MARS - Snickers Hungerithm, stock exchange for chocolate bars

The Hungerithm is like a stock exchange for Snickers chocolate bars. Prices of the bars fluctuate in real-time, based on the online mood of the internet.

When anger goes up, Snickers prices goes down at every 7-Eleven stores. Prices are updated 144 times a day, and drop as low as 82% off the normal shelf price.

My role

As Technical Director, my role involved to:

Liaise with internal and external stakeholders to help with creative, scoping, client pitches, and global markets negotiations.

Design and architect the end-to-end solution for global rollout across Australia, America, Russia, Europe and South-East Asia.

Lead a small team of x3 developers and a tester to build, test and deploy the project worldwide.

Prototype solutions for the unknowns and the most complex parts of the build.

Key results

67% sales increase in Australia.

30 millions media impression and 1,740% rise in social traffic after the initial launch.

Won multiple awards, including a Gold Cannes Lion and a Webby Award.

Technology

Built upon Google Deep Learning models for Natural Language, the Hungerithm is a bespoke service scanning over 14,000 Twitter posts a day. The mood of the internet, affecting the chocolate bars price in real-time, is calculated by extracting sentiment and processing the output into a bespoke algorithm that can normally distribute and deviate results over time. It can understand slang, sarcasm, emoji’s and differences in context.

The interactive client-side interface was built with VueJS and allowed customers to monitor fluctuating prices and redeem coupons in real-time. Engineered around AA accessibility requirements and the ability to switch locales, currencies, and reading direction based on local market needs. A bespoke interface was also created for community managers to moderate all public content appearing on the client-side interface.

The infrastructure was built and deployed on Amazon AWS using infrastructure-as-code. With a centralised GraphQL API and a serverless architecture leveraging services likes Fargate, Aurora, Lambda, S3 and CloudFront. Designed to be easily deployed in different regions and markets.