SKIM presented new findings on effect of visualization modes on choice behavior at 2013 ART Forum, Chicago

SKIM took part in the annual Advanced Research Techniques Forum for 2013 in Chicago from June 9-12. At the previous year’s ART conference in 2012, the audience received our presentation on the effect of visualization modes on choice behavior with great enthusiasm. To us it demonstrated that research into visualization is a current topic and it motivated us to further explore this field. This time, we took on a different aspect of visual communication: advertisement and package design.

Generating perfect package designs by using dynamic optimization methods

Presented by: Carlo Borghi, Eline van der Gaast, Virginie Jesionka and Gerard Loosschilder from SKIM

http://www.slideshare.net/SKIMgroup/dynamically-converging-to-the-best-package-designs-skim-at-art-2013-submitted-6-june2013

If you are unable to view the slides on this page or you’d like to download the presentation file, click here to view our slides on SlideShare.

The problem – how to design an optimal package?

Imagine you are to design the best possible package for a new feline medication. You may want to use a few inspiring images showing an expensive, precious pet and an elegant package design. You will try to entice your customers by using well-chosen claims about the product qualities, for example its long-lasting effect. The number of elements – claims, pictures, and so on – and the way you can put them together to come up with a product design – varying size, position, and so on – can easily be infinite.

When facing this problem, traditional research techniques revolve around either a dangerous reduction of the parameter space in the research design (at the risk of missing the real optimum) or a lack of granularity in the most promising area of the parameter space, with many observations collected about irrelevant combinations.

The solution

To overcome these limitations, we developed a dynamic, on-the-fly optimization algorithm that combines traditional experimental research with an optimization learning process. We use this algorithm during data collection to:

• Explore the vast parameter space of all possible packages, while being able to account for heterogeneity in consumer preferences with a variety of optima that might exist – where each optimum represents the preference of a segment.
• Set the path through the solution space and generate new concepts based on previously collected answers, directing the search to obtain a higher granularity with more potential
• Gauge market impact by collecting absolute attractiveness scores, as opposed to common choice-based exercises

Results validation

To validate our algorithm, we have collected consumer survey data relative to a currently available product. First, we have tested whether the optimal design identified by the algorithm performs better than the current one. Furthermore, we have tested whether our algorithm is more effective than the expert’s opinion, by testing the algorithm optimum versus a set of designs created by a team of communication experts.