Maven Rewards Challenge - Cafe Segmentation Tool

Maven Rewards Challenge: Customer Segmentation and Offer Performance Analysis

Executive Summary

We conducted a comprehensive test by sending different combinations of promotional offers to our existing rewards members at Maven Cafe. The primary goal of this test was to gain insights into customer behavior, segment our customer base effectively, and optimize our future promotional strategies.

The key findings from our analysis have significant implications for how we engage with our customers and tailor our offers. This report outlines the major outcomes of our customer segmentation efforts, offer performance evaluation, and strategic recommendations for future campaigns. For a hands-on exploration, please visit the following link to interact with our data and insights directly: Maven Rewards Analysis App.

1. Data Overview

Dataset

Our analysis was based on three key datasets that capture the behavior of our Cafe Rewards members over the 30-day period:

1. Offers Data: Details of the promotional offers sent to customers.

2. Customers Data: Demographic information of our rewards members.

3. Events Data: Records of customer activities, including transactions, offers received, viewed, and completed.

These datasets were meticulously cleaned and processed to ensure accurate and reliable insights.

2. Key Insights from the Analysis

Customer Segmentation

Using Recency, Frequency, and Monetary (RFM) analysis, we identified distinct customer segments within our rewards members. The segmentation revealed the following key insights:

• Cluster 1: This segment demonstrated the highest Customer Lifetime Value (CLV), making it our most valuable customer group. Tailoring offers specifically for this segment could further enhance their engagement and spending.

• Cluster 2: Customers in this segment had the highest average transaction amounts across all offer types, suggesting they are willing to spend more when incentivized appropriately.

• Cluster 3: Though slightly less valuable than Cluster 1, this segment shows strong engagement with informational offers, indicating they value content and communication from Maven Cafe.

Strategic Recommendations:

• Cluster 1 should be a focal point for high-value offers, possibly through exclusive deals or premium rewards.

• Cluster 2 can be targeted with offers that encourage higher spending, such as tiered discounts or bundled deals.

• Cluster 3 might benefit from content-driven offers, leveraging their interest in informational engagement.

Offer Performance

We evaluated the effectiveness of different offer types—Buy One Get One (BOGO), Discount, and Informational—across various customer segments:

• BOGO offers achieved the highest completion rates across all customer segments, making them the most effective promotional tool.

• Informational offers had the highest viewed rates, particularly in Clusters 2 and 3, indicating they are effective in capturing customer attention but may need enhancement to drive completions.

Strategic Recommendations:

• Focus on BOGO offers for broad-based campaigns, as they consistently drive high engagement and completions.

• Reevaluate Informational offers by possibly integrating them with rewards or discounts to boost completion rates.

Channel Performance

Our analysis of promotional channels revealed the following:

• Social Media: Drove the highest average transaction amounts, indicating that customers are more likely to spend more when offers are communicated through these platforms.

• Web: Achieved the highest offer completion rates, suggesting that our website is a reliable channel for converting customer interest into action.

Strategic Recommendations:

• Dual-channel strategy: Use Social Media for higher spending campaigns and the Web for broad reach and consistent offer completion.

• Explore cross-channel promotions that combine the strengths of Social Media and Web channels to maximize both engagement and conversion.

3. Recommendations for Future Campaigns

1. Segment-Specific Strategies:

• Cluster 1: Offer exclusive, high-value promotions to maintain and enhance their engagement.

• Cluster 2: Provide offers that encourage higher spending, such as bundled deals or premium products.

• Cluster 3: Focus on content-driven campaigns that inform and engage, possibly tied to specific rewards.

2. Channel Optimization:

• Leverage the Web for reliable completion rates and Social Media for driving higher transaction values.

• Consider cross-channel promotions that leverage the strengths of multiple channels to enhance overall campaign effectiveness.

3. Offer Refinement:

• Continue to focus on BOGO and Discount offers, as they show the highest effectiveness.

• Reevaluate and possibly enhance Informational offers by integrating rewards or incentives to drive higher completion rates.

Conclusion

The Maven Rewards Challenge provided valuable insights into our customer base and the effectiveness of our promotional strategies. By leveraging these findings, we can tailor our offers and campaigns to better align with customer preferences, thereby driving higher engagement and maximizing revenue.

To explore the data and insights interactively, visit the Maven Rewards Analysis App: Maven Rewards Analysis App.

This strategic approach will ensure that Maven Cafe remains competitive in the market, with a loyal and engaged customer base. We look forward to implementing these strategies and continuing to grow the success of our rewards program.

Summary: Optimizing Streamlit App Performance

Issue:

The Streamlit app was slow, especially during machine learning tasks and data processing, due to heavy and repeated computations.

Causes:

1. Heavy Computation: Resource-intensive clustering algorithms like KMeans.
2. Repeated Computation: Machine learning models and data processing steps rerun with each user interaction.

Solution:

1. Storing Results: Switching from csv files to using Parquet files a column-oriented data file format designed for efficient data storage and retrieval.
2. Data Caching: Used @st.cache_data in Streamlit to cache database results, minimizing repeated computations.

These optimizations significantly improved the app’s performance by reducing redundant operations and efficiently retrieving stored results.