The competitive landscape of e-commerce demands proactive measures to manage inventory, especially for clearance items like shoes at Kakaobuy. Traditional methods often lead to losses. This article explores how Kakaobuy employs machine learning to predict stagnation and automate promotions for shoes, exemplified by New Balance, ensuring optimal inventory turnover.
The Challenge of Stagnant Clearance Inventory
Clearance sales, while necessary, present a significant risk. Products like shoes can quickly become dead stock, tying up capital and warehouse space. Manually monitoring hundreds of SKUs for sales decay is inefficient and reactive. Kakaobuy identified that a sharp, consistent decline in daily sales velocity is the primary indicator of future滞销 (zhìxiāo - stagnant stock). Catching this trend early is the key to mitigating risk.
For more industry insights on e-commerce trends, you can visit Kakaobuy News.
The Power of Sales Decay Curve Machine Learning
We implemented a proprietary ML model that analyzes the real-time sales decay curve for every product in our catalog. It goes beyond simple daily sales figures and examines:
- Rate of Declining Velocity:
- Consistency of Understanding product sales trends is crucial:
- Comparatives with Similar Items:
- Inventory Level Correlation:
- Consistency of Understanding product sales trends is crucial:
Proactive Stagnation Risk Alerts
The system automatically flags products at risk. For instance, the moment a specific New Balance shoe model experiences a 35% drop in its 7-day rolling average of daily sales, an alert is triggered. This alert, often presented in a dashboard like the one below, provides a clear, early warning signal to the merchandising team.
Live Sales Alert Dashboard
| Product SKU | Product Name | Previous 7-Day Avg. Sales | Current 7-Day Avg. Sales | Sales Decay (%) | Stock On Hand | Risk Level |
|---|---|---|---|---|---|---|
| NB550CLR | New Balance 550 Clearance | 20 units/day | 13 units/day | -35% | 210 units | HIGH |
This table is not just for display; it's the trigger for automated action
Automated Tiered Promotions and Optimized Information Spread
Instead of a generic store-wide sale, our system responds with surgical precision. Based on the severity of the decay and the stock level, it generates a dynamic, multi-stage promotional strategy for that specific product.
- Stage 1 (Early Decay):
- Stage 2 (Sustained Decay):
- Stage 3 (High-Risk Stagnation):
- Stage 2 (Sustained Decay):
The system uses A/B testing data to match the product with the historically most effective coupon type (percentage-off vs. dollar-off vs. bundling) for that category, maximizing the conversion rate.
Achieving Superior Results
This data-driven merchandi clearancesing approach has transformed our clearance process. By moving from a reactive to a predictive model, Stores like Kakaobuy have significantly reduced the average clearance time for shoes and improved overall inventory turnover rates. Capital is freed up faster, and storage costs are minimized, directly contributing to a healthier bottom line.