Our Remarkable Cases

See some of Hyper Morphing Technologies cases below


CNET is a website, based in America, with all types of multimedia content about technology and electronics. We worked with them on testing the best banner for their website. In this real-world test, we used the actual CNET website, visited by 8.4 million visitors every day. Over 100.000 consumers viewed over 450.000 banners on CNET.com. On the relevant webpages, CNET’s click rates almost doubled relative to control banners.


General Motors

General Motors (GM) is a big, multinational, car company with brands like Opel, Chevrolet and Vauxhall. We did a project with GM to find out if cognitive-based advertising better helps than traditional targeting. In this virtual test, we used laboratory-based websites, that simulate the actual General Motors’ website. The sample consisted randomly recruited panel of consumers. Between this group, every person saw the banner that best fits to his own cognitive-style. The experiment demonstrates that matching banners to cognitive-style segments is feasible and provides significant benefits above and beyond traditional targeting.

BT Group

BT Group is a British communications service company in the UK and in 180 countries around the world. For this project, we ran an experimental BT Group website using data from 835 priming respondents. Every visitor saw a website that best fits to their cognitive style. Given the results, purchase intentions can increase by 20%, which means that if implemented system-wide, such increases represent approximately $80 million in additional value.

Suruga Bank

Suruga Bank is a Japanese commercial bank in the greater Tokyo area. By using a strategy of openness and honesty, Suruga sought to demonstrate that its products (low interest rates, high limits, but a careful screening process) would meet the needs of many customers. Suruga’s managers found website morphing intriguing and authorized a small-scale field experiment. The established morphing algorithm infers latent segments from a preset number of clicks and then selects the best “morph” using expected Gittins indices. Switching costs, potential website exit, and all clicks prior to morphing are ignored in this established algorithm.
In our project with Suruga Bank we evaluated the improved algorithm, including these ignored elements, with synthetic data and with a proof-of-feasibility application to Japanese bank card loans. The proposed algorithm generalizes the established algorithm, is feasible in real time, performs substantially better when tuning parameters are identified from calibration data, and is reasonably robust to misspecification.