Exploring visual representations of hotel brand image: framework development using mixed methods

Purpose This study aims to achieve two purposes: to construct multilevel visual representations regarding how a hotel brand is portrayed or perceived via various pictorial attributes and to develop a quantitative measurement framework that identifies these pictorial attributes by using deep learning models. Design/methodology/approach A mixed-methods approach was used across three phases. First, two pretrained deep learning models were used to identify manifest and photography-level attributes. The second phase involved qualitative surveys and quantitative techniques to generate attributes related to latent cognitive perceptions and brand personality. In the third phase, a labeled data set was created, and VGG16 models were trained to automatically identify these higher-level attributes from images. Findings A four-level conceptual framework was constructed to link hotel-generated photos with brand image perception. In addition, a measurement framework was developed using multiple deep learning algorithms (e.g. Place365, NIMA and VGG16) to identify and classify these pictorial attributes. Practical implications It offers actionable tools for marketers to strategically use photographs to project a distinct brand image and craft effective marketing communications on social media platforms, helping attract target audiences and differentiate their brand in a competitive market. Originality/value This multidisciplinary study innovatively integrates qualitative methods and deep learning to generate theoretical insights into the visual representation of hotel brand image. It bridges the gap in understanding how photos can be leveraged to develop brand image, strengthen brand competitiveness and improve the effectiveness of social media content.