My first Maven Data Challenge: Cafe Rewards Offers

Maven Cafe Reward Project

Overview: Maven Coffee conducted a 30-day test involving various promotional offers sent to existing rewards members.

Test Details:

• Duration: 30 days
• Frequency: 6 rounds of offers sent to customers
• Offer Variations: 10 combinations, including types, durations, difficulties, rewards, and channels

The goal was to identify key customer segments and develop a data-driven strategy for future promotional messaging and targeting.

Key Insights

Based on activity and spending, four key customer segments were identified:

1. Active and High-Spending Customers (21% of Customers):

   • Activity Level: Above average
   • Spending: Higher than average
   • Best Offer: Discount offer on a $10 spend with a $2 reward via all channels
   • Insight: This segment was highly responsive and engaged more with offers.

2. Highly Active but Low-Spending Customers (26% of Customers):

   • Activity Level: High
   • Spending: Low
   • Offer Sensitivity: Higher spending rate in connection to offers
   • Best Offer: Discount offer on a $10 spend with a $2 reward via all channels
   • Insight: Despite low spending, this group frequently utilised offers.

3. Low-Activity but High-Spending Customers (17% of Customers):

   • Activity Level: Low
   • Spending: High
   • Offer Sensitivity: Low spending rate in connection to offers
   • Best Offer: Discount offer on a $20 spend with a $5 reward (offer with the highest minimum spend ) via web and e-mail channels.
   • Insight: This group shows high spending regardless of offers but responds better to more valuable offers. Due to the high average age, web and email channels work better than mobile and social channels.

4. Low-Activity and Low-Spending Customers (36% of Customers):

   • Activity Level: Below average
   • Spending: Lower than average
   • Offer Sensitivity: High (allocate a significant portion of their spending to offers)
   • Best Offer: Discount offer on a $7 spend with a $3 reward via all channels
   • Insight: These customers are not very active or high spenders, they do make substantial use of the offers available.

Strategy for Future Promotions:

Channel Performance:

• Best Performing Channels: Combination of all channels
• Insight: The highest number of successfully completed offers was achieved through a multi-channel approach. Older, high-value customers responded best to web and email channels.

Recommendations for Offer Details:

• Offer Duration: Avoid sending offers with a validity of less than 3 days, as the average time from receipt to completion is 3 days.

To optimize future promotions:

• Tailor offers to target the identified customer segments more effectively.
• Active and High-Spending Customers: Use multi-channel strategies with the $10 spend-$2 reward offer.
• Highly Active but Low-Spending Customers: Increase the frequency of value-focused promotions to boost spending.
• Low-Activity but High-Spending Customers: Emphasize higher-value offers through web and email channels.
• Low-Activity and Low-Spending Customers: Align offers with spending habits, leveraging the $7 spend-$3 reward offer.

Customer Segmentation Methodology:

We focused on 2 main factors that couls drive succes promotion: the activity and the spending os the customers. Based on the customer's activity - which indicates their interest in the offers- and the customer's value - the total amount spent by the customer-.

Based on activity and spending, four key customer segments were identified:

1. Active and High-Spending Customers: This segment views more offers than average and has above-average spending.
2. Highly Active but Low-Spending Customers: This segment views more offers than average but has lower-than-average spending.
3. Low-Activity but High-Spending Customers : This segment views more offers than average but has lower-than-average spending.
4. Low-Activity and Low-Spending Customers: This segment views less offers than average and has lower-than-average spending.

---

Methodological comment

To validate whether this segmentation was realistic, we performed K-means clustering on the data. I created four clusters and used a scatter plot to visualize and verify that the customer segments aligned with the results of the K-means analysis. The results clearly show the identified four customer groups.

Technical Notes

• Data Issues:

  • Offer Data: Some timestamps were associated with multiple offer-driven transactions; however, this did not significantly affect results.

  • Customer data: Addressing missing demographic data. Total customer base: 17,000; offers received: 16,994; demographic data available for: 14,825. For the demographic analyses, the smaller base population was used, while the revenue analysis was performed with all customers included.

  • Visualization Notes: Power BI's public visualization does not support key factor analysis. A summary of the key indicators has been prepared and is presented in the public report.