Why You’ll Quit Your Data Science Job

Original Source Here

Expectation 1: I get to use cutting-edge machine learning algorithms to solve complex problems that will impact the business.

Sadly this isn’t the case with most data science jobs because you have to balance business needs with time. It’s difficult to justify to senior management the time you’ll spend chasing marginal improvements in accuracy with shiny new algorithms when a simple one will yield similar results.

As a data scientist, I supported email marketing and spent the majority of my time building purchase propensity models. I didn’t have opportunities to try other types of models because there was no business need. My work had a low impact because email marketing made up a small portion of overall revenue. I wanted to leave so badly I almost quit without finding another job first. Learn from my mistakes.

How to get closer to your expectation:

  • Read the job description carefully. — Is the main responsibility building machine learning models? Beware of job titles because data analyst jobs can have data scientist titles.
  • Clarify expectations of the role during your interview. — If the job description includes creating dashboards or ETL pipelines ask what percentage of your time will be allocated to these responsibilities. If modeling is only expected to take a small portion of your time then that’s not the right job for you.
  • Ask what groups in the company you’ll support. — Not all groups are created equal. Revenue generating groups such as sales or product get more visibility if your contributions can impact their KPIs. Ask questions about the business model and how the company makes money. What groups help grow the business? If your role supports those groups then your models can potentially have a big impact on revenue. This doesn’t mean you should decline an offer if the role supports a cost center but you should take it into consideration.

Expectation 2: I get to build models all day.

In order to have the luxury of building models all day you need the right data infrastructure and technology in place to support this. In all the companies I’ve worked for, I had to deal with data that was either decentralized, dirty, undocumented or a combination of the three.

How to get closer to your expectation:

  • Ask about the technology stack.— If the interviewer mentions Excel spreadsheets and an Access database run for the hills. ( Unfortunately, I was a consultant at the time and this was a client so I had to stay put. ) If the interviewer mentions recent software and tools you’re familiar with that’s a good sign.
  • Ask about the data availability.— Confirm the company has a database with data. Some companies are just starting out and there may be very little data to work with for modeling. Are there data engineers to help you load new data or are you expected to do the ETL development yourself? If there’s little data or you spend your days loading data, you’ll have no time to build models and that’s not the job you want.
  • Ask how the data team is organized. — Check if you’re the first data hire because you’ll likely work on any data-related task and modeling will be the last thing you get to do. Ask how many other data scientists are on the team. Check if there’s a separate team of data analysts. The best scenario is to work for a company with separate data science and data analytics teams. Having two teams means there’s a higher chance you’ll be building models while the data analysts create dashboards and run data analysis.

Expectation 3: The company understands the value of data science because why else would they have hired me.

A Deloitte report states “data analytics is growing in importance and plays a critical role as a decision-making resource for executives”. However, data literacy is a common problem in organizations. Companies want to be “data-driven” like everyone else but most don’t understand how to leverage data science to add value to the business.

How to get closer to your expectation:

  • Ask for examples of past problems and what models were used to solve them. — This question will help you gauge the data science maturity in the company. If you can’t get a good answer, this is a red flag. Chances are they don’t have a good idea of how data science can contribute and you’ll get frustrated due to the lack of data literacy. Beware of working in companies where data science is still unproven because you’ll encounter more resistance in adoption.
  • Learn how to present machine learning to non-technical people. — You’ll likely have to present your model results to many types of audiences. Many of them won’t understand what machine learning is. Learn to explain how your model can benefit the business. Having this skill will benefit you in any data science role and help your stakeholders appreciate the value you’re adding.


Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot

%d bloggers like this: