▤ 목차
Data
The dataset provided consisted of six interconnected tables, each containing crucial information about orders, products, customers, and other related attributes. To better understand the relationships within the data, an Entity-Relationship Diagram (ERD) was created, which highlighted the unique and foreign key relationships between the tables. For example, order_id was identified as the primary key in the order table and as a foreign key in the order_product_prior table, allowing us to merge these datasets effectively. This consolidation enabled a unified view of orders and their associated products.
Following the merging process, data cleaning was performed to improve reliability and accuracy. Duplicates and null values were systematically removed to ensure that the dataset was consistent and free from redundancy. This step was critical for avoiding biases or inaccuracies during the analysis phase.
In the exploratory analysis, we discovered notable patterns in ordering behavior. Most orders were placed during business hours, particularly between 9 AM and 4 PM. Weekly trends showed orders were evenly distributed across weekdays, with pronounced peaks on Sundays and Mondays. A deeper analysis combining day-of-week and time-of-day dimensions revealed that Mondays between 9 AM and 3 PM saw the highest order volumes, while Sunday afternoons (12 PM to 3 PM) were also highly active.
Reordering behavior emerged as a significant trend. Customers appeared to reorder items rather than purchasing entirely new products. This insight suggested a recurring pattern where customers frequently replenished items they consumed regularly. A closer examination of the reorder frequency confirmed this hypothesis, indicating that many products were reordered within a 7-day cycle.
Further analysis by department and aisle highlighted which categories were in high demand. Dairy, fresh produce, and snacks consistently ranked as top-selling categories, reflecting customer preferences for staple and perishable goods. In contrast, categories such as pet products, bulk items, and personal care saw relatively lower demand, underscoring the primary focus on consumable essentials.
These findings not only provided insights into customer behavior but also laid the groundwork for developing targeted strategies to improve product availability, optimize delivery operations, and enhance customer satisfaction.
Positioning VIPs
Our analysis identified that customers frequently reordered fresh items within a 7-day period, indicating consistent consumption patterns. This behavior not only contributed to product turnover but also increased revenue through service and delivery fees. Recognizing the importance of these recurring transactions, we defined VIP customers as those who significantly impacted revenue through their high order frequency.
To position these VIPs, we analyzed the order frequency (the number of times a customer reordered). The data revealed that customers who placed at least 40 orders had an outsized influence on revenue generation. While these high-frequency customers represented only 8.68% of the total customer base, they accounted for a remarkable 32.97% of total revenue. This disproportionate contribution underscores the critical role these customers play in driving business growth.
To further validate our findings, we applied the RFM (Recency, Frequency, Monetary) model, a widely-used framework for customer segmentation. The analysis confirmed that our VIPs consistently exhibited high recency, indicating they had ordered recently; high frequency, reflecting their frequent purchases; and significant monetary value, showcasing their strong buying power. These attributes reinforce their importance as a key segment of customers who order numerous times within short intervals.
By identifying and understanding this VIP segment, we can develop targeted strategies to enhance their experience, such as personalized promotions, loyalty programs, and exclusive offers, ensuring continued engagement and revenue growth.
What is RFM (Recency, Frequency, Monetary)?
RFM (Recency, Frequency, Monetary) is a customer segmentation technique used in marketing and data analysis to evaluate and categorize customers based on their transactional behavior. By analyzing these three metrics—recency, frequency, and monetary value—businesses can gain deeper insights into customer engagement and identify high-value segments. Each dimension provides critical information about customer activity, enabling companies to create personalized strategies for retention, cross-selling, and upselling.
- Recency measures how recently a customer made their most recent purchase. This metric is crucial because customers who have engaged with the business recently are more likely to respond to marketing efforts or make additional purchases. The idea is to prioritize customers with higher recency scores as they are typically more engaged and active.
- Frequency assesses how often a customer makes purchases over a specified period. A higher frequency indicates a loyal customer who interacts with the business consistently. Understanding frequency helps identify the most committed customer segments and allows businesses to nurture relationships with these repeat buyers.
- Monetary evaluates the total spending of a customer over a specific timeframe. Customers with higher monetary values contribute significantly to revenue and often represent the most profitable segment. This metric helps businesses focus on customers who provide the greatest financial return.
By combining these three dimensions, businesses can create an RFM score for each customer, typically calculated by assigning numerical values (or scores) to each metric and then combining them. Customers with high RFM scores are often classified as VIPs or key accounts, deserving special attention and tailored strategies to maintain their loyalty.
RFM analysis is a versatile tool that can be applied across industries, including retail, e-commerce, and subscription-based services. It not only helps in identifying valuable customers but also informs targeted campaigns, predictive modeling, and resource allocation. Overall, RFM is a data-driven approach that empowers businesses to optimize their marketing efforts and maximize customer lifetime value.
VIP Purchase Patterns
Our analysis of VIP purchasing behavior revealed that this customer segment predominantly buys fresh produce, dairy products, and snacks. These purchases often occur on Mondays between 9 AM and 11 AM, with most items reordered within a 7-day cycle. The nature of these products—perishable and frequently consumed—makes them integral to the company’s revenue. However, their consistent demand also highlights an opportunity to expand VIP purchasing behavior beyond these high-revenue items to underperforming product categories.
To diversify VIP purchases, we examined categories with lower sales volumes that are not tied to specific customer demographics, such as pet owners or families with young children. This analysis identified household items, personal care products, alcohol, and bulk goods as areas of potential growth. These categories present untapped opportunities, as they cater to a broader audience and have a longer shelf life compared to perishable goods. By encouraging VIPs to explore these categories, we aim to enhance overall revenue and optimize inventory turnover.
To achieve this, we propose implementing a targeted A/B testing strategy. The core idea revolves around offering discount coupons to VIP customers with the following parameters:
- Discount Coupons for VIPs: Provide exclusive coupons as an incentive, reinforcing their status as valuable customers and fostering loyalty.
- Focused Discounts on Low-Sale Products: Limit the coupons to apply only to the identified low-performing categories (household items, personal care, alcohol, and bulk goods). This ensures that the promotional effort directly addresses underperforming segments rather than cannibalizing sales from already high-demand products.
- Time-Limited Coupons: Activate the coupons on Sunday and set a 48-hour expiration window. The urgency created by this limited timeframe leverages loss aversion—a psychological principle where customers are more likely to act to avoid losing out on an opportunity.
We hypothesize that this approach will encourage VIP customers to expand their purchasing habits, resulting in increased sales of low-performing items. The combination of targeted discounts and time-sensitive offers is expected to drive short-term revenue growth while also improving the lifecycle and turnover of underperforming products.
By monitoring the results of the A/B test, we can assess the effectiveness of this strategy and fine-tune our approach. If successful, this initiative could not only boost revenue from previously low-demand categories but also strengthen VIP loyalty through personalized incentives. Ultimately, this strategy aligns with the broader goal of optimizing the product mix and ensuring sustained growth across all categories.
Analysis Methodology
The analysis methodology for this report primarily involved examining transaction data to identify key customer behaviors and segmentations, specifically targeting VIP customers. The process began with data cleaning and merging, followed by exploratory data analysis (EDA) to uncover patterns in purchasing behavior. Descriptive statistics, such as order frequency and recency, were used to identify segments with high revenue potential. Additionally, the RFM (Recency, Frequency, Monetary) model was applied to further segment customers and identify those with the highest purchasing power.
We also employed A/B testing to assess the impact of potential promotional strategies, including targeted coupons for low-sale products. This methodology allowed us to evaluate how different customer segments (VIPs vs. non-VIPs) responded to incentives, and how such incentives influenced revenue growth.
Improvements
While our analysis provided valuable insights and actionable strategies, there are areas where further exploration and refinement could enhance the results. Below are the key areas for improvement:
- Identifying and Addressing Additional Challenges
- Beyond the primary focus on VIPs, there may be other underlying issues within the data that require attention. Identifying and solving these problems would present new challenges but could lead to innovative approaches and additional opportunities for growth. For instance, examining patterns among non-VIP customers could reveal untapped potential or overlooked behaviors that contribute to revenue. A strategy focused on increasing engagement and loyalty among non-VIPs could serve as a complementary initiative to the VIP-focused approach.
- Exploring Alternative Data Approaches and Methods
- Our analysis followed a structured methodology, but different analytical techniques could provide new perspectives and uncover additional insights. For example, adopting advanced machine learning models, clustering algorithms, or predictive analytics could help identify trends, segment customers more precisely, and forecast future behavior with greater accuracy. Expanding the scope of data analysis tools and methods could significantly enhance decision-making.
- Demographic-Based Insights
- While our approach prioritized behavioral patterns, incorporating demographic data could provide a deeper understanding of customer preferences. Investigating factors such as age, location, household composition, and income levels could reveal how these characteristics influence purchasing behavior. This additional layer of analysis could inform more targeted marketing strategies and product recommendations, ensuring that promotions are aligned with customer needs and preferences.
By addressing these areas, the analysis could become more comprehensive, enabling the development of more nuanced strategies. These improvements would not only strengthen the overall impact of our findings but also ensure that future initiatives are adaptable to a broader range of customer needs and market dynamics.
'AI > Project' 카테고리의 다른 글
| 2nd K-AI Manufacturing Competition (3) (1) | 2025.01.15 |
|---|---|
| 2nd K-AI Manufacturing Competition (2) (1) | 2025.01.15 |
| 2nd K-AI Manufacturing Competition (1) (2) | 2025.01.15 |
| 2nd K-AI Manufacturing Competition (0) (2) | 2025.01.15 |
| Marketing Strategy Proposal with Instacart Data Analysis (0) (0) | 2024.12.23 |
