Data Science Problem Finding: Practical Methods and Techniques
- Amara James Moosa
- Jan 21
- 3 min read
Updated: Feb 26
Introduction
How do you find meaningful data science problems before your boss assigns them? It's a question I've been asked countless times, and after seeing nearly a billion Google results on the topic, I knew it was time to share my approach. Like many data scientists, especially early in my career, I struggled with this too. This article reveals the proven methods I use to proactively identify impactful data science opportunities.
Data Science Problem Finding Practical Methods and Techniques
"Gemba," a Japanese term, translates to "the place where the truth can be found" or simply, "the actual place." In the world of manufacturing, a "Gemba Walk" involves observing work processes firsthand to identify inefficiencies. This concept, while rooted in manufacturing, can be incredibly valuable for data scientists.
Data Science Gemba Walks: Observing Users in Action
For a data scientist, a Gemba Walk isn't about physically visiting a factory floor. Instead, it's about observing how users interact with the product or service we aim to improve. This involves:
Becoming the user: Experience the product or service as a customer would.
For an e-commerce site, this means navigating the purchase funnel, adding items to the cart, and completing the checkout process.
Identifying pain points: Observe user frustrations, roadblocks, and areas of confusion.
Gathering user feedback: Actively listen to user feedback, both verbal and nonverbal.
Documenting observations: Record key insights and observations.
Benefits of Data Science Gemba Walks:
Uncover hidden problems: By directly observing user behavior, you can uncover issues that may not be readily apparent through surveys or data analysis alone.
Generate actionable hypotheses: Observations can lead to specific, data-driven hypotheses about user behavior and areas for improvement.
Improve data collection: Understanding user behavior helps you design more effective data collection methods.
Build empathy with users: By experiencing the product or service firsthand, you gain a deeper understanding of user needs and perspectives.
Real World Example
In a previous role as an Analytics Manager, I partnered with product managers to improve the Cart & Checkout experience for a major retailer. We focused on improving conversion rates through A/B testing.
One observation I made was that some customers were attempting to add new credit cards during checkout, but then abandoning the process. To understand why, I conducted a "Gemba Walk":
I experienced the checkout flow myself: I added a new credit card and completed a purchase. Then, I tried adding the same card again.
The unexpected outcome: The card wasn't added, and the transaction failed without any error message displayed to me.
To identify the root cause of declining card additions, I analyzed user sessions with Quantum Metrics. This revealed the system was silently rejecting duplicate cards due to fraud prevention measures. This unexpected behavior negatively impacted user experience. I collaborated with the product manager and fraud team, proposing a solution to overwrite existing cards with new ones.
To validate the solution's impact, I conducted a sizing exercise to estimate potential gains in conversion and GMV. Following approval, an A/B test was implemented. Results demonstrated a significant improvement in customer experience and business outcomes with the new design.
Who Uses It And When?
By conducting Gemba Walks, these roles can gain valuable insights into user behavior, identify critical areas for improvement, and ultimately drive better business outcomes.
Product Managers: Understand user pain points, identify areas for improvement, and prioritize features.
UX/UI Designers: Observe user interactions, identify usability issues, and improve the user experience.
Web Analysts: Analyze user behavior, identify website bottlenecks, and improve conversion rates.
Marketing Analysts: Understand customer journeys, identify marketing campaign effectiveness, and optimize customer acquisition strategies.
Customer Success Managers: Improve customer onboarding, identify areas for improvement in customer support, and increase customer satisfaction.
Data Scientists: Gain context for data analysis, identify relevant data sources, and generate more impactful hypotheses.
Important Notes
By being mindful of these limitations, data scientists can effectively use Gemba Walks to gain valuable insights into user behavior.
Confirmation Bias: Focusing on observations that confirm existing beliefs.
Observer Effect: User behavior changes when observed.
Limited Scope: Observing only a small subset of users.
Over-reliance on Anecdotes: Drawing broad conclusions from limited observations.
Lack of Structure: Conducting Gemba Walks without a clear plan.
Ignoring Context: Failing to consider the broader environment.
Neglecting Follow-up: Not acting on the insights gained.
Conclusion
In the dynamic world of data science, proactively identifying impactful problems is crucial for career growth and business success. By embracing the "Gemba Walk" principle, data scientists can move beyond data analysis and directly observe user behavior. This firsthand experience fosters empathy, uncovers hidden problems, and generates actionable hypotheses. By combining user observation with rigorous data analysis, data scientists can drive significant improvements in product, customer experience, and overall business performance.
Data Analytics Training Resources
Analysts Builder
Master key analytics tools. Analysts Build provides in-depth training in SQL, Python, and Tableau, along with resources for career advancement. Use code ABNEW20OFF for 20% off. Details: https://www.analystbuilder.com/?via=amara
This was a very insightful post on the importance of effective problem finding in data science. The tips on framing problems and understanding business needs were particularly helpful. Thanks for sharing!