Ipsos enhances AI Pack Sprint: AI and its experts evaluate packaging together

A predictive model supplemented by the SPEED analysis grid

The automated component simulates buyers' attention and their immediate reactions to a package, compared with those of the main competitors. It evaluates each project on more than 250 variables and calculates its probability of success based on more than 30,000 behavioural data points.

For Ipsos, artificial intelligence "can bring speed" but, without the input of human experts, it "cannot confidently take into account" the brand's specific objectives, the intent behind the design or the evolution of category codes.

Experts therefore review the results using the proprietary SPEED grid, which covers five criteria: ability to capture attention, immediate brand recognition, clarity of information, emotional appeal and purchase incentive. They then interpret the scores in light of the brand's objectives, category trends and the competitive environment, in order to translate them into operational recommendations.

A client case and reliability indicators put forward by Ipsos

The group illustrates this approach with TéPone, an alcoholic iced tea launched in Spain in 2025, which was to move a packaging designed for festivals toward a format suited to distribution.

Within a few days, the tool identified several obstacles, including limited shelf visibility, a logo that was too dominant and insufficient visual cues for the product category.

According to Ipsos, these findings guided a redesign, and the new packaging achieved a purchase rate 3.5 times higher than the initial design. The group claims to have been the first research agency to launch, in 2018, an AI model for packaging pre-selection, tested alongside consumer surveys with an accuracy of 85%.

Ipsos states that it carries out more than 500 packaging studies each year and draws on more than 50 years of behavioural expertise. It argues that an early evaluation can raise the success rate of packaging redesigns from 30% to 70%.