Reported execution: the discount followed the mood
Clemenger BBDO describes Hungerithm as a live Australian promotion for Snickers and 7-Eleven. A custom system read social posts, classified the internet's mood and changed the price offered through the campaign. D&AD records a 3,000-word lexicon, more than 14,000 posts analysed daily and phone-based barcode redemption. The campaign also used dynamic display, social and other media to make a fluctuating price visible.
That is a more consequential use of an algorithm than a targeting footnote. The offer consumers saw could change because the system had made a reading of public language. The brand's long-running hunger line supplied the logic: if the internet was losing its temper, a cheaper bar became the remedy.
Our reading: make the model legible in the offer
The useful creative move is not that a machine measured sentiment. It is that the measurement altered something a person could understand and redeem. A live score with no consequence would have been a novelty; a price makes the same data into a compact proposition.
For a comparable brief, the audience should be able to explain the rule without an explainer video. What signal changes? How often? What does a change unlock? That clarity also makes the system easier to challenge when its reading feels strange or unfair.
Operational lesson: a live mechanic needs a fallback
Mood classification, especially around slang, irony and breaking news, is a noisy input. The case describes human-designed language rules and retail redemption infrastructure; neither should be treated as invisible plumbing. Teams should decide in advance how extreme or ambiguous inputs are handled, who can pause the system, and whether the advertised price can be honored at every participating store.
The media plan matters too. A changing offer needs its time sensitivity communicated without creating a promise that an old screenshot can still fulfill. Versioned terms, a clear redemption window and a support route are part of the creative's credibility.
Evidence boundary: a good mechanism is not a universal result
The agency and award accounts report strong sales and engagement outcomes. Those figures are useful evidence of what the campaign participants say happened, but the public records do not provide a counterfactual that isolates the algorithm from discount depth, retailer reach, existing brand equity or media weight.
The durable case is therefore not ‘sentiment always sells chocolate.’ It is a documented example of algorithmic data becoming a consumer-facing rule rather than merely an optimization claim.
Sources & limits
Agency case study and award record; reported sales and reach figures are campaign claims, not an independent causal audit.
- Clemenger BBDO — Hungerithm — source publication date not stated.
- D&AD — Hungerithm — source publication date not stated.
Send a correction with the passage and supporting source.