Skip to content

Multi-Modal Data-Driven Design Concept Evaluator

AdaTech, a registered company is using an AI-driven platform to assess customer sentiment, for the development of design concepts

Published: 14th July 2022
Multi-Modal Data-Driven Design Concept Evaluator
Source: NicoElNino,,


Evidence suggests that product design decisions often lead to significant social impacts concerning population change, gender inequality, education, stratification, health and well-being, and human rights, among others. Yet, current product design practices are predominantly focused on economic and environmental impacts, and little is known about the social sustainability of designed products. 

There are several underlying issues with integrating social sustainability and product design. First, existing methods to elicit the needs of users are mostly based on human interaction and assessment activities, such as interviews, surveys, focus groups and design thinking methods, which are limited in the scope of the data collected and rely on human assessment, experience and intuition to develop unique insights. Additionally, bias, a lack of diversity, and limited inclusion are challenges that designers must overcome as they design new products for the global community. The Internet, digital platforms, social media, and mobile devices have enabled users to interact, share opinions, and discuss the products and services they use at scale. However, existing sentiment analysis methods are mostly limited to extracting general favorability metrics (e.g., like/dislike, rating) from user reviews, leaving the fundamental “latent” needs of individuals/society unknown. There is a lack of computational methods to translate these insights into new, explorative ways to increase the quantity, quality, and diversity of new design knowledge and concepts.

Technology Overview

AdaTech envisions a substantial opportunity to devise an AI framework to fundamentally enhance designers’ ability to innovate “socially aware” products by identifying pressing and latent societal needs from social media and e-commerce platforms and automatically generating design concept recommendations for designers informed by those needs.

The platform will assist designers, marketers, and product line managers in the development and evaluation of new products, reducing or eliminating traditional marketing and design resources. The SaaS platform will make this process more efficient and increase the quality, quantity, and diversity of knowledge used in the innovation process. It will use novel natural language processing algorithms based on state-of-the-art language models (e.g., T5) for extracting user needs from online reviews as attribute-level performance aggregates and opinion summaries. Additionally, it will use generative modeling algorithms for incorporating both the visual and functional aspects of past successful designs in new design concepts, by coupling them with a Deep Multimodal Design Evaluation (DMDE) model that can predict how users will rate a new concept based on past sentiment data


  • Significantly cheaper, faster, and less biased than existing methods that rely predominantly on designers’ judgement and/or inherently biased data generated from user interviews and focus groups. 
  • Ability to integrate the user feedback and knowledge from all competing products to inform future designs–existing approaches are limited to the company’s products or at most, to a handful of products selected by designers for comparison 
  • A first-of-its-kind data-driven approach for concept evaluation-existing approaches are subjective and based on the knowledge of the design team or feedback from a small number of retailers or focus group users 
  • Predicts the desirability of a new design concept with over 99% accuracy


  • Concept evaluation in early-stage product development: Predicting the potential success or failure (e.g., desirability among users) of a new design concept with high confidence and accuracy 
  • Generative design: The model can be combined with generative design algorithms to automatically create new concepts with high potential desirability 
  • Data-driven comparison between competing models and new trends 
  • Trend analysis by comparing the predictions obtained from analyzing user reviews from different time periods 


 Testing and validating MVP with pilot partner(s), Connecting with prospective customers

IP Status
  • Provisional patent
  • Patent application submitted
  • Development partner
  • Commercial partner
  • University spin out
  • Seeking investment