Letting attention guide self-optimising creative.

Two years after the United Nations launched the Global Goals in 2015, a coalition of charities and businesses brought the Goalkeepers campaign to DOOH. The ambition was not only to reach people at scale, but to make complex global challenges feel relevant in everyday life. The campaign ran across streets, stations, and public spaces where people do not arrive intending to watch advertising; they encounter it as part of their daily movement.

The challenge was therefore not just visibility, but resonance. Could a public-space campaign learn which messages actually held attention, and adapt itself accordingly?

To explore this idea, Quividi introduced Campaign Genius, a self-optimising creative system built on an evolutionary algorithm.

As the campaign ran across participating DOOH networks, the system measured how long people paid attention to different creative combinations and message sequences. These attention signals were continuously fed back into the algorithm, which iteratively recomposed the campaign while it was being played, reinforcing sequences that held attention and progressively retiring those that did not.

Creative elements that attracted more sustained attention were shown more often. Those that under-performed were gradually removed. The campaign did not follow a fixed media plan; it evolved in place, screen by screen, guided by how real people responded in real environments.

Across the network, attention increased as the campaign learned. On the best-optimised screens, average attention time rose by up to 32 percent, from 4.2 seconds to 5.5 seconds. Certain creative sequences consistently outperformed others, revealing how subtle changes in rhythm, colour, and ordering influence real-world engagement.

More importantly, the campaign demonstrated something that had rarely been possible in physical media: creative could be refined based on observed human response, not just creative instinct or post-campaign surveys.

In 2017, Campaign Genius was the first AI-based self-optimising campaign engine.