Using social network analysis to understand consumer behaviour

What are social networks and how have you used them in your research?

With a grant from the Warwick Spotlights and working with colleagues at the Warwick Medical School, we are analysing large and anonymised real-time data and using this to construct social networks in the domains of health and business. For example, we have developed a health questionnaire with a team at Warwickshire NHS to gather data from pupils (over 13) to try and find out how mental health is spread from peer groups. In my own PhD research, I am working more in the domain of banking and finance, trying to understand how peer groups can affect the diffusion of an innovation. For this project we have collaborated with a European Bank to understand how consumers interact with a new technology or a new programme that has been introduced by a business. We then explore how peer groups affect this diffusion process, for example, by calculating the probability that a consumer will opt to use a new innovation because their peers and those in their social network have opted to use it. There's lots of literature on social networks and peer groups, but the intersection with mental health is something that has only been discussed in theory. In relation to business, we want to test the methodology in relation to very different types of data sets than it has been used with previously. One of my research projects involves large data from the London Stock Exchange Group (LSEG) database for testing a psychological bias using IBES and CRSP estimates. In addition, I am also working on empirically testing the theoretical utility functions for financial misconduct episodes arising from moral dilemmas that resonate with either emotional (System 1) or non-emotional (System 2) thinking processes.

What challenges have you experienced in using this methodology?

One challenge has been in understanding the data itself because they are both very different and the expectations are very different. In the worst-case scenario, for example, we might not be able to use the data to construct social networks and there will have been no point to the research. Also, different types of data present different challenges relating to time. With the data relating to mental health, there is no time challenge but, with the data from business, people are gradually adopting the innovation, and we are looking at whether this can be speeded up and, if so, how this might be done.

Some data can also be research-friendly while some is not. The data the NHS provided tends to be really precise and is usually clean. You do not have to do lots of screening before using it. You can start your analysis with a few minor tweaks. When it comes to other institutions like Banks, there are usually multiple files of data and multiple identification numbers for different products that are not stable across customers. So, for my PhD project involving banking data I have to construct a consumer identification variable that stays stable across the different data sets across time and when I merge them together. It can be very challenging.

Working with this scale of data and knowing what to do with it can feel overwhelming at times. I think because I have experience of working in business, I felt more comfortable with that data. But when it comes to asking children about their mental health, it feels like we are making a difference in people's lives: engaging with them, and asking them how they feel, rather than just taking their data and then publishing your results. For me, both are very different and both of them are important.

What advice would you give to researchers who are new to these types of methodologies?

I would say one size does not fit all. One methodology cannot fit all kinds of data and, similarly, one data can be looked at in thousands of different ways. With big data, for instance, I can use multiple models to interpret it, or I can even run a randomised control trial to test whether a theory is correct. Be as open as you can when it comes to data analysis. I had little experience with qualitative methodologies, but I learnt and found it to be interesting. I am still more comfortable doing quantitative research, but I can only say this now because I have tried a whole range of different methods. Have a taste of every single methodology that is out there. Even now I am learning to use large language models which is interesting but also frustrating because sometimes the model does not learn from the initial data sets provided.

I would also say be brave and be prepared to let go of some of the things you have worked on. Results, for example, can be so frustrating. Sometimes you work hard on writing code, but the code does not run, or the experiment does not produce the results that you had thought. So, you have to discard lots of things. That is a challenge for me right now and I have had to let go of things I have worked on for some time. It's frustrating. It's disheartening, but you have to learn to get up and keep going.

I would also suggest learning basic programming then building on it. Even if you are not going to be working with quantitative data, you need to have some basic proficiency in how different programmes work. For me, R Studio was the one, but it could be any. Just pick up one up and try to work on that: find some literature to support you and try to get better in it. Mastering one programming language will help you to keep up with the dynamic nature of research methodologies.