Solving Problems Like a Data Scientist (Part 3)
Solving Problems Like a Data Scientist (Part 3)
Asking Better Questions
“The important thing is not to stop questioning. Curiosity has its own reason for existing.” — Albert Einstein
Introduction: The Difference Between a Good Analyst and a Great One
Imagine walking into a doctor’s office with a persistent headache.
Before prescribing medication, a good doctor asks questions.
When did it start?
How severe is the pain?
Where exactly does it hurt?
What were you doing before it began?
Do you have any other symptoms?
Have you experienced this before?
Now imagine if the doctor skipped every one of those questions and simply prescribed painkillers.
You would probably find another doctor.
Yet this is exactly how many analytical projects begin.
A stakeholder requests a dashboard.
An executive asks for a report.
A manager wants a prediction model.
The analyst immediately opens Excel, SQL, Python, or Power BI.
Very few people pause to ask the questions that determine whether the work will solve the actual problem.
Exceptional data scientists understand a simple truth:
The quality of your analysis can never exceed the quality of your questions.
Questions shape investigations.
Investigations shape insights.
Insights shape decisions.
And better decisions create better outcomes.
Why Questions Matter More Than Answers
People often associate expertise with having answers.
In reality, expertise often begins with asking better questions.
Think about some of history’s greatest scientific breakthroughs.
Isaac Newton didn’t begin with the answer to gravity.
He began with a question.
Why do objects always fall toward the Earth?
Marie Curie asked questions about radioactivity that few others were willing to pursue.
Alan Turing asked whether machines could think.
Tim Berners-Lee asked how researchers could better share information.
Every significant discovery begins with curiosity.
Data science is no different.
The purpose of analysis is not simply to answer questions.
It is to answer the right questions.
Questions Determine the Direction of Your Analysis
Consider this simple request from a retail executive:
“Why are sales declining?”
At first glance, this appears to be a perfectly reasonable question.
But it’s far too broad.
A data scientist immediately begins narrowing the focus.
Instead of accepting the question at face value, they ask:
- Is the decline happening everywhere or only in certain regions?
- Which products are affected?
- When did the decline begin?
- Did anything change before the decline started?
- Are we losing customers or selling fewer products to existing customers?
- What external factors may have influenced demand?
- What decision are we trying to support?
Notice what happened.
The original question didn’t disappear.
It evolved.
The better the questions became, the clearer the analytical path became.
Curious People Solve Better Problems
One characteristic appears repeatedly among exceptional data scientists:
Curiosity.
Curiosity is more than wanting to know facts.
It is the willingness to challenge assumptions.
To investigate uncertainty.
To remain comfortable saying,
“I don’t know yet.”
Curious professionals resist jumping to conclusions.
Instead, they become investigators.
They recognize that every dataset tells only part of the story.
Their job is to uncover the rest.
Four Types of Questions Every Data Scientist Should Ask
Not all questions serve the same purpose.
Experienced data scientists intentionally move through different categories of questions as their understanding deepens.
1. Descriptive Questions
These questions help us understand what has already happened.
Examples:
- What happened?
- How many customers left?
- Which region experienced the greatest growth?
- What trends exist over time?
These questions establish the facts.
Without them, there is no foundation for deeper analysis.
2. Diagnostic Questions
Once we understand what happened, we investigate why.
Examples:
- Why did customer churn increase?
- Why are sales declining?
- Why are delivery times increasing?
- Why are certain branches outperforming others?
Diagnostic questions reveal relationships and possible causes.
This is where analytical thinking becomes more valuable than reporting.
3. Predictive Questions
Once patterns become clearer, attention shifts toward the future.
Examples:
- Which customers are most likely to churn?
- What will demand look like next quarter?
- Which patients are at greater risk of readmission?
- Which products will likely become best sellers?
Prediction helps organizations prepare before events occur.
4. Prescriptive Questions
This is where data science creates its greatest value.
Examples:
- Which customers should receive retention offers?
- Which marketing strategy should we invest in?
- Which stores should receive additional inventory?
- Which intervention is most likely to improve patient outcomes?
Prescriptive questions move beyond understanding.
They guide action.
Get Mills Clifford’s stories in your inbox
Join Medium for free to get updates from this writer.
Ultimately, this is why organizations invest in analytics.
The Question Ladder
One practical way to improve your thinking is to climb what I call the Question Ladder.
Each level brings greater clarity.
Level 1
What happened?
Example:
Sales declined.
Level 2
Why did it happen?
Customer complaints increased.
Level 3
What else might explain it?
A competitor launched a lower-priced product.
Consumer spending also declined.
Seasonality may have influenced purchasing behavior.
Level 4
What should we do?
Adjust pricing?
Improve customer experience?
Increase marketing?
Launch new products?
Level 5
How will we know if our decision worked?
Define measurable success.
This final question is surprisingly overlooked.
Without it, organizations struggle to evaluate whether their decisions actually solved the problem.
A Business Case Study: Asking Better Questions in Telecommunications
Let’s return to the telecommunications company from the previous articles.
Initially, executives believed the problem was straightforward:
“Customers are leaving.”
An inexperienced analyst accepted that statement and began building a churn prediction model.
An experienced data scientist asked additional questions.
Which customers are leaving?
When do they leave?
Which products are they using?
Do they contact customer support before cancelling?
Are cancellations concentrated in specific geographic regions?
Did network outages increase?
Have pricing plans recently changed?
Does churn differ between prepaid and postpaid customers?
One particularly revealing question emerged:
“What happens during the customer’s first 30 days?”
That single question changed the entire investigation.
The team discovered that customers experiencing poor network performance within their first month were significantly more likely to cancel their subscriptions.
The solution wasn’t simply a better churn model.
It was improving network reliability for new customers.
The breakthrough came from asking a better question.
Questions That Data Scientists Should Ask Stakeholders
Whenever a new project begins, consider asking questions like these.
About the Business
- What problem prompted this request?
- Why is this important now?
- How will this affect organizational goals?
About Success
- What would success look like?
- How will success be measured?
- What decision will change because of this analysis?
About Data
- What data already exists?
- What data is missing?
- How reliable is the available information?
About Constraints
- What timeline are we working within?
- What budget or operational limitations exist?
- What assumptions should we challenge?
These conversations often prevent weeks of unnecessary work.
Common Mistakes
Many analytical projects lose momentum because people ask:
- Questions that are too broad.
- Questions that assume conclusions.
- Questions based on opinions instead of evidence.
- Questions focused on reporting rather than decisions.
- Questions that ignore stakeholder objectives.
Another common mistake is believing the first question is the right question.
Often, the first question is simply the beginning of the conversation.
Key Takeaways
Asking better questions is one of the most valuable skills a data scientist can develop.
Remember:
- Curiosity is a professional advantage.
- Every answer should generate a better question.
- Good questions reduce wasted analysis.
- Better questions produce better decisions.
- Questions create the roadmap for every successful analytical project.
The next time someone asks you for a dashboard, a report, or a predictive model, resist the temptation to immediately begin working.
Instead, begin asking questions.
You may discover that the original request wasn’t the real problem at all.
Reflection Questions
Think about your most recent analytical project.
- What was the very first question you were asked?
- Did you challenge that question?
- What additional questions would you ask if you could start again?
- Did your questions lead to action—or simply more information?
Take a moment to reflect.
Sometimes the fastest way to solve a problem is to slow down and ask one more thoughtful question.
Closing Thoughts
Technology continues to evolve.
Data grows exponentially.
Artificial intelligence becomes increasingly capable.
Yet one skill remains timeless.
The ability to ask thoughtful, purposeful, and insightful questions.
The future belongs not only to those who can build sophisticated models but also to those who know which questions are worth answering.
Because in data science—as in life—
Better questions almost always lead to better solutions.
Join the Conversation
Think about a project, at work or in your personal life, where asking one additional question completely changed your understanding of the problem.
- What was the question?
- How did it reshape your thinking?
- What lesson did you take away from the experience?
I’d love to read your stories and learn from your experiences.
In Part 4, we’ll explore “Thinking in Hypotheses” and discover why the best data scientists don’t begin with conclusions—they begin with ideas that can be tested.
