Ward W.

When we said we wanted to personalize XboxOne, we meant much more than just allowing our customers the ability to select their favorite color. For us, personalization meant giving customers quick access to their favorites with deep-linked Pins. It meant solving the cold-start problem of an engaging experience from Day Zero. And it meant growing that experience by adopting to their implicit preferences and likes over time. While XboxOne was inarguably built as a game console, we were determined to also make it a multi-media platform.

As such, when we began to imagine XboxOne, we concentrated our efforts on how our customers discover and enjoy content, regardless of whether it is a game, a movie, or a song. Historically, many sites approach the problem of a large catalog by slapping on search (aka solve for the lean-forward user intent) and calling it a day, letting you find the content. But that is not enough. Some days you do not know what you want (aka lean-back user intent) and you want the content to find you. Again, many sites slap on a taxonomy for customers to browse, but this can be unwieldly for a large corpus of content types and genres, not to mention that how you construct your taxonomy varies by culture. Even when using sophisticated browse approaches — such as recommendations or editorial picks — any single approach cannot support every user intent. Our challenge, then, was to figure out how to harmonize multiple discovery vectors into a single, cohesive experience.

Beyond the aforementioned search and browse vectors, we wanted to include a variety of new discovery vectors. These included editorial picks, recommendations, and social suggestions, to name just a few. We knew from experience that each discovery vector worked better for certain media-types, and could vary by customer. We were aware that some customers were motivated by top hits, while others would watch a movie a friend had watched. In order for us to truly provide a personalized experience, we needed to be able to boost one vector at the expense of another based on a customer's behavior. For example, if you tended to watch sci-fi movies on Friday nights, we combined this with the fact that you converted from a top 10 list better than a recommendation to ensure we placed "top 10 sci-fi movies" front and center.

This is where I entered the picture. During the design of XboxOne, I played a key role in promoting this next-level personalization on both the services and experiential sides. I introduced the abstraction of an ascriptive spectrum that was necessary to support these different discovery vectors. More specifically, we created a syntax that allowed one side of the spectrum (perscriptive vectors such as Editorial) to be interwoven with the other side (descriptive vectors such as Recommendations). I championed — and led the team that eventually delivered — the architecture that now powers Xbox Home, Pins, Store, and Search.

As noted previously, the goal that I set for my team was to find a way to bring together any number of disparate discovery vectors ( e.g. search, recommendations, top 10, editorial picks) and mix them together in a way that allowed us to change (boost or decrease) a signal based upon A/B testing. An additional goal was to do this entirely from the services side, thus obviating the need for an XboxOne flash update whenever we wanted to change the console experience, as is still the case with Xbox360.

We achieved these goals to the extent that — from our customers perspective — it is now perceived as a single unified user experience ,