Responses by Madison Palasini, copywriter and Haihang Du, lead developer, Watson Design Group, Inc.

Background: Though it’s now over, Mystery Box Live! was a wedding-friendly guessing game that was live every day for a week. Each day, a new box was shown and the first person to guess what’s inside won a fancy wedding prize. Since we worked with retailer The Black Tux in partnership with The Knot, the game was targeted towards the wedding audience. Ultimately it’s a fun mobile experience that challenges the imagination and rewards those who think outside the box.

Highlights: In the splash page, the objects animate with rotating text slides leading users to sign up and play. We also created a way of rendering mechanics to offer different color styles each day programmatically. That way, the whole site’s look and feel, including the run-time shading of WebGL objects, will all respond to match that day’s unique color theme. The mobile experience also included custom special shoot assets and a group of floating WebGL objects as the core design elements.

Challenges: Keeping each game fresh for users and keeping up with traffic spikes. Unlimited guess submissions and live-streamed pop-ups displaying all players’ guesses also made the server super busy. We reduced the ajax calls and leveraged Socket.io to deal with that real-time data to maintain a smooth, seamless experience.

Favorite details: We were able to combine meticulously-styled special shoot assets and custom design elements to create an engaging experience that encourages users to dream big, think differently and have some fun while they’re at it.

Anything new: We gained insights on server stresses, spike traffic solutions and efficient WebGL animations that we had to optimize for a variety of mobile devices. We also learned a lot about how to position clues to keep the game fun, yet challenging. People are very smart and resourceful.

Technical features: On the server side, our site stack includes EC2, DynamoDB and Redis. To accommodate spike traffic, we had one-second TTL cache on most of our endpoints. We also auto-scaled the app servers whenever the CPU reached 50 percent utilization.