- Analytics Wisdom
- Posts
- Data Science Segmentation for Credit Card Retention Strategies
Data Science Segmentation for Credit Card Retention Strategies
Targeted Retention: How Data Can Shape Credit Card Offers
S P
18 Nov
Introduction
Welcome to the latest edition of Analytics Wisdom. In this issue, we cover a case study for a multinational credit card company and advise them on targeted retention strategies for their users.
Case Study
Due to the year coming to an end, credit card company, Kapital Two wants to reduce the amount of churn within their users and hence decides to introduce a retention offer. The standard annual credit card fee is $399, and the retention offer is respectively discounted to $199.
The offer will be sent through their weekly newsletter as a special edition and they wish to offer the heavy discounted fee of $199 to a selective group of users that prove to be high value.
Goal
The aim is to assist Kapital Two in identifying their most valuable users. Since there are limited seats available for the discounted offer, we need to narrow down the users even more. We will only include those who consistently open their emails to improve the chances of them redeeming the retention offer.
We’ve been provided the data for the user id, week , email open rate, click rate, and revenue generated by the user for 8 weeks.
Our Analytical Approach
In tackling this case study, we employ focus on in-depth analytics to accurately segment customers. Our strategy is centered on leveraging email data, correlating key metrics such as open rates and click rates with revenue generation, to identify the most valuable customer segments.
Rolling Metric Analysis: We calculate rolling averages for revenue and engagement metrics to capture trends over time, providing a dynamic view of customer behavior.
Percentile-Based Segmentation: By assessing the 25th, 50th, and 75th percentiles of revenue and engagement, we effectively categorize customers into different value tiers.
Trend Comparisons: We plot and compare various trends, such as the average revenue contribution of high-value versus other users, to understand the impact of each segment.
Categorizing High Value Users
We will define high-value users as those whose revenue generation and engagement metrics every week were above certain percentile thresholds. Specifically, users who were above the 75th percentile in terms of total revenue and had higher than average open and click rates were categorized as high-value.
This approach will ensure that we can target users who were not only contributing significantly in terms of revenue but also consistently engaged with the content.
Total Revenue Generation: We look at the total revenue each user generated over the 8-week period. Users are considered high-value if their total revenue was at or above the 75th percentile compared to all users. This means we focus on the top 25% of users in terms of revenue generation.
Engagement Metrics (Open and Click Rates): We also consider how engaged users are with the newsletter content. This engagement is measured using two key metrics: the average open rate and the average click rate of the newsletters.
Here’s the code for sorting users as high value or not using both criterias, rolling revenue and rolling open rates.
Now to Visualize Some Stats
Quantitative Insights and Takeaways:
High-Value User Composition: While they make up a smaller percentage of the total user base, high-value users are crucial due to their larger economic impact. Even a small shift in their behavior can have a significant revenue implication.
Revenue Contribution: The consistent gap in average revenue between high-value and other users suggests that targeting high-value users could yield a higher return on investment for marketing campaigns, such as the discounted annual fee offer.
Stability in User Classification: The percentage of high-value users does not show large spikes or dips, which could imply that once a user becomes high-value, they are likely to remain so. This stability is beneficial for forecasting and planning marketing initiatives.
Marketing Strategy: Given the consistent performance and higher revenue contribution of high-value users, focusing the limited seats of the discounted tier on them is likely to maintain or increase their loyalty and spending, enhancing long-term revenue potential.
Let’s dig in even further. The retention offer will have a higher chance of being redeemed by users that are consistent at opening their emails. Hence, we now look the percentile values and segment users by the percentile of their rolling open rate.
Here’s what this graph tells us.
Higher Engagement and Response Likelihood: The P50 high-value users represent those who are consistently above the median in terms of open rates among the already engaged high-value segment. This suggests they are more likely to engage with and respond to the offer, maximizing the potential impact of the campaign.
Optimized Resource Allocation: With limited seats available, it's crucial to focus resources on users who are most likely to take advantage of the offer. Targeting the P50 high-value users allows for a more efficient use of resources, ensuring the offer reaches those most inclined to appreciate and use it.
Enhanced Customer Lifetime Value (CLV): High engagement users are often more loyal and have a higher CLV. By focusing on this group, the company can strengthen relationships with its most valuable customers, potentially leading to increased long-term revenue.
Reduced Risk of Offer Saturation: Spreading the offer too thinly across a broader group could dilute its perceived value. Concentrating on the P50 high-value users ensures the offer retains its exclusivity and perceived value, making it more appealing.
Conclusion
Hence targeting the P50 high-value users instead of all users for the limited discounted annual tier offer would be a strategic approach for these very reasons.
In conclusion, the company should prioritize high-value users for the campaign due to their significant and consistent contribution to revenue, as well as their stable representation within the user base.
