The AI hype has been dominating tech headlines over the last couple of years. We've moved past the big data and machine learning fervor of the 2010s. So it might feel like the data science job market is slowing down.
That's far from the truth. Earlier this month, Forbes published an article titled "The 5 Fastest Growing Six-Figure Jobs In 2025". Data Scientist is featured in this list, referred to as "the hottest trend in business right now". According to the U.S. Bureau of Labor Statistics, a 36% growth in data science jobs is projected by 2033. This is certainly a positive signal for those considering data science as a career path.
Having spent a decade plus in this field, I can attest that data science can be a satisfying profession. But there are both positives, and the less positives. Here are three things I genuinely love about being a data scientist, as well as three challenges that anyone considering this career path should be aware of.
Highlight #1: Making Discoveries as a Data Detective
There's a unique thrill in uncovering insights from data. For a brief moment when you spot a pattern, you're the only person who holds this new knowledge. Whether it's identifying an emerging customer behavior, or validating a business trend, these "aha moments" are what make data science genuinely exciting. You are a data detective, gathering and piecing together evidence, and revealing to the world what was previously invisible.
Highlight #2: The People
As data scientists, our job is to extract knowledge from data. And as if through self selection, I often find data scientists to be highly committed to truth and rigor. When data scientists discuss findings, it's rarely just about the conclusions – you'll hear about methodology, limitations, and levels of certainty. This intellectual honesty, reminiscent of Hans Rosling's "Factfulness" approach, makes every interaction enriching.
This kind of exchanges gounded in truth and reasoning can be a breath of fresh air in a world rife with overconfidence and misinformation. For example, Professor Daniel Kahneman, a Nobel Laureate in Economics and expert in behavioral economics, has written about the danger of overconfidence. Kahneman warned, "people come up with coherent stories and confident predictions even when they know little or nothing. Overconfidence arises because people are often blind to their own blindness."
Of course, data scientists are not immune to mistakes and biases. But the culture of thoughtful analysis and evidence-based discussion creates an environment where statements are not made without evidence (and confidence intervals!).
Highlight #3: Tools
The breadth of data science tools is quite remarkable. From the elegant symbolic mathematics of Mathematica, to MATLAB's powerful numerical computing, to R's statistical prowess - each of these softwares opens new doors. Then there is Python's data ecosystem: my first encounter with Pandas DataFrame was a revelation - suddenly, data manipulation became intuitive and expressive. Seaborn made visualizations beautiful by default, scikit-learn makes machine learning accessible, and PyTorch enabled complex deep learning architectures. Natural language processing? SpaCy makes it approachable.
But what's even more exciting is how the landscape keeps evolving. Google Colab makes Python notebook collaborative, and brings powerful computing to your browser. Streamlit turned data scientists into web app developers overnight. Observable made data visualization not just interactive, but truly explorable.
The craft of data science is defined by people and tools. With such powerful tools at our disposal, we can focus on the pursuit of truth through data.
Lowlight #1: Getting "P-value Shopped"
A while ago, a few data science colleagues and I set up data science "office hours" where we'd help any team with data-related questions. One day, a marketing team member came to us with a statistics request. They had just run a quarter-long campaign and wanted to know if it was impactful. We performed the data analysis, and informed them that we couldn't detect statistically significant impact.
Later on, through the grapevine, we learned that this marketing campaign was eventually presented as a success. Apparently the marketing team was able to find another analyst who sliced and diced the data in a more favorable way. We felt deflated.
Such "p-value shopping" is more common than you might think. In an organization, everyone is under pressure to demonstrate impact. And data scientists are often the ones tasked with the proof. This can create a challenging dynamic where teams go from data scientist to data scientist, looking for the most favorable result. Meanwhile, the data scientist is put in the position of "justifying" a desired conclusion, instead of rigorously performing the analysis.
Lowlight #2: Not in Control to Affect Change
Speaking of impact, one of the most frustrating experiences as a data scientist is identifying a clear opportunity for improvement, but lacking the authority or resources to actually implement the change. You might uncover that, for example, a particular feature would significantly improve user experience, or a specific model could automate a manual task. You have the data. You can even estimate the potential impact. You make beautiful plots within a persuasive presentation. Everyone agrees that it's a good idea. But fast-forward several months, and no action is taken.
Data scientists often operate in an advisory capacity. While we can surface insights and make recommendations, the actual implementation typically depends on other teams – engineering for technical changes, operations for process improvements, or business teams for strategic shifts. Your carefully researched recommendations need to wait for prioritization among many competing initiatives. If you are action-oriented (and impatient!), this thumb twiddling can be grueling.
Lowlight #3: There isn't Always Discovery
While making discoveries is thrilling, the reality is that not every analysis leads to an "aha moment." Sometimes, after days of exploring data, testing hypotheses, and trying different approaches, we simply confirm what was already known. Other times, we might find that there isn't enough signal in the data to draw meaningful conclusions.
In academia, we have begun to appreciate the importance of null results. There is concrete value in documenting "this path has been explored; search elsewhere." But in an industry setting, as a data scientist, null results do not bring stakeholder excitement. Worse, if there are company goals built around making data discoveries, then instead of honest progress, the project would be seen as a failure.
Mitigating Challenges
I don't want to end this article with the challenges of being a data scientist, without offering solutions. In fact, there is an extremely effective strategy: shift how we think about data science work by moving from one-off analyses to building productized data solutions. For example:
- Instead of running analyses that might get "p-value shopped," build monitoring systems that continuously track metrics with clear statistical guidelines
- Rather than waiting for other teams to implement your recommendations, create self-service tools that empower stakeholders to directly access and act on data
- When faced with fishing expeditions, develop systematic approaches that can be reused across similar problems, turning exploration into scalable investigation frameworks
This shift from analysis to product isn't always easy – it requires additional technical skills and organizational buy-in. But it's often the most effective way to ensure your work as a data scientist creates lasting impact.
Are you curious about a data science career? Wondering how to stand out in a sea of job applicants? Check out our job search coaching service and see how we can help!
