How do you make an impression at SXSW? Start with something authentic (and delicious).
At the Cloud Experience Lab, we were presented with a unique opportunity to build an exhibit at SXSW with our partners at the IBM Design Studio. Simple requirements—create a fun and memorable experience that draws traffic to the event by showcasing IBM’s relevance to a millennial audience.
LET WATSON PICK THE CRAFT BREW FOR YOU
The solution turned out to be something both local and near to our hearts—craft brew. By mashing up a physical beer tasting experience with some digital concepts from the lab—image recognition and smart shelves—we were able to create a new kind of taste-testing experience. Add in a bit of Watson machine learning to predict your favorite local Austin craft brew, and we had a winner on our hands.
The basic experience begins with a user being asked a series of colloquial questions about food preferences. Equipped with this information, Watson makes a recommendation of the three craft brews we think the taster will like best. As the beers are poured and tasted one by one, we used object recognition to trigger the display of blind taste information on a large screen behind the bartender. The taster then arranges the beers on the bar top according to preference. As the final reveal, the name and details of the top beer are unveiled to the taster along with similar local craft brews they may like. The taster’s rankings are are automatically captured by the same object recognition technology, providing feedback to Watson to improve our recommendations over time. Machine learning at its best!
LET’S TAKE A LOOK AT HOW WE BUILT THE EXPERIENCE…
We built an iPad app for the bartender to run the experience, which corresponds in real time to the app seen by the taster.
Each taste reveals a multitude of information on the screen about the brew while still retaining anonymity, including a six-point sensory chart visually highlighting particular individual flavors. Watson’s recommendation confidence is also revealed. Once the taster has tried and ranked all three beers, image recognition reveals additional details of the top pick.
We built custom bars with clear tops and outfitted them with hidden cameras for image recognition. In addition, we created custom display stands to mount the TVs showing all the beer detail. Once those were assembled we gave them a custom paint job and applied vinyl graphics.
DESIGN + DEVELOPMENT OF ANIMATIONS
To display compelling, real-time information about the brew to the taster, we implemented some slick animations in the tvOS app. This was made possible by design prototyping using Flinto, followed by development building them out in Xcode.
FRIDAY, MARCH 11, 2016
1,700 DAYTIME ATTENDEES
700 NIGHTTIME ATTENDEES
167 INDIVIDUALS COMPLETED
501 BEERS POURED
MOST POURED (72): YELLOW ARMADILLO
MOST POPULAR: YELLOW ARMADILLO
Real-time analytics gathered from the duration of experience run at IBM Design Hive.
Post-SXSW, we continue to collect data to further validate a correlation between food flavor preferences and beer. A revised version of the beer tasting experience was installed in a local craft beer store, Whichcraft, for the public to enjoy.