Customer attrition, often referred to as “churn,” is a critical issue for businesses across industries. Losing customers not only impacts revenue but can also indicate deeper problems within a company’s products, services, or customer experience. When a client approaches a data science team stating that they are facing severe customer attrition and are unclear about what to do, it is vital for the data scientist to proceed strategically. This article delves into the right approach to tackle such a scenario, covering both descriptive and predictive solutions. We will explain essential concepts, provide actionable steps, and suggest options for clients to help them make informed decisions. This phase of client interaction and problem scoping, although often underrated, is fundamental to delivering true value through data science.

Data Science Interview Question: Tackling Customer Attrition

Understanding the Problem Statement

Imagine a client emails you: “We are facing severe customer attrition. We aren’t clear on what could be done. We want your help. How will you proceed?” Many aspiring data scientists immediately think, “Let’s build a churn prediction machine learning model!” However, this instinct to leap straight to machine learning can be premature and may not align with the client’s true needs.

The right approach involves careful, structured interactions with the client to define the project scope, clarify their expectations, and suggest potential avenues for solution. This critical phase—often called business problem scoping—lays the foundation for a successful data science project. Let’s break down how to proceed, the concepts involved, and best practices for delivering real business impact.

The Importance of the Problem Scoping Phase

The problem scoping phase is often overlooked by technical professionals, who may favor jumping straight into analysis or modeling. However, this phase is essential for the following reasons:

• Ensuring Alignment: It aligns the data science team’s work with the client’s business goals.
• Avoiding Misguided Solutions: It prevents wasted effort on solutions that don’t address the root cause or business context.
• Building Trust: It demonstrates to the client that you care about their actual problem, not just showcasing technical skills.
• Defining Success: It helps define what success looks like for the project.

Key Questions to Ask the Client

• What do you mean by “customer attrition”? How are you measuring it?
• Which customer segments are most affected?
• What data do you currently collect about customers and their journeys?
• Have you identified any patterns or triggers for attrition?
• What actions do you want to take if you could identify at-risk customers?
• What is the desired business outcome: reduce churn rate by X%, improve retention, increase revenue, etc.?

Two Main Approaches: Descriptive and Predictive Solutions

After clarifying the above, you can present the client with two main approaches to address customer attrition: a descriptive solution and a predictive solution. Each has its own merits and should be considered based on the client’s needs, data maturity, and business context.

1. Descriptive Solution

A descriptive solution uses data analysis and visualization to understand and communicate what is happening and why. This methodology is especially useful when the patterns leading to customer attrition are not complex and can be identified with simple analysis.

Steps in a Descriptive Approach

1. Root Cause Analysis (RCA) Dashboard:
   • Aggregate and visualize key metrics that may influence attrition (e.g., usage frequency, customer complaints, service interactions).
   • Highlight trends and anomalies over time.
   • Identify patterns or events that precede attrition.
2. Data Analysis and Pattern Recognition:
   • Segment customers by relevant factors (demographics, product usage, tenure).
   • Compare behaviors of retained vs. lost customers.
   • Look for statistically significant differences using hypothesis testing.
3. Share Key Insights with Client:
   • Summarize findings in business-friendly language.
   • Recommend actionable steps (e.g., improve onboarding, offer targeted incentives, address common complaints).

When is Descriptive Analysis Sufficient?

If the reasons for attrition are straightforward (e.g., customers leave after a price hike or poor support experience), descriptive analytics alone can empower the business to take effective corrective actions. Not every problem requires machine learning—sometimes, “good enough” insights are all that’s needed to make a big impact.

2. Predictive Solution

If the descriptive phase shows that attrition patterns are complex, or if the client explicitly wants to anticipate which customers are likely to leave in the near future, a predictive solution may be warranted. This involves building a machine learning model to predict customer churn, coupled with customer segmentation for targeted interventions.

Steps in a Predictive Approach

1. Data Inspection:
   • Assess data availability, quality, and completeness.
   • Check if data meets minimum requirements for modeling (sufficient size, relevant features, labeled outcomes).
2. Feature Engineering:
   • Create features that may influence churn (e.g., last login date, purchase history, support tickets, engagement scores).
   • Consider temporal trends and customer lifecycle stages.
3. Model Development:
   • Choose appropriate algorithms (e.g., logistic regression, random forest, XGBoost, neural networks).
   • Train, validate, and test the model using appropriate metrics (accuracy, precision, recall, ROC-AUC).
4. Customer Segmentation:
   • Score customers by their risk of attrition (e.g., probability output of model).
   • Segment into buckets: “Very Risky,” “Medium Risky,” “Low Risky.”
   • Enable business to prioritize retention efforts efficiently.
5. Deployment and Real-time Tracking:
   • Integrate model into business systems for real-time or periodic scoring.
   • Build dashboards to monitor churn risk and intervention results.
6. Discuss Frameworks, Infrastructure, and Costs:
   • Evaluate required tools (e.g., cloud platforms, ML frameworks, BI tools).
   • Estimate infrastructure and maintenance costs.
   • Ensure your solution aligns with client’s technology environment and budget.

When is Predictive Modeling Appropriate?

If the client needs to proactively contact customers at risk of leaving—especially in large-scale, competitive industries—predictive modeling provides the granular insights necessary for targeted action. However, predictive solutions require good data, technical expertise, and ongoing maintenance, so it’s important to ensure readiness before proceeding.

Explaining Key Concepts in Customer Attrition Analysis

What is Customer Attrition?

Customer attrition (or churn) is the loss of customers over a given period. It is typically expressed as a churn rate:

\( \text{Churn Rate} = \frac{\text{Number of Customers Lost in Period}}{\text{Total Customers at Start of Period}} \)

High churn rates can signal service issues, product-market misfit, or competitive pressures. Retaining existing customers is almost always less expensive than acquiring new ones, making churn reduction a key business lever.

Descriptive vs. Predictive Analytics

• Descriptive Analytics: Answers “what happened?” and “why did it happen?”
• Predictive Analytics: Answers “what is likely to happen next?”

Descriptive analytics provides the foundation for understanding churn by summarizing historical data and identifying patterns. Predictive analytics builds on these insights to forecast future churn, enabling proactive retention strategies.

Root Cause Analysis (RCA)

Root Cause Analysis in churn projects involves investigating the underlying factors that lead customers to leave. RCA dashboards typically visualize multiple dimensions:

• Customer demographics
• Product usage trends
• Support interactions
• Feedback and complaints
• External events (e.g., competitor launches, economic factors)

The goal is to move beyond surface-level symptoms and uncover actionable insights.

Predictive Modeling Process

1. Data Preparation: Cleaning, transforming, and creating features from raw data.
2. Model Training: Fitting a model to historical data with known outcomes.
3. Evaluation: Using metrics such as precision, recall, F1-score, and ROC-AUC to measure model performance.
4. Deployment: Integrating the model into real-time or batch processes for scoring customers.

For example, a simple logistic regression model for churn might take the form:

\( P(\text{Churn}) = \frac{1}{1 + e^{-(\beta_0 + \beta_1 x_1 + \beta_2 x_2 + \ldots + \beta_n x_n)}} \)

Where \( x_1, x_2, \ldots, x_n \) are input features, and \( \beta_0, \beta_1, \ldots, \beta_n \) are coefficients learned from data.

Customer Segmentation

After scoring customers by churn risk, segmentation divides them into actionable groups. For example:

• Very Risky: Probability of churn > 80%
• Medium Risky: Probability of churn 40-80%
• Low Risky: Probability of churn < 40%

This enables the business to tailor retention campaigns and allocate resources where they matter most.

My Solution Approach: Step-by-Step Guide

Step 1: Engage in Discovery and Clarification

Begin with a workshop or interview series with client stakeholders. Use open-ended questions to clarify:

• How is attrition measured?
• What are the business impacts?
• What interventions have already been tried?
• What data is available, and where does it reside?

Document findings and validate them with the client.

Step 2: Conduct Data Inventory and Quality Assessment

Work with the client’s data team to inventory all available data sources:

• Customer profiles and demographics
• Transactional and usage data
• Support and complaint logs
• Marketing and communication records

Assess data for completeness, accuracy, and recency. Use Python or SQL for initial data exploration:

import pandas as pd

# Load customer data
df = pd.read_csv('customer_data.csv')

# Check for missing values
print(df.isnull().sum())

# Summary statistics
print(df.describe())

Step 3: Start with Descriptive Analysis and RCA Dashboard

Develop visualizations (using tools like Power BI, Tableau, or Python’s matplotlib/seaborn) to surface key trends:

• Churn rates by product, segment, geography
• Time series of churn events
• Correlations between churn and customer behavior

Example Python code for plotting churn rates:

import matplotlib.pyplot as plt

# Calculate churn rate by segment
churn_by_segment = df.groupby('segment')['churned'].mean()

# Plot
churn_by_segment.plot(kind='bar')
plt.title('Churn Rate by Customer Segment')
plt.ylabel('Churn Rate')
plt.show()

Step 4: Share Insights and Discuss Options

Present findings to the client. Discuss:

• Are there obvious patterns that suggest actionable next steps?
• Is there appetite for deeper, predictive analysis?
• What are the business constraints (budget, resources, timelines)?

Step 5: If Needed, Move to Predictive Modeling

If descriptive analysis is insufficient, propose a predictive modeling project:

• Define the modeling objective (binary classification: churn vs. retain).
• Engineer relevant features from historical data.
• Split data into training and test sets.
• Experiment with simple models first (e.g., logistic regression), then try more complex algorithms.
• Evaluate models using cross-validation and appropriate metrics.

from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import roc_auc_score

# Feature matrix X, target y
X = df[['feature1', 'feature2', 'feature3']]
y = df['churned']

# Split data
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Train model
model = LogisticRegression()
model.fit(X_train, y_train)

# Predict probabilities
y_pred_proba = model.predict_proba(X_test)[:,1]

# Evaluate
print("ROC-AUC:", roc_auc_score(y_test, y_pred_proba))

Step 6: Enable Customer Segmentation and Actionable Dashboards

Use model outputs to segment customers and build dashboards that track risk segments over time. Integrate these dashboards into business processes so teams can act on high-risk customers immediately.

Step 7: Plan for Deployment, Monitoring, and Iteration

Discuss deployment options (batch scoring, real-time APIs, integration with CRM or marketing automation platforms). Estimate infrastructure needs and costs. Plan for model monitoring, periodic retraining, and feedback loops

for ongoing improvement. Emphasize the importance of close alignment between the data science, IT, and business teams to ensure successful adoption and measurable business impact.

Deep Dive: Detailed Steps and Considerations

Step 1: Discovery and Clarification (The Foundation)

The first and most critical step is to thoroughly understand the business context. This phase is often rushed or skipped by inexperienced data scientists, but it’s where the path to a successful project is laid.

• Stakeholder Interviews: Schedule sessions with key stakeholders—product owners, customer success managers, marketing leads, and sales teams. Their perspectives will help you grasp where, when, and why attrition hurts most.
• Process Mapping: Visualize the customer journey. Identify touchpoints where customers drop off or express dissatisfaction.
• Review Historical Interventions: What retention efforts have been tried before? What worked, what didn’t, and why?
• Define Project Boundaries: Is the scope to analyze all customers, or specific segments (e.g., high-value, new users)? Clarify time frames and desired outcomes (e.g., reduce churn by 10% in 6 months).

Step 2: Data Inventory and Quality Assessment

High-quality data is the bedrock of both descriptive and predictive analytics. At this stage, you should:

• Catalog Data Sources: List all potential sources—CRM systems, transaction logs, web/app analytics, support platforms.
• Assess Data Health: Check for missing values, inconsistencies, and outdated information.
• Evaluate Data Granularity: Is data available at the customer level? Are time stamps present?
• Secure Data Access: Work with IT and compliance teams to ensure privacy and security considerations are met (GDPR, CCPA, etc.).

Step 3: Descriptive Analytics and Root Cause Analysis

This is where you begin to bring data to life for your client. Use descriptive statistics and visualization to answer questions like:

• What is the monthly churn rate over the past two years?
• Are certain products, regions, or customer types more likely to churn?
• What are the most common support issues among churned customers?

Here’s a simple example using Python’s seaborn for visualization:

import seaborn as sns
import matplotlib.pyplot as plt

# Assume df has 'churned' (0/1), 'region', 'product' columns
sns.barplot(x='region', y='churned', data=df, estimator=np.mean)
plt.title('Churn Rate by Region')
plt.ylabel('Churn Rate')
plt.show()

Create a dashboard that integrates these insights using tools like Power BI or Tableau. A good RCA dashboard will:

• Visualize customer attrition trends over time
• Break down churn by key segments (e.g., by cohort, product line, geography)
• Highlight high-risk periods (e.g., after onboarding, after price changes)
• Allow drill-down to individual customer journeys

Statistical Analysis

You may also perform hypothesis testing to validate if observed patterns are statistically significant. For example, a t-test to compare average tenure between churned and retained customers:

from scipy.stats import ttest_ind

# Assume df has 'tenure' and 'churned'
tenure_churned = df[df['churned'] == 1]['tenure']
tenure_retained = df[df['churned'] == 0]['tenure']

t_stat, p_val = ttest_ind(tenure_churned, tenure_retained)
print("T-statistic:", t_stat, "P-value:", p_val)

A low p-value (<0.05) indicates a significant difference in tenure between groups.

Step 4: Insight Sharing and Action Planning

Hold a workshop with stakeholders to present findings. Use storytelling techniques—combine visualizations with concise explanations. Encourage discussion around:

• Immediate actions (e.g., customer outreach, process improvements)
• Areas needing deeper investigation
• Whether to proceed with predictive modeling

During this phase, your ability to communicate clearly and relate findings to business objectives will build trust and foster buy-in for more advanced analytics.

Step 5: Predictive Modeling for Churn (If Needed)

If the business seeks to proactively prevent attrition, predictive modeling is the next step. The goal is to predict, for each customer, the probability they will churn in a future period.

Feature Engineering

Good features are the key to model performance. Typical features for churn models include:

• Recency, frequency, and monetary value of purchases (RFM analysis)
• Product/service usage patterns
• Support ticket history
• Engagement with communications (emails, push notifications)
• Demographics and tenure

Modeling Algorithms

Start simple (logistic regression) and move to more complex models if needed (random forests, gradient boosting, neural networks). For interpretability, tree-based models often perform well and provide feature importance out-of-the-box.

Model Evaluation

Use appropriate metrics. In churn prediction, recall and precision for the “churn” class are usually more important than overall accuracy (since the class is often imbalanced). ROC-AUC is a common metric:

\( \text{ROC-AUC} = \text{Area under the Receiver Operating Characteristic curve} \)

Customer Risk Segmentation

Once you have probability scores from your model, segment customers:

df['churn_score'] = model.predict_proba(X)[:, 1]

def segment(score):
    if score > 0.8:
        return 'Very Risky'
    elif score > 0.4:
        return 'Medium Risky'
    else:
        return 'Low Risky'

df['churn_risk'] = df['churn_score'].apply(segment)

Use this segmentation to guide retention campaigns, focusing efforts and incentives on the “Very Risky” group.

Deployment Considerations

• Model can be deployed as a batch job (e.g., nightly scoring) or as a real-time API (for instant scoring in CRM systems).
• Dashboards should update automatically, with clear action triggers for business teams.

Discuss infrastructure costs (cloud compute, storage, ML platform fees) and ensure business teams are prepared for operational change.

Step 6: Monitoring, Feedback, and Continuous Improvement

Deploying a churn model is not the end—it’s the start of an iterative process. Track real-world outcomes to ensure the model’s predictions are accurate and useful.

• Monitor model performance (e.g., drift in prediction accuracy, changing customer behavior)
• Solicit feedback from users (e.g., sales or customer success teams)
• Retrain the model periodically as new data comes in

Step 7: Ensuring Business Alignment and ROI

Throughout all phases, maintain a focus on business alignment. Use clear KPIs to measure the impact of your work—such as reduction in churn rate, increase in customer lifetime value, or improved retention revenue.

• Present business cases for further investment in predictive analytics
• Quantify ROI using before/after analyses
• Adjust strategies based on what works in practice

Common Pitfalls and How to Avoid Them

• Jumping to Machine Learning Too Soon: Not every churn problem needs a predictive model. Sometimes simple descriptive analytics (even a well-designed dashboard) provides all the actionable insights required.
• Poor Data Quality: Garbage in, garbage out. Always start with a thorough data audit.
• Lack of Business Buy-in: Data science solutions are only as good as their adoption by business teams. Engage stakeholders early and often.
• No Plan for Maintenance: Models degrade as customer behavior changes. Plan for monitoring and retraining from day one.
• Overcomplicating Solutions: Begin with simple models and interventions. Only add complexity if needed and justified.

Best Practices for Data Science Projects on Customer Attrition

• Start with the Problem, Not the Solution: Understand the business pain before suggesting any technical approach.
• Iterative Delivery: Deliver insights continuously, not just at project end. Early wins build momentum and trust.
• Transparent Communication: Explain assumptions, limitations, and uncertainties. Use visual aids and analogies for non-technical audiences.
• Collaborate Across Functions: Work closely with marketing, sales, support, and IT. Churn is a cross-functional challenge.
• Document Everything: Keep detailed records of data sources, modeling decisions, and business feedback. This streamlines future work and troubleshooting.

Case Study: Illustrative Example

Consider a SaaS company with high churn among small business clients. The company’s leadership is unsure why this is happening. Here’s how a data scientist might apply the above process:

1. Discovery: Interviews reveal that churn often occurs after the first billing cycle.
2. Data Audit: Data includes product usage logs, support tickets, and customer demographics.
3. Descriptive Dashboard: Shows that customers who don’t use the platform in the first week are much more likely to churn.
4. Actionable Insight: Recommend a new onboarding campaign targeting users within their first week.
5. Predictive Modeling: For further improvement, build a model using first-week engagement metrics to predict churn, enabling proactive outreach.

This approach delivers value at every stage, from quick wins to long-term predictive power.

FAQs: Data Science and Customer Attrition

1. What are the most important features for churn prediction?

Features capturing customer engagement (e.g., login frequency, purchase activity), support interactions, account age, and demographic information are typically most predictive.

2. How do you deal with imbalanced churn data?

Techniques include resampling (oversampling minority class, undersampling majority class), using class-weighted loss functions, or evaluating with metrics like ROC-AUC, precision, and recall rather than accuracy.

3. How often should a churn model be retrained?

Retrain at least quarterly, or whenever there are significant business changes (e.g., new pricing, major product updates) that might shift customer behavior patterns.

4. What’s the biggest mistake in churn analytics?

Jumping straight to technical solutions without understanding the business context and what actions the business can realistically take.

Conclusion: The Right Approach to Customer Attrition in Data Science

Addressing customer attrition is not just about building sophisticated algorithms—it's about understanding business needs, delivering actionable insights, and enabling the organization to act. The most successful data scientists recognize that:

• Problem scoping and stakeholder engagement are just as important as modeling.
• Descriptive analytics and RCA dashboards are often enough to drive impactful change.
• Predictive modeling should be considered when business complexity and scale justify it.
• Success depends on clear communication, cross-functional collaboration, and a focus on business outcomes.

By following a structured, step-by-step approach—starting from discovery through to deployment and monitoring—you’ll not only answer your client’s questions, but empower them to reduce attrition, boost retention, and drive sustainable growth.

Further Reading & Resources

• How to Develop a Customer Churn Prediction Model in Python
• Churn Prediction: Predictive Analytics and Machine Learning
• Tableau: Customer Churn Analytics
• scikit-learn: Supervised Learning

Summary Table: Descriptive vs. Predictive Solutions for Customer Attrition

Final Thoughts

Data science is at its best when it bridges the gap

between business problems and actionable solutions. Addressing customer attrition is a prime example of how effective data science extends far beyond algorithms and coding—it’s about collaborating, communicating, and delivering value. By investing time in the discovery phase, leveraging descriptive analytics for initial insights, and only advancing to predictive modeling when justified, you’ll ensure your client’s needs are met with clarity and impact.

Remember, the journey doesn’t end with deploying a churn prediction model. Ongoing monitoring, stakeholder education, and iterative improvement are essential to maintain long-term success. By focusing on both the “what” (understanding attrition patterns) and the “why” (enabling business action), data scientists can become invaluable partners in driving customer loyalty and business growth.

Above all, never underestimate the underrated phase of scope definition and business engagement. It’s here that trust is built, the right questions are asked, and the most fitting solutions are found. Whether your client needs a powerful dashboard to surface actionable insights or a sophisticated machine learning pipeline for proactive retention, your ability to guide them through these options—and align analytics with their true goals—will set you apart as a data science professional.

Practical Tips for Data Scientists Facing Customer Attrition Problems

• Listen First, Analyze Second: Always begin with in-depth conversations to understand business pain points before touching the data.
• Visualize Early: Create simple visualizations as soon as possible. They can reveal trends, spark discussions, and quickly engage stakeholders.
• Document Assumptions: Clearly state what you know, what you don’t, and how assumptions might affect outcomes.
• Choose Simplicity: Simple models and interventions are easier to explain, implement, and maintain. Complexity should only be added when absolutely necessary.
• Build for Action: Ensure that your deliverables—whether dashboards or predictive tools—fit seamlessly into business workflows and empower teams to act.
• Measure Impact: Define and track KPIs, such as reduction in churn, increased retention, or improvements in customer satisfaction, to demonstrate the value of your work.
• Communicate Continuously: Keep stakeholders informed throughout the project lifecycle with regular updates, demos, and feedback sessions.

Sample Interview Answer: How Will You Proceed with a Customer Attrition Problem?

Here’s a concise, interview-ready answer that incorporates the key points discussed in this article:

Suppose a client reports severe customer attrition but is unclear on what to do. I would start by engaging in detailed discussions to clarify their definition of attrition, understand business impact, and identify available data. I’d propose two main approaches: First, a descriptive solution—building a root cause analysis dashboard and conducting pattern analysis to surface key drivers of churn. If actionable insights emerge, the business can often address the problem directly. If patterns are complex or the client needs to predict future churn, I’d recommend a predictive solution—assessing data quality, engineering features, and building a machine learning model to score and segment customers by churn risk. This enables targeted retention strategies. Throughout, I’d ensure alignment on goals, feasibility, costs, and deployment, maintaining open communication and focusing on solutions that drive business value.

Customer Attrition: Glossary of Key Terms

• Churn (Attrition): The rate at which customers leave a product or service over a specific period.
• Descriptive Analytics: Analytical techniques that summarize historical data to understand what happened.
• Predictive Analytics: Analytical approaches that forecast future outcomes based on historical data.
• Root Cause Analysis (RCA): Systematic process to identify the underlying factors behind a problem.
• Feature Engineering: Creating new variables from raw data to improve model performance.
• ROC-AUC: Receiver Operating Characteristic - Area Under the Curve, a metric for classification model performance.
• Customer Segmentation: Dividing customers into groups based on similar attributes or behaviors.
• Model Drift: When a predictive model’s performance degrades due to changes in underlying data patterns.

Conclusion: Your Roadmap to Effective Customer Churn Analysis

Tackling customer attrition effectively is about much more than deploying machine learning models. It requires a holistic, business-first approach—starting with discovery and scope definition, moving through descriptive analysis and actionable insights, and only advancing to predictive analytics when justified by business needs and data readiness.

As a data scientist, your role is to guide the client through these options, ask the right questions, and build solutions that not only identify but also enable action on the root causes of churn. Whether you’re designing RCA dashboards or deploying real-time predictive models, remember that your ultimate goal is to help the business retain more customers and grow sustainably.

Stay curious, communicate clearly, and always keep the business context front and center. That’s how you’ll deliver truly impactful data science solutions for customer attrition—and stand out in any interview or real-world project.

Further Steps for Aspiring Data Scientists

• Practice exploratory data analysis (EDA) and dashboard creation on open churn datasets (e.g., the Telco Customer Churn Dataset on Kaggle).
• Simulate stakeholder conversations with peers to refine your business problem scoping skills.
• Stay updated on advances in explainable AI and model monitoring for production systems.
• Read case studies from industries like telecom, SaaS, and banking for practical churn reduction strategies.

Ready to Tackle Your Next Data Science Interview?

Prepare to explain not just how you would build a churn model, but also how you would guide a client from confusion to clarity—helping them choose between descriptive and predictive solutions, emphasizing actionable insights, and ensuring business alignment every step of the way.

With this holistic approach, you’ll demonstrate both technical and consulting acumen, setting yourself up for success in data science interviews and real-world projects alike.

Good luck! And remember: In data science, the journey to clarity is as important as the solution itself.

Related Articles

• Data Science Interview Question - Ecommerce
• Data Science Interview Question - Retail
• Data Science Interview Question - Banking
• Data Scientist Interview Question - Amazon
• Data Science Interview Question - Healthcare