Best Buy LogoJay Myers, Lead Web Development Engineer at BestBuy, has moved the proverbial ball forward yet again by creating an implementation of the vocabulary in BestBuy’s Black Friday web pages.

First, a bit of history…

Myers began incorporating structured data into BestBuy web pages in 2009. Starting initially with basic store information (hours of operation, location, contact information), Myers soon expanded the project to include product pages, music data, and the 600,000+ item product catalog. This work quickly became a widely cited use-case for semantic markup. In particular, it brought a lot of attention to the RDFa syntax and the GoodRelations vocabulary. The effort resulted in improved page rankings, richer display of BestBuy search listings in browsers, and — after putting user-friendly tools in the hands of store managers —  enabled Myers to tackle the retail problem of Open Box returns.

Timeline of Semantic Markup at BestBuy

Recently, Myers announced that he has started experimenting with, a vocabulary released by Google, Yahoo and Microsoft last June. uses the Microdata syntax. We spoke to Jay about this work to learn what it was like to implement and the progression of his Semantic Web markup work. What led you to try and Microdata?
Jay Myers: I was fortunate enough to participate in the workshop in Mountain View last month where a bunch of really smart people were talking structured data on the web. If you know anything about the history between some of the individuals attending, you’d figure we would have several opposing viewpoints and many arguments would ensue. To my surprise, this was not the case — we had a great day of very constructive talks. And with this warm and fuzzy spirit of goodwill, I figured it was time to put the rubber to the road and release a new standard for all to test.

SW: So, what have you done to “put the rubber to the road?”
JM: We saw an opportunity to get a live working example out on our site through a Ratings and Reviews enhancement project that was already being worked on at the time of the announcement. This project was a natural fit; the [Google] Rich Snippets and [Yahoo!] Search Monkey teams had laid a very solid foundation a while back by constructing and supporting reviews vocabularies and markup examples even before This established vocabulary was largely adopted by the new standards, so it was relatively easy to implement.

SW: What did you expect when you set out to tackle this and what did you experience?
JM: I expected there would be a slight learning curve — as a developer I have found that coding with data and presentation in mind is much different than just presentation. It’s not that the markup is more difficult, it puts me in a different mindset delivering human readable code while visualizing what a machine might get out of the rich markup.

SW: How smooth was that work compared to your earlier efforts? Were you creating new markup or modifying existing markup to make it work with
JM: In terms of time, we spent around 2 days creating, testing and validating the code before deploying. This markup replaced a prior solution using “plain old” HTML. Rather than try to retrofit, we worked from scratch to create new code. The beauty of good data is that can be shaped into many forms. In the case of this solution, we had the raw data and simply needed to use a transformation mechanism (for us it’s NetKernel) to make it human and machine readable.

SW: You had done so much work with RDFa previously. Were you able to build on that work?
JM: One of the themes of my announcement was that this solution was about the schema, not the syntax. That being said, I did find that the Microdata syntax was straightforward to implement. I am still a massive RDFa fan, and am really excited about RDFa 1.1 Lite. I have been reading the editor’s draft and have plans to deploy it into an upcoming solution.

SW: What challenges did you encounter?
JM: I work with a number of very talented developers on an extremely large codebase. The biggest challenge is keeping the code pristine to ensure that machines can parse it and create accurate RDF triples. With a bunch of cooks in the kitchen it can be hard to maintain more stringent markup standards. In addition to experimenting with new vocabularies and syntax, I also need to continue to educate my fellow developers to be cognisant of the impact and benefits that this markup can provide.

SW: What benefit(s) do you hope/expect to see? How will you measure success?
JM: I hope that consumers will benefit from this valuable data being more visible on the web. In commerce, reviews are vital to our customers to assist in making an informed purchasing decision. I see challenges in getting this data to customers. First, there is an abundance of review data on the web for us to utilize, but so much that we humans may not be able to consume it all. Second, implementation of reviews data on the web is often inaccessible to machines. In my opinion, these challenges can be addressed by laying a foundation of rich structured data — something that can be accomplished by any dev, given they take a little extra time to be mindful of their markup. For a brief moment we have seen search engine results pages display the three latest reviews for some of our products — I feel that is very beneficial to shoppers.

I also expect that the Semantic Web community will take these production examples and provide feedback in order to make things better. There are great things happening in groups like the HTML Data Task Force and Web Schemas Task Force that should be able to benefit from live examples.

In the holiday spirit of giving I am excited to reveal our next rich data experiment: look for the Best Buy “doorbuster” page before the US Thanksgiving holiday.  I am happy to announce our 2011 Black Friday “Doorbusters” page contains enhanced markup, utilizing for one set of products and for a second set of products. These can be found at the BestBuy Doorbusters page.

Happy parsing!