A new resource has been announced on Best Buy’s BBYOpen blog: Metis Alpha. Like Best Buy’s earlier forays into Semantic Web use, this one started with a business problem. As the announcement states: “These days, consumers have a rich variety of products available at their fingertips. A massive product landscape has evolved, but sadly products in this enormous and rich landscape often get flattened to just a price tag. Over time, it seems the product value proposition, variety, descriptions, specifics, and details that make up products have all but disappeared. This presents consumers with a ‘paradox of choice’ where misinformed decisions can lead to poor product selections, and ultimately product returns and customer remorse.”
Metis proposes to tackle this problem by surfacing raw data in a machine-readable format. This alpha release is merely the first phase of several, but it’s a big one. Data about the entire 500K+ active products from BestBuy’s catalog are now available as Semantic Endpoints. Metis uses the GoodRelations vocabulary to offer detailed product information, and even offers reviews from the BestBuy site in the form of unique semantic endpoints (the output is RDF/XML). Including the reviews, this brings the total number of endpoints currently in Metis to more than 1 Million. Phase one also represents a strong effort in Linked Data implementation, as the project achieves four out of five stars on Tim Berners-Lee’s “Five Stars of Linked Data” scale.
Jay Myers, who has the job title of “Emerging Digital Platforms Product Manager — BBY Open” at Best Buy, led a small team to create the Metis project, and when we spoke, he stressed the business benefit of not having to hire or train a big technical group. The team of developers consisted of consultants Brian Sletten, Randy Kahle, and Tom Geudens. Metis was built on the NetKernel platform (Our coverage here).
Myers hinted at future plans for Metis. While phase one was all about publishing the data in a machine-readable format, he expects that in a future phase, the team will build a SPARQL endpoint, and eventually develop systems that would allow external partners and business units within BestBuy to reuse this data. He sees this approach as another way to prove the business value of Linked Data systems. He also anticipates the team being able to add that fifth star as they tie into other data sets on the web.
The starting point for navigating and consuming the catalog is http://metis.bbyopen.com/product/sitemap_index.xml. Within that sitemap are links to several additional sitemaps representing the entire active catalog. As for freshness of the data, the sitemaps are built daily and contain information on when they were last modified, so as the catalog is updated, so is the data at Metis.
If you’re struggling to understand what, exactly, is being offered here, let’s look at an example. BestBuy.com offers the CD, “In the Aeroplane Over the Sea” by the band, Neutral Milk Hotel, and shows a human-readable page at the main, consumer-facing site. The same database that drives that site, also provides the data to Metis, which represents the same information in a unique product semantic endpoint (based on SKU), and another endpoint for the review.
As Metis gains traction, Myers and team are looking for developers to experiment and provide feedback on the data and the implementation. For service updates, questions, errors, or other general feedback on Metis, see @BBYMetis on Twitter.