping costs to the extent where we’d risk losing the sale).
I’m very excited about the emerging tools specifically geared toward merchandising. There have been metadata search engines before (like Verity, which allows a weighted search similar to my barnesandnoble.com example), but these tended to be large applications not geared toward ease of use for merchandisers. Because these new tools are Web-based, they’re easy to integrate and don’t require the IT infrastructure larger metadata databases do. In the past, we had to build these applications ourselves. That’s costly and requires IT staff maintenance.
These tools are easier for marketers and merchandisers to use and are modularized to solve specific problems. The end result is the ability to quickly harness the power of metadata in ways we’ve been dreaming about for years.
Questions, thoughts, comments? Let me know.
Until next time…
When measuring the effectiveness of discount codes, retailers often get it wrong. In this article, we'll look at how data-driven attribution can help businesses better understand where discount codes produce the best ROI.
Data. It’s the latest ‘buzzword’ in the digital marketing world when it comes to content.
Digital has quite forcefully overturned the entire media industry, causing even the most traditional companies to adapt or be left behind.
What’s behind a successful data-driven marketing strategy?