Why Data Scientists and Engineers Quit Their Jobs

It’s not just that they are underpaid

A recent study of data scientists made the following quite stark conclusion: “The typical data scientist works for a large tech firm — where they have been employed for roughly one year with an average of 6.2 years of prior experience in the field. Notably, they have switched companies twice (or more) since 2017 and will likely remain with their current employer for a mere 1.7 years on average. Only a tiny minority of those surveyed (2%) had not changed jobs within the last five years.”

Turnover is a big problem in the data science and data engineering professions, and it hurts everyone. Data scientists and engineers themselves do not want to be jumping frequently from position to position, as that does not help them build long term skills and expertise and looks bad on their CVs. Equally, for employers, replacing these skills is hard and expensive. Finding the specific technical background is not easy, and the ramp up time of new joiners is a further opportunity cost. Often, the desire to quickly replace someone leads to a vicious cycle of poorly qualified hires which exacerbates the turnover problem.

So let’s lay out the main reasons why data scientists and data engineers leave their jobs. Broadly, these reasons fall into three categories: economic, technical, and environmental. Employers need to address all three to stand a chance of building long term retention in their data science staff.

Economic reasons

The economic reasons are fairly clear and well-known, I think. Many, many employers have no idea how to price the skills of data scientists and engineers, and in the absence of good intelligence, they price too low.

In an example on LinkedIn, Moez Ali called out a job posting requesting someone with 6+ years of work experience, but describing it as “entry level” and offering a salary of around USD 65,000. There are many, many examples of these kinds of job postings, including naive job postings that ask for for a 5+ years of experience in a software package that has existed for less than 3 years.

Making job postings like this give away that employers have no idea about the roles and skills they are hiring for, and are a good indication to any data scientist/engineer worth their salt that they will be joining an environment that does not appropriately value their skills. As well…