My Data Guest — An Interview with Francisco Villarroel
I am thrilled to have Francisco Villarroel, proficient KNIME user and innovative educator, renowned scholar and marketing analyst in this My Data Guest interview.
Francisco talked about his work and teaching experience with KNIME, necessary preparations for a data job, and future possibilities on automation and analytics. For this episode, Anthony Ballerino and Simone Di Gregorio, two of Francisco’s former students and former interns at KNIME, are our special interviewers.
Francisco Villarroel is an Assistant Professor of Marketing at Luiss Guido Carli University in Rome, and a prolific researcher in Marketing Analytics. His research is centered around unstructured data, such as online reviews and social media content, where he relies on advanced text and image mining techniques to extract marketing insight. Prof. Villaroel was recently awarded a Teaching Excellence Award from the Chancellor of his home institution, and holds a KNIME Contributor of the Month award for April 2022.
Simone: What is your background and area of scientific expertise?
Francisco: My Bachelors was in Business and Economics. I graduated in 2006 from the University of Chile. I worked for about 6 years, mainly on topics related to branding. This led me to start an MSc in Marketing in 2012 in Manchester (UK).
I had a course that was called Data Science in Marketing and that is how I fell in love with the idea of analytics, especially for unstructured data. In 2012, I started a Ph.D. at the University of Maastricht with a focus on the use of text data for marketing insight. Since then, I have been deepening my knowledge at the intersection of marketing and analytics. From 2016 until 2020, I was an Assistant Professor of Marketing at UMass Amherst (US) with a focus on teaching and research in the areas of Text Analytics for Marketing, Branding, Customer Experience and Sentiment Analysis.
Simone: You have traveled a lot during your career. You started in Chile, then moved to the U.S. and Europe. Do you see any differences in the teaching approach between the U.S. and Europe?
Francisco: They were somehow different. Chile was similar to Italy, with a more hierarchical type of teaching. There is more distance between professors and students. I would say that it is similar in the U.S. but that depends on the type of university and the size of the class. In the Netherlands during my Ph.D., we had a problem-based learning approach. I used to have classes of 10–15 students, where it was more of a conversation with the students to discuss how to solve a problem.
Anthony: When was the first time you encountered KNIME and do you use it in your research?
Francisco: It was in 2012 in the beginning of my Ph.D. I took a course called text mining, a specialized course on text analytics from the Computer Science department. One of my colleagues in the class recommended me using KNIME Analytics Platform for my projects. I used KNIME for two research projects and I enjoyed using it as it was very intuitive. At my Assistant Professor of Marketing position at the University of Massachusetts, Amherst, I started teaching with KNIME. KNIME was not frequently used in the classroom back then but the students liked it immediately. Nowadays, there are so many support materials like the courses, books and the documentation but back then I had to teach almost every process.
Anthony: How did you come up with the live repository for Machine Learning and Marketing solutions on the KNIME Community Hub that you develop together with the Evangelism team here at KNIME?
Francisco: I always wanted to have a paper that could be a sort of resource or repository of analytics resources for marketing researchers and students.. In the Journal of Business Research, there was a special issue about Machine Learning in Marketing and I thought about writing an article on how to do machine learning with KNIME. I immediately thought about Rosaria to ask for some help. The team at KNIME is super hands-on, helpful and motivated.
Note. The scientific paper “Machine learning for marketing on the KNIME Hub: The development of a live repository for marketing applications” is available here.
Simone: Do you plan to add more deep learning-driven solutions to the repository?
Francisco: Yes. We’re currently working on a project that uses deep learning for image classification, which will be in the repository at the end of this year or the beginning of next year. We plan to make an app that could help content managers decide which type of image to use and how to combine it with text.
Anthony: How can students or researchers access such repositories?
Francisco: They can just go to the KNIME Community Hub and search for tags like “Marketing Analytics”, “Sentiment Analysis” or author names like mine. It is an amazing resource for scholars.
Anthony: Which collaboration with other fellow scholars are you the most proud of?
Francisco: It is hard to pick one. I work on several interesting projects. I love the KNIME live repository project because it is about my experience with KNIME and has uses for marketing analytics. Another project of mine is about language theory and its connection to an NLP application. Combination of theory and analytics makes the projects very unique.
Anthony: What is the main reason that convinced you to adopt KNIME Analytics Platform for your courses?
Francisco: I think that combining resources and optimizing time is important. In 2016 when I started to teach, I asked myself: why would I teach analytics with SPSS or R when I use KNIME for my research? I did not have too many expectations in the beginning but I noticed that the students liked KNIME and I got positive feedback from them. It also helps people from different backgrounds, such as Computer Science or Marketing, work together with it.
Anthony: Do you include use cases from Machine Learning and Marketing live repository in the KNIME Community Hub in your classes?
Francisco: I started using the Customer Experience and Sentiment Analysis workflows last year. The first one measures customer experience from reviews and the second one measures sentiment analysis from Twitter. I plan to integrate the Attribution Models and Brand Reputation Tracker workflows this year.
Simone: How did you come up with the idea of organizing the final project for a Bachelors course of yours with KNIME?
Francisco: I have always liked applied learning and I like to innovate a little bit in my courses. I don’t like to do the same things for my courses every year. In the U.S. in 2015, I enrolled all my students for a ‘Digital Marketing Competition’. Two of my teams got to the final and one of them got second. The year after when I started to teach at Luiss Guido Carli University, I came up with a challenge that included an Italian analytics company. The second year at this university, I thought about including KNIME in this challenge since our research collaboration had been really good and I was familiar with the team, with Rosaria and Roberto.
Simone: Do you think that a collaboration between students and businesses can impact the overall outcome of a project?
Francisco: The best scenario is to involve outside businesses and that is why I wanted to involve KNIME. It exposes the students not just to my evaluation but also to the industry and other professors.
Anthony: What was the most rewarding part of your work at Luiss Guido Carli University?
Francisco: It is encouraging to get recognized and be awarded by the students. When you see engaged students in the classroom who end up obtaining internships in interesting places like KNIME, for me that is the most rewarding part of the job.
Simone: Where do you see marketing and analytics in the next few years and will businesses increasingly adopt data-driven solutions?
Francisco: I think it is going to be about Marketing Attribution, the use of analytics to understand which marketing channel leads to a conversion/purchase. Organizations have better data about consumer interaction and models to understand the links between channels, which is a combination of Econometrics and Statistics. Interestingly, there is still very little machine learning on these topics but there can definitely be more.
Simone: What is your opinion about AutoML, will everything be automated in the future?
Francisco: There is going to be a lot of automation in the future. I really like some reflections of two of the most prolific authors in marketing: Roland Rust and Ming-Hui Huang. They work in the field of artificial intelligence and the intersection with marketing. In their book ‘The Feeling Economy’, they describe how AI has its stages, its current focus on mechanical tasks, and future analytical tasks, such as the type of advertisement to send to a customer, language-based models or image selection. Analytics teams have to better understand the role of human touch, empathy, feelings and business relationships.
Roberto: What would students need to learn to prepare for the data job market? Perhaps something in the direction of further complementarity between the human and the machine?
Francisco: Yes, absolutely. This makes me think of a recent project that I am currently working on. It started as a research project but ended up with the development of an app, which can help content managers make better decisions about the content that they post. This relates to the fact that education needs a more interdisciplinary approach. Organizations will also have to customize the development of their apps for better quality products or services.
Anthony: We are reaching the end of our interview. Before we say goodbye, we’d like to ask a few more questions. Which KNIME nodes would you select as your top 3?
Francisco: The top node for me would be the GroupBy node. It is one of the simplest nodes but it is incredibly powerful. I like the String Manipulation node too as it allows me to handle string and text data. And I love the nodes of the KNIME Deep Learning — Keras integration ‒ they are very nice and intuitive to use!
Anthony: Are there any upcoming projects that you are working on?
Francisco: I am working on different projects related to influencer marketing, chatbots, customer service and content marketing as well. So a little bit of a mixture but the common pattern is them having unstructured data.
This interview was first published on Low Code for Data Science.