The Campaign

To modernize “You’re Not You When You’re Hungry” and also make price something worth following outside the store, we created the Hungerithm. A hunger-algorithm that that monitored online mood and lowered SNICKERS® prices accordingly at every 7-Eleven in Australia. The angrier the Internet got, the cheaper SNICKERS® became, with prices dropping to as low as 82% off the normal price.   The Hungerithm was built on a 3,000-word lexicon and analyzed over 14,000 social posts a day to determine sentiment. It even understood slang and sarcasm.   Once mood was established, it was assigned to one of ten price points, ranging from $1.75 AUD (roughly $1.30 USD) when people were “PRETTY CHILL” to 50 cents AUD (roughly 30 cents USD) when they were “LOSING IT”.   The Hungerithm ran live 24/7 (updating 144 times a day) at SNICKERS.com.au, where users simply clicked a button to generate a 7-Eleven barcode right on their phones.

Creative Execution

Using a 3,000-word lexicon that covers 95% of commonly used words in the English language, the Hungerithm determines online mood by analysing over 14,000 social posts a day. It even understands slang, sarcasm and variations in context. Each social post is pulled apart word-by-word and given a score based on its polarity (how negative or positive it is), subjectivity (how objective or subjective it is), and intensity (how strongly each word modifies adjacent words). Once the general mood of the Internet is established, it’s assigned to one of ten mood states and price points. So if people are raging on Twitter (we’re bracing ourselves for Donald Trump being endorsed), the Hungerithm can help by dropping the price of a SNICKERS to as low as 50 cents. To claim their SNICKERS, users simply visit Snickers.com.au on their mobile device and click “Get A Snickers” to generate a unique barcode, redeemable at any 7-Eleven store Australia-wide. No downloading apps or printing vouchers is required.

We evaluated different approaches to sentiment analysis. Rather than an off the shelf tool where we can’t determine what was happening under the hood, we opted for a raw lexicon and used live tweets to tune how we interpreted that information programmatically within our code. We trained up the lexicon to interpret meaning in a place where natural language is limited by the 140 character format of the platform. Fine tuning was critical in ensuring we interpreted combinations of words, shorthands, hashtags and captured the intent of each tweet accurately. This engine also gives us future proofing to allow us to roll out to different countries and languages where nuances of language can be very different. From our database of captured tweets, we curated and plotted the most interesting and amusing ‘angry tweets’ into our visualisation - ensuring we exposed the underlying humanity of what was powering the Hungerithm

The SNICKERS® Hungerithm was the first algorithm to tie online mood to in-store price on a real-time, 24/7 basis. Built upon a 3,000-word lexicon licensed by Massachusetts Institue of Technology, the algorithm analysed over 14,000 social posts a day, word-by-word, to determine the current sentiment. Prices were then updated accordingly 144 times a day at every 7-Eleven in Australia. To lock in a price, users simply visited SNICKERS.com.au on their mobile and clicked “Get a Snickers” to instantly generate a 7-Eleven barcode right on their mobile.