Many large companies have format requirements for data science jobs, such as a Masters degree or even a PhD. These roles are competitive, so companies typically increase the requirements to filter out the less dedicated applicants.
Still, there is a shortage of highly skilled workers who can combine knowledge of programming and statistics in a business environment.
Whether it is worth it to get a PhD is another topic, widely discussed elsewhere. There is a different ROI for different jobs, and it is not universally valuable to get a PhD. An exception is focusing on CS or Machine Learning, where there is high demand at big tech companies for these skills.
Big vs Small Companies Hiring
Data Science jobs basically fall into two categories: large companies and early-stage startups.
At a large company, where there are more established processes to follow, the HR department will often restrict applicants who meet certain criteria, such as degrees and years of experience. They sort through hundreds to thousands of applicants for each role. Every factor that helps narrow the scope of qualified candidates means less applicants to review and less work for the hiring manager.
Startups, on the other hand, tend to be more flexible with requirements. The best way to gain experience is learning by doing, and often startups will allow people to start who don't have formal or specialized education, as long as they are willing to prove themselves through an internship. Many startups will take ambitious interns who can demonstrate their abilities within a few months, assuming they have the funds and capacity to manage them.
That being said, many data scientists do have PhDs. Whether it is an advantage or not, depends entirely on the relevance to the job and the skills learned in the process. I discuss this in my interview with Natalia Bielczyk, PhD.
Completing a PhD was a highly personal choice, which I have seen other engineers, including former head of Google Research Jakob Uszkoreit, dropped out his PhD, much to the chagrin of his supervisor.
Data Science vs Data Engineer
While Data Scientist was the sexiest role of the last decade, the data engineer role has been rising in popularity for fresh grads, due to the wider demand and the difficulty of finding and defining the data scientist skill set.
If your skills extend beyond a Jupyter Notebook to backend infrastructure, it is worth considering a role as a data engineer, which practically never requires a PhD, and is almost purely hands-on and skills based. Do you prefer the theoretical or statistical aspects of data analysis, and the ability to expand what is possible with a dataset? Data science is more likely your gig. Do you rather prefer coding and coordinating with a team to solve technical challenges and build maintainable software. Engineering is likely a faster path to success.
I personally choose to complete a PhD after already having landed a job as a data scientist. Why? Perhaps a combination of perfectionism and curiosity - I like that I can speak at eye level with other researchers who have dived deep into a topic.
The willingness to understand a scientific topic to an extent that only a handful of others can relate makes it however sometimes a lonely process. I happen to think it is a beautiful thing to be willing to sacrifice a material reward in order to get closer to an understanding of the world, but I am not easily motivated by extrinsic rewards. When I meet someone else who has similar motivations, there is often mutual respect for the commitment to knowledge, and that adds a lot of personal value to my life.
Teaching
The ability to teach down the road is a clear advantage for large companies that wish to keep open their recruitment pipeline. Teaching courses means access to a wider talent pool and is an advantage to keeping skills fresh with current approaches and technologies. It is not uncommon for students to be more skilled at the latest frameworks than their professors or even hiring supervisors in industry. WIthout a PhD, it may be more difficult to be visible as an instructor and to provide value to the university system.
Social Impact
With a PhD, you will be seen as having achieved the highest level of formal education in your field. To some, this means a wider audience as an author on scientific or engineering topics, not only academic but also lay audiences. This is a bonus to companies small and large who like to brag about the number of PhDs on their data science team. Not a few investors are also impressed to see people with so much training passing up opportunities at larger companies to work with a small company.
What is necessary for the job?
What is actually necessary, usually depends on the job, and usually boils down to a combination of engineering skills, analytical skills, and communication skills. These can be learned in a doctoral program, or, perhaps even better, in an industrial environment matching the position. Your 3-month data science boot camp may give you the confidence to proceed, just as much as your 3-year PhD. If the goal is a position at a startup, maybe skip the lengthy PhD and go strait to a boot camp, where you will learn the basic skills necessary, and can decide later if a PhD is right for you.
Conclusion: It Depends
In short, do it if you want to, and can afford the opportunity cost. It can be a liberating period of time to explore a topic you are passionate about, and gain skills in programming, statistics, and academic collaboration. But if you see it as a requirement on a job application, consider that it might simply be there to filter out candidates who have not committed so much to their technical education, for reasons independent of actual hands-on skills. You can ask yourself if it is worth it to belong to the club of PhDs for reasons independent of getting the job.