Transforming market whitespaces into new product concepts: a text embedding and inversion approach

Conventional approaches to new product development rely on explicit customer needs but fail to capture latent opportunities owing to functional fixedness. While market whitespace-driven approaches address this limitation, current approaches require human experts to manually interpret market maps and generate product concepts, limiting scalability and consistency. This study proposes an automated framework that transforms market whitespaces directly into human-readable product concepts without human interpretation. The proposed approach is composed of four steps. First, product descriptions are preprocessed into standardised formats. Second, descriptions are embedded using text-embedding-ada-002 and compressed via autoencoder into two-dimensional coordinates. Third, market whitespaces are identified through kernel density estimation and grid-based detection. Finally, market whitespaces are inverted using decoder reconstruction and the Vec2Text model to generate interpretable product descriptions. The framework creates a visual product map where market whitespaces represent potential market opportunities, then converts these spatial coordinates back into textual product descriptions. A case study using time management/focus applications from Apple App Store is conducted to demonstrate the effectiveness of the proposed approach. Generated product concepts showed high semantic similarity to real applications subsequently launched in the market, indicating the potential utility of the approach for identifying market opportunities and generating corresponding product concepts.