Gijs Overgoor Research

In his research Gijs Overgoor focuses on applying techniques from AI and Econometrics to solve marketing problems. His latest work, published in the California Management Review, discusses a framework for implementing Marketing AI projects. Other work in progress is focused on consumer interaction with visual content on websites and social media. 

Gijs’ research was featured here. It provides an excellent summary of some of the questions asked in his research. Based on this article he was also invited to speak at NPO Radio 1. 

Gijs was featured on The World Transformed podcast to talk about his work on image analytics in marketing.  Click here to listen.

Publications

+ G. Overgoor, M. Chica, W. Rand, A. Weishampel (2019), ”Letting the Computers Take Over: Using AI to Solve Marketing Problems”, in press at California Management Review. https://doi.org/10.1177/0008125619859318

Artificial intelligence (AI) has proven to be useful in many applications from automating cars to providing customer service responses. However, though many firms want to take advantage of AI to improve marketing, they lack a process by which to execute a Marketing AI project. This article discusses the use of AI to provide support for marketing decisions. Based on the established Cross-Industry Standard Process for Data Mining (CRISP-DM) framework, it creates a process for managers to use when executing a Marketing AI project and discusses issues that might arise. It explores how this framework was used to develop three cutting-edge Marketing AI applications.

+ G. Overgoor, M. Mazloom, R. Rietveld, M. Worring, and W. Van Dolen (2017), ”A Spatio-Temporal Category Representation for Brand Popularity Prediction”, in Proceedings of the ACM International Conference on Multimedia Retrieval, Bucharest, Romania

Social media has become an important tool in marketing for companies to communicate with their consumers. Firms post content and consumers express their appreciation for the brand by following them on social media and/or by liking the firm generated content. Understanding the consumers’ attitudes towards a particular brand on social media (i.e. liking) is important. In this paper, we focus on a method for brand popularity prediction and use it to analyze social media posts generated by various brands during a specific period of time. Existing instance-based popularity prediction methods focus on popularity of images, text, and individual posts. We propose a new category based popularity prediction method by incorporating the spatio-temporal dimension in the representation. In particular, we focus on brands as a specific category. We study the behavior of our method by performing four experiments on a collection of brand posts crawled from Instagram with 150,000 posts related to 430 active brands. Our experiments establish that 1) we are able to accurately predict the popularity of posts generated by brands, 2) we can use this post-level trained model to predict the popularity of a brand, 3) by constructing category representations we are improving the accuracy of brand popularity prediction, and 4) using our proposal we are able to select a set of images for each brand with high potential of becoming popular.

Working Papers:

+ G. Overgoor, W. Rand, W. Van Dolen, M. Mazloom, ”Simplicity is not Key: Understanding Marketer-Generated Social Media Images and Consumer Liking”.  Under second round review at International Journal of Research in Marketing

Social media channels are becoming increasingly important marketing channels, and recently these channels are becoming more and more dominated by content that is not textual, but visual in nature. Relating textual content to sales and conversions is difficult enough, but visual content is even more difficult to analyze. In this paper, we explore how consumers engage with visual content. Speci fically, we explore the role of the complexity of images in creating consumer liking. To carry this out, we use a number of different features of the images posted on Instagram by brands and relate these features to likes on the images. We use a convolutional neural network that can automatically identify objects contained in the images, to create features used to build the model. We show that there is a u-shaped relationship between the complexity of images and the amount of likes that they generate from consumers, and provide insights into how this knowledge can be used to generate and choose better social media images.

+ G. Overgoor, W. Rand, W. Van Dolen, ”The Champion of Images: Understanding the Role of Images in the Decision-Making Process of Online Hotel Bookings”. Under review at Marketing Science
 – Job Market Paper
 – Nominated for best paper at HICSS 2020

Images are vitally important in engaging consumers and helping them to make decisions. On many online travel agency (OTA) websites, the hotel`s image can take up 33 % of the space of the search result listing, but the importance of this image in the decision-making process has yet to be studied. In this research, we use deep learning to extract information directly from hotel images and we apply visual analytics to understand the importance of this information during consideration set formation. We perform a hotel-level prediction and find that we are able to accurately predict what hotel will be more likely to be clicked on based on the information we can extract from the image. We then complement these findings using LambdaMART to predict consumer clicks during search and find that on average there is a 10 % improvement when we incorporate image information as compared to just the textual and numerical features. In addition, we find that the imagery impacts the importance of other attributes such as price, with a decrease in importance of over 70 % in some locations. Finally, in a neuroscience experiment we show that our results can be explained by the fact that the human brain processes images with a high click-through rate differently than low click-through rate images. Overall, we present one of the first visual analytics frameworks that can be used at a large-scale to help understand the impact of imagery online. Our research has valuable theoretical and methodological implications that advance the study of unstructured data in marketing. 

+ H. Lee, G. Overgoor, W. Van Dolen, ”Who Has the Real Power? Identifying the Opinion Leader in an Online Brand Community”. In preparation for submission to Journal of Marketing.

What makes an influencer? Analyzing comments from a brand community, we examine the effect of social norms, personal characteristics, social network positions, and writing style on how influential users and comments shape and change collective sentiments. The results suggest that as users are heavily influenced by the majority of others’ sentiments, with the sequential biases in place, the first person to make the comments can be influential in dictating the rest of the discussion. To shift the sentiments from positive to negative, traditional influencer identifiers such as having a high social status, and being active and well-embedded in the network are essential, as well as writing with an authentic, confident, positive but not overly emotional tone. Conversely, to mitigate the sentiments from negative to positive, writing very positive comments is the only characteristic that matters. To sustain the impact, there need to be users who share similar characteristics to follow the sentiment pattern to create a new majority, i.e., the social norm. The findings highlight the need to reevaluate how we identify opinion leaders.

+ G. Overgoor, R. Mestri, W. Rand, “In the Eye of the Reviewer: An Application of Unsupervised Clustering to User Generated Imagery in Online Reviews” – accepted to HICSS 2021, expanding for submission to Journal of Marketing Research

Mining opinions from online reviews has been shown to be extremely valuable in the past decades. There has been a surge of research focused on understanding consumer brand perceptions from the textual content of online reviews using text mining methods. With the increase in smartphone usage and ease of posting images, these reviews now often contain visual content. We propose an unsupervised cluster method to understand the user-generated imagery (UGI) of online reviews in the travel industry. Using the deep embedded clustering model we group together similar UGI and examine the average review ratings of these clusters to identify imagery associated with positive and negative reviews. After training the method on the entire dataset, we map out individual hotels and their corresponding UGI to show how hotel managers can use the method to understand their performance in particular areas of customer service based on UGI. The performance in a  cluster relative to the population can be a clear indicator of areas that need improvement or areas that should be highlighted in the hotel’s marketing efforts. Overall, we present a useful application using visual analytics for mining consumer opinions and perceptions directly from image data.  

Research in Progress:

+ G. Overgoor, “Deep Choice: A Deep Learning Approach to Consumer Choice”

+ R. Mestri, G. Overgoor, W. Rand, “Generative Embedded Clustering: Developing a Continuous Latent Space for Generative Capabilities”

+ G. Overgoor, C. Chan, Y. Bart, K. Pauwels, “What is the price of a social media influencer?”

+ G. Overgoor, A. Colicev, Y. Bart, K. Pauwels, “The Impact of TV Ads on Consumer Mindset Metrics: An Application of Video Analytics”