In short, the answer is…it depends.

A PhD is a great way to get deep exposure to a number of core data science domains. That said, because the field of data science is moving so fast, experience is often more important than a degree – and the actual research of a PhD might be outdated in a relatively short amount of time (although the experience of a PhD will certainly be useful regardless, especially if you want to get into research at some point!)  

At Peak, you definitely don’t need a PhD. We look for skills, values, and potential for data science when we are recruiting, rather than purely at your academic achievements. It is hard to know what the field will look like in five years, so we want adaptable people who are enthusiastic about change. 

We know that good data scientists can come from anywhere, and this is reflected in the diversity of our data science team’s academic backgrounds. In our team, just over half have a PhD. Just under half have a Master’s degree, and 5% of the team have neither.

The academic backgrounds of the Peak team are also quite varied. In fact, 82% of the team didn’t study data science specifically. Any subject that teaches coding, maths, and technical research can act as a great spring-board for a career in data science. Maths, the Physical Sciences, and the Computer Sciences are a few academic domains with a particular focus on these skills – and the bulk of our team are from these sorts of backgrounds. That said, the team at Peak covers quite a broad range of disciplines – everything from Astronomy to Neuroscience!

As detailed in another recent blog post on landing a first job in data science, at Peak we want data scientists who are good at problem solving, coding, and maths. Doing a PhD in a relevant field is a good way to learn some of these skills, but there are other avenues (such as a Master’s in data science or a graduate scheme) available too. 

Why did you choose a PhD?

There was a large range of responses to this question. It is fairly common for a PhD to be a next step for people who are not sure what they want to do, and enjoy studying. To some people, it feels like a natural extension – they love their area of study and a PhD allowed them to really follow their curiosity! 

Some members of the team had planned for a PhD since they knew what a PhD was, and targeted specific research groups known to be the best in the field. Others saw a PhD as a good way to change the direction of their career.

What was the best thing about doing a PhD?

PhDs are a great way of learning a lot of skills independently. You will learn project management, teamwork, coding, statistics, and you should become an expert in your particular field. On top of the extra skills, a PhD itself can make your CV look a lot better, and can make you stand out from the crowd as you will get to add extra letters before or after your name (something that proved to be a strong motivator for people in the team while doing their PhD, but also something they stopped using pretty quickly!). 

Some of the opportunities you will get during a PhD, such as public speaking, and solving really hard problems, can also improve your confidence.

On top of that, for many, day-to-day PhD life can be really fun. As well as continuous learning, PhDs often involve a good amount of travel, can sometimes have a lot of social interaction with people who have similar interests, and often grant the freedom to work the hours that you choose to, take holidays as you like, and work remotely as much as you want!

What was the worst thing about doing a PhD?

It might be surprising to learn that not every data scientist at Peak enjoyed their PhD, and some actually left mid-degree to pursue something else. In fact, there were four main challenges faced during a PhD that were consistently raised by the team. These were:

• Stress
• Loneliness
• A feeling of a lack of impact
• A change in career direction

Stress

PhDs can be hard, and stressful. When you are writing up a PhD, it can take a few months of working 12-hour days (including weekends.) This work will usually fall entirely on you, with nobody to help. Unlike a regular work environment, where there can be a lot of small things due regularly, a PhD will often have very large things due less regularly, which can make the deadlines much higher pressure when they do come along.

Loneliness

Often, you and your supervisor will be the only people in the department working on the specific problem of your PhD. If your supervisor leaves mid-degree, or your communication with them breaks down for some reason, you can feel very isolated. It can be very difficult to find support within the university system sometimes, so having a good personal support network can be key to whether you enjoy your PhD.

Impact

Unfortunately, many PhDs have little tangible impact, especially while the degree is being completed. Newton spoke evocatively of standing upon the shoulders of giants, but in many ways the biggest giant in modern science is a bunch of regular sized people on top of each others’ shoulders in an oversized trench coat! In particular, in niche fields, or fields with few practical applications, the years of hard work somebody puts into a PhD could result in a piece of work that is only read by a few people.

Career direction

Often, it takes until halfway through a PhD for people to decide that the academic career path is not for them. The PhD itself gives a better view of what academic life is like than what an undergraduate student experiences, which can be enough to help people change their mind. This is especially true in data science, where a lot of research and innovation is being driven within industry. In fact, we have several data scientists in the team who left their degrees early so they could get into industry faster! 

So, which option is right for you? We asked the team to share their thoughts…

General advice

If you do plan to complete a PhD in before moving into the data science industry, here is some advice from the team.

? Be prepared, and make sure you are motivated enough to finish. A PhD can get very difficult towards the end. Final years are often filled with 12-hour days, six or seven days a week.

? A PhD is not necessarily the best thing for a career afterwards, especially in data science. There are usually other paths available to the job you want which might be better, especially if the PhD project available to you isn’t a perfect match for the role you want. Don’t do a PhD just because it’s the next natural step.

? Make sure you get on with the supervisor, and have a passion for the subject. A lot of people who end up struggling with their PhDs do so because they don’t have a good working relationship with supervisors. Definitely try and meet them beforehand.

? Make sure you join a good research group – it makes travelling to conferences much more fun. This isn’t just ‘good’ in the academic sense either (though that might help.) If there is an opportunity to chat with the research group informally beforehand, it can be helpful to get a feel for whether you might enjoy working with them!

? Make sure you are comfortable with being a student for at least another three years. That means less disposable income than some of your peers, and potentially delaying other life events a bit longer too. Since a PhD is research-based, be prepared for the possibility of everything going wrong, having to change direction part way through, and the degree taking an extra two years on top of what you originally planned (even if funding has ran out)!

? Be wary of unfunded PhDs. Given all the work pressure, doing it without getting paid just adds another layer of stress that may be better avoided. If it’s something you are incredibly passionate about it may be worth it, but really think carefully before committing.

? If you are looking for a funded option but don’t exactly know what to do, the Centres for Doctoral Training (usually called CDTs) can be a great place to start. They offer fully funded MSc + PhDs, so you get a year to figure out what you want to do during your MSc while getting paid to do it. The CDT helps you be near potential supervisors, have talks with each of them, and know the departments of the university, before you make a decision. A CDT is particularly useful if you don’t know the university from the inside. The downside to a CDT is that they usually have set intakes, so you will need to delay the PhD by six to 12 months, but it may be worth it.

PhDs are often very flexible around work hours and remote working. This can be a double-edged sword. On the upside, it is much easier to set your own work schedule, and work at times and in a place that suits you. On the downside, holidays aren’t normally tracked at all, and some parts of academia have something of a culture of overwork (including taking too few holidays!) This often means research students don’t take enough time off either.

If you are choosing to be somewhere for a few years with very little money, make sure it is a beautiful place, with people you like, doing something you love.

Conclusion

Choosing to do a PhD in a relevant subject area can be an excellent choice for people who want to learn independently. Over the course of your PhD, you will likely teach yourself a lot of skills that will stand you in good stead for the commercial world, including how to code, how to research, how to read and write technical documents, and how to solve problems. A PhD is also a chance to travel, meet interesting people, and to ultimately complete a technically demanding research project.

On the other hand, PhDs can be very stressful, especially near the end. Self-learning can be challenging, and the whole degree can feel very isolating (especially if there are difficulties with your supervisor). PhDs generally have a very narrow technical focus, which can lead to a feeling that your work has no impact. Finally, it will delay you from starting your data science career, which might frustrate you if you aren’t completely sold on the PhD itself. 

Compared to a PhD, a Master’s degree is much quicker to complete, meaning you can typically get a job quicker. Master’s degrees expose you to a broad range of data science techniques, with learning in a much more structured environment, so you are less likely to ‘miss’ an important lesson or concept than you might be in a PhD. You will also get a good working knowledge of all areas of data science.

The problem with Master’s programs is that they often don’t have time to explore the application of techniques to real problems and real data, or to build highly complex systems. These are critical skills for any data scientist (but can be gained in other ways too, such as by building up a portfolio of data science projects.)

So, should you do a PhD if you want to be a data scientist? In short, the answer is…it depends. But, hopefully this blog has helped you decide which path best suits you!

Stuart is Head of Data Science at Peak, and Tom is a Data Science Team Leader. Got a question for them about landing a job in the industry? You can connect with Stuart on LinkedIn here, and find Tom here.

As a starter for ten, here’s some key data science terminology to be aware of if you’re looking to learn more about AI and its benefits!

Artificial intelligence
This seems like a good place to start! At Peak, we think of AI as a general term that refers to computer hardware or software that behaves in a way which appears intelligent. The term was first coined by Dartmouth Assistant Professor John McCarthy way back in 1956 when he said: “AI is the science and engineering of making intelligent machines, especially intelligent computer programs.”

Computational neuroscience
This term was first used at a conference by Eric L. Schwartz in 1985, who was providing a review of a field that had no previous name. Computational neuroscience explains how electrical and chemical signals are used in the brain to represent and process information.

Cybernetics
Cybernetics is the study of how information is communicated in machines and electronic devices – which is then compared to how information is communicated in the brain and nervous system.

Data science
Data science is at the heart of what we do here at Peak. It combines domain expertise with programming skills, as well as mathematical and statistical know-how, in order to extract valuable insights from data. We apply machine learning algorithms to a whole host of data types to allow our AI System to execute tasks that would ordinarily require human intelligence, with a focus on delivering profitable outcomes for businesses.

Neural networks
Neuroscientists have discovered that the brain is composed of billions of interconnected processing units called neurons. A neuron receives inputs from multiple sources, integrates that information, and sends an output to many other connected neurons: a neural network. Due to the highly interconnected nature of a neural network, it’s ideal for learning and recognising relationships between many related pieces of information. Because neurons are analogue processing units, they don’t rely on strict logical rules like a normal computer program, and can therefore adapt to new situations when they arise. Neural networks have found use in many fields such as computer vision and speech recognition, but they can also be used to learn and predict relationships between many other types of real-world information streams like market trends and patterns of consumer behaviour. Peak takes advantage of the power of neural networks to process and understand vast quantities of data to help businesses make better business decisions.

Deep learning
Neural networks benefit from having multiple layers. For example, if an image recognition neural network is shown a picture of a white fluffy cat, the first layer might recognise nothing more than the presence of fluffiness in the image. The next layer might recognise the eyes and ears, and as deeper layers are added, the model can begin to piece together the general patterns of fluffiness, eye position, ears and general cuteness required to recognise that this is a cat and not a dog. Deep neural networks therefore enable deeper insights in a given dataset and can sometimes outperform humans in decision making and planning.

Fuzzy logic
Fuzzy logic was a term coined by University of California’s Dr. Lotfi Zadeh back in the 1960s. It’s defined as an approach to computing that’s based on the idea of “degrees of truth”, as opposed to the usual black and white, “true or false” logic that modern computers are based on.

Machine learning
Also an increasingly-used term in the world of business, machine learning (often abbreviated to ML) is an application of AI. It allows systems to be able to learn automatically, improve from experience, and get smarter over time, without being explicitly programmed in a certain way.

Natural language processing
Often abbreviated to NLP, natural language processing is a branch of AI that helps computers to understand and interpret human language – both text and speech. So, when you ask Siri to skip that aforementioned earworm, that interaction is made possible because of NLP – the AI recognises your voice, understands the action that you’re requesting, executes the action, and responds appropriately within seconds.

If you’re interested in how NLP can be beneficial in a business setting, be sure to check out our case study with multinational pharmaceutical firm GSK.

Skynet:
Uh oh. How did this one sneak in here? The evil antagonists from the Terminator franchise probably don’t give off the best first impression when it comes to AI, but rest assured that Peak’s approach and our solutions are a lot less terrifying – honest.

If you’re interested to learn how you can apply some of the above data science terminology in your business to drive growth and deliver value, we’d love to hear from you.