SUCCESS STORY

Smart Creatives: real-time banner optimization

Display the most relevant banner according to the context thanks to Artificial Intelligence.

The challenge

Context

For La Banque Postale Consumer Finance, presenting its different products through different visuals and choosing carefully which to show is a key issue. Indeed, the banner displayed plays a vital role in the ability to convince future customers.

Faced with the disappearance of tracking methods linked to third-party cookies, maintaining this capacity for congruence becomes a major challenge: how to adapt a banner’s choice in the most granular way possible, in order to maximize its impact on the user, but without relying on the 3rd-party cookie?

Goals

  • Optimize the campaign in real time, by displaying the right banner according to the target and the distribution context.
  • Use an optimizing method of the displayed banner that does not rely on third-party cookies.

The solution

In order to increase the performance of the display campaign through banner optimization, Numberly has implemented its Smart Creatives solution for La Banque Postale Consumer Finance.

This solution uses advanced artificial intelligence models to decide, at each auction, which banner to display based on the context.
The algorithm learns continuously, depending on the broadcast environment (subject of the web page, time of day, etc.), which banner works best, in order to choose the best banner auction by auction.

Smart Creatives thus provides numerous advantages:

  • Continuous optimization (data-driven and performance-oriented) and in real-time (decision in a few milliseconds) via a reinforcement learning algorithm.
  • A 100% cookieless solution, based solely on clicks and contextual information, hence maximizing the congruence effect (the match between the advertising message and the environment in which it is presented).
  • A transparent learning model: analyses are made available to better understand the banner/message/visual choices made by the AI depending on the context.
  • Flexibility and power significantly greater than traditional A/B testing.

The results

The results were very satisfactory, making it possible to significantly increase qualified user engagement with the banners:

+16 %

click-through rate compared to random distribution of the same banners

+ 22 %

rate of engaged post-click visits (visits lasting more than 5 seconds or with scrolling or clicking on a button on the landing page, having taken place after a click on a banner)

Discover the case in less than 2 minutes! (subtitles available in English)