Research and data lessons from a non-academic job:

One year on.

Five lessons from a non-academic job

It’s been one year after leaving academia and starting to work for a charity organisation. For a bit of context, I did my PhD in Psychology, after which I spent four years across two postdocs (one in France, one in England) and I got to the point where I could no longer see my career progress in academia. There are many reasons why I came to this crossroad, but above all, I realised that my passion is with research and data, and less so with the other aspects of academic life. To my delight, my non-academic journey so far has been full of development and challenges and here are some of the lessons I shared, with some additional thoughts. Here are five things I learnt:

The three data lessons

  1. I learnt a lot about data. Real life data is messy AF. I learnt how to join complicated datasets with missing data and how to deal with types of variables I rarely encountered like strings and dates. I cannot underestimate how neat data from academic research is compared to what real life data looks like.
  2. I learnt how organisations store their data. Unlike any other datasets I worked with before, organisational data is dynamic. There isn’t just one database but rather multiple sets that are related to one another. You need tools like SQL to pull data from internal data systems. SQL queries are also used to power live dashboards which are used to display data in the simplest form. This is where most teams take their data from. Tools like PowerBi and Tableau are the major players here when it comes to data visualisation and internal data sharing. This was a completely new lesson to me because like most academics, I have been used to dealing with only static data, pulled from Qualtrics or arranged neatly within one Excel sheet. The good news about SQL is that I was able to pick it up relatively easily with my programming skills in R but what I still find tricky is finding the piece of data I need in columns and columns of data that aren’t always neatly labelled. I really miss having a neat codebook with all variable names listed!
  3. I also had to revisit my thoughts on descriptive statistics. In academia, inferential stats are king because we infer something about a population from a sample. With organisational data, you have access to the whole population so descriptive stats become very powerful. I had to resist an urge to ask “yes, but is this increase statistically significant?” because if all 100 programme members answered the same question about the satisfaction in the following year, and we saw an increase of 5% in satisfaction, this in itself is informational and quantified in terms of the effect size - we don’t need inferential statistics to tell us if this difference was significant at the 0.01 level or at the 0.001 level and whether this would apply to the wider population - this isn’t the point.

The two (new) methodological approaches lessons.

  1. I learnt about quasi-experimental approaches using matching. While in academia I often ran experiments, establishing causation to evidence the impact of established programmes and interventions is more tricky as you often can’t randomly assign participants to treatment and control. This led me to learning about quasi-experimental statistical approaches and methods like propensity score matching to estimate the likelihood of treatment and select the ideal matched pairs. Lack of random assignment is also the least of all issues. The environment in which programmes and interventions happen in the real world is far from the controlled experiments we run in labs/online. So how do you assess whether participating in a long-term teacher training programme has a real life impact for the pupils they teach? How do you establish causation and the process by which it happens? These are a lot more tricky questions than I ever thought I had to approach. It turns out there is a whole host of methodologies to address this issue - these are a lot more prevalent approaches in disciplines such as policy (think: policymakers often want to know what impact introducing policy X had on the local population but they aren’t always able to randomly assign two groups of population into an intervention with policy application versus not).
  2. I learnt how to run a discrete choice experiment. This is a method that is often employed in marketing, I’ve rarely heard of it before but I found an excellent application of this method in the work we were doing. By designing a discrete choice experiment, we were able to assess what changes we could make to programmes we offer to increase their propensity. Would this programme be more popular if we introduced element X or increased Y? This technique looks at implicit preferences for various factors (not just two factors but many at the same time). It puts them directly in competition with one another to calculate propensity of each factor. It’s also often used in medical sciences and economics.I had a lot of fun designing this research and I am currently working with internal stakeholders to develop actions from this research - how can we turn this knowledge to attract more people to our programmes? It’s fascinating how quickly you can turn research into action.

I find the process of reflecting on these lessons somewhat therapeutic because this time last year, I was full of excitement but also filled with some worries as I was embarking on my non-academic path. These lessons tell me that I not only was able to build on my research and data skills that I developed during my academic career, but also this move has allowed me to expand my development in ways I didn’t anticipate before.

I originally went into academia because I loved this process of learning and developing, stumbling across new challenges and new approaches, but after 7 years I started getting worn out by discussing the same issues while using the same approaches and I craved new challenges. This first year of the non-academic pathway has certainly fulfilled that goal. I’m not saying that academia can’t be a fulfilling path with lots of challenges - I think it’s more about the fit. Digging deeper into what it is really about your job that makes you happy is going to lead you down the right pathways, academic or not.

Contact information: Karolina Urbanska (, Twitter @karo_urb, & website, and link to the original thread:

Editor’s note: The present text is an adapted and expanded version of the widely shared Twitter thread by Karolina which resonated with so many of us. We invited the author if she would be willing to adapt & expand her thread to an Open Scholarship audience.