This project reflects my approach to real-world analytical projects at my current company. It involves understanding business goals to align reports and analyses with stakeholder expectations, ensuring the right questions are answered. I also deep dive into different user segments to uncover insights, providing a clearer understanding of customer behaviors and suggesting appropriate actions for each group.

You can find the full code on GitHub here.

There are two important business question to answer in this project:

I. What are the most effective approaches for sending customer offers?

II. What strategies are recommended for targeting specific customer segments?

About The Dataset

The Starbucks Customer Dataset is a dataset designed for customer segmentation and offer performance analysis. It includes the following files:

• Portfolio data: details of promotional offers sent to customers, including offer type (discount, BOGO (Buy One Get One), informational), reward amount and duration.

• Profile data: contains customer demographic info such as age, income, gender and membership tenure.

• Transcript data: logs of customer response with offers (receiving/ viewing/ completing an offer), and also regular transactions.

Exploratory Data Analysis (EDA)

To understand the relationships between variables, I performed univariate and bivariate analyses.

• Univariate analysis provided statistical summaries of all variables within each table.

• Bivariate analysis was applied to customer profile data, examining the relationships between age, gender and income. Similarly, bivariate analysis was applied to transcript data to track how customer activities evolved during the campaign.

Key findings from this EDA part:

Customer Demographics

• Gender Distribution: The customer base is predominantly male (57%).

• Age Distribution: Customer ages range from 18-101, with a roughly normal distribution centered around a mean of 54 and a standard deviation of 17. The majority of customers (60%) fall within the age range of 46-75.

• Income Distribution: Customer incomes range from $30k-120k, with a right-skewed distribution. The mean income is $65.4k with a standard deviation of $21.6k.

• Income by Age Group: Customers aged 18-35 have an average annual income of around $51k, while those aged 36-55 earn about $65k on average. For customers aged 56 and older, the average income is slightly higher at approximately $70k.

• Gender and Age Correlation: There is a higher proportion of younger customers among males compared to other genders.

Transactional Activities

• This table includes approximately 140,000 transactions, which is slightly more than 4 times the number of offers completed. This indicates that many transactions occurred without the use of offers.

• Customers received 76,000 offers, with 76% viewed and 44% completed. Offer completions peaked on the day offers were sent, followed by a brief 2-day increase in transactions. This suggests that well-timed offers can temporarily boost customer purchases.

These analyses provided valuable insights into customer behavior and their interaction with offers, laying the groundwork for deeper analysis of offer popularity and customer segmentation.

Identifying Popular Offers & Key Success Factors

To address the first business question, I analyzed which offers performed best and what factors contributed to their success. The popularity of an offer was defined by its completion rate - the number of times an offer was completed divided by the number of times it was received. Key findings:

• Overall Completion Rate: The top 3 offers by popularity were Offer F, E, and B, with overall completion rates of 70%, 67.4%, and 56.7%.

• View Rate (viewed/received): The most significant factor influencing the view rate of offers is the promotion channel used: Social media and email were the most effective channels, with mobile also contributing significantly. The web had the least impact on view rates.

• Completion Rate (completed/viewed): The type of offer is the primary determinant of its completion rate. Customers preferred 'discount' offers over 'BOGO' offers, which means they are more responsive to promotions that provide immediate savings.

Recommendations for improving offer effectiveness:

The company should distribute offers at regular intervals to help establish and reinforce purchasing habits among customers.

To maximize the overall offer completion rate, efforts should be focused on improving both the offer view rate and completion rate. The key factors influencing these rates are promotional channels, offer types, and offer durations.

Improving Offer View Rate:

• Leverage Social Media: Social media has proven to be the most effective channel for offer views.

• Utilize Email and Mobile Apps: If the budget allows, send offers via email and mobile apps, as these channels also show strong performance in driving views.

Improving Offer Completion Rate:

• Prioritize Discount Offers: Customers across all age, gender, and income groups tend to prefer discount offers over BOGO. Sending more discount-based offers can boost completion rates.

Customer Segmentation

To answer the second business question - recommending strategies for specific customer groups - I performed customer segmentation based on their behavior using K-Means Clustering. The variables used for clustering included the number of offers viewed, the number of offers completed, the number of transactions made, and the total amount spent.

Using both the Elbow method and Silhouette score, I determined that 5 clusters would provide the best segmentation. Key metrics were also calculated for each cluster to guide the final interpretation and inform strategies for each customer group, including:

• Frequency: The number of transactions made during the campaign month.

• Average Purchase Value (APV): The average amount spent per transaction.

• Offer Completion Rate: The ratio of offers completed to offers viewed.

• Sensitivity to Offers: The ratio of offers completed to total transactions made.

Insights from Clustering & Recommendations

Once K-Means clusters are created, here are some customer segments that I analyzed based on their average transaction frequency, APV, offer completion rate and offer sensitivity:

Cluster 1 – Most Valuable Customers

• Customers in this cluster are highly valuable due to their frequent purchases (12.2 on average, around one purchase every 2-3 days) and high spending, with the highest APV of $18.3. Their Offer Completion Rate is high (92%), and they show a moderate level of Offer Sensitivity (34%), meaning they respond well to offers but are not overly reliant on them.

• Recommendation: These customers generate significant revenue and should be prioritized for retention strategies. The company can further engage them by encouraging more frequent purchases, perhaps through exclusive promotions.

Cluster 0 – Responsive Purchasers

• Customers in this cluster make frequent purchases (7.7 on average) with a high APV of $18.1. They have the highest Offer Completion Rate of 100%, indicating that, on average, they used every offer they viewed, showing strong responsiveness to offers. Their Offer Sensitivity is moderate (37%), suggesting they often use offers to make purchases.

• Recommendation: These customers are valuable due to their responsiveness, and the company should continue targeting them with well-tailored offers to maintain engagement.

Cluster 2 – Moderate Spending, Moderate Engagement

• Cluster 2 customers make fewer purchases (4.5 on average) and have a lower APV of $11.1. Their Offer Completion Rate is low at 32%, and their Offer Sensitivity is similar at 32%, indicating they are less likely to use offers for their purchases.

• Recommendation: This group may need stronger incentives or better-targeted offers to increase engagement, as they are less influenced by existing promotions and spend less per transaction.

Cluster 3 – Occasional Low-Spenders

• Customers in this cluster make only 5 purchases on average and have the lowest APV of $6.9. They are the least responsive to offers, with both an Offer Completion Rate and Offer Sensitivity of 31% and 13%, respectively.

• Recommendation: These customers are less engaged with both offers and purchasing overall. The company could focus on nurturing this group by offering significant discounts or more personalized offers to increase both their purchase frequency and spend.

Cluster 4 – Regulars

• This cluster consists of the company’s most frequent purchasers, with an average of 17.1 transactions during the campaign, showing strong loyalty. However, they have the lowest APV of $4.2, indicating they consistently make low-value purchases. Their Offer Completion Rate is moderate at 58%, but their Offer Sensitivity is very low at 12%, showing that offers do not significantly drive their purchasing behavior.

• Recommendation: These regular customers are crucial for stable revenue, but the company could focus on increasing their APV by encouraging them to try higher-value products through targeted promotions or sampling programs.

Demographic data was then analyzed to provide a clearer profile of each customer segment, key insights summarized:

• Responsive-Purchasers and Most-Valuable are older and wealthier, with balanced gender ratios.

• Regulars are younger, lower-income, and heavily skewed towards male customers.

• Low-Spenders and Moderate-Spenders also show male dominance but have a broader spread across income and age groups.