Data Anomalies | Ender Sari

Addressing Inventory and Data Anomalies

Due to confidentiality agreements, I cannot disclose the actual name of the company involved; hence, I refer to it as "REXX". REXX operates multiple specialty stores across various regions, each focusing on consumer electronics. The transaction data collected by REXX includes the order date, time, store location, SKU, sales quantity, and the remaining inventory after each transaction. This dataset also accounts for transaction cancellations, facilitating accurate inventory adjustments.

The challenges I addressed during this project included:

Inconsistent Inventory Replenishments:
- Identifying precise replenishment timings for each SKU across different stores, a complex task given the irregularities in transaction records.
Record Anomalies:
- Detecting abnormalities such as negative inventory levels or sales that did not logically affect inventory figures, pointing to potential data recording or processing errors.
Data Recording Practices:
- Existing data handling methods were error-prone and required enhancements to ensure higher accuracy and reliability.

You can access the sample Python notebook code.

Button

Initial Assumptions

To ground my analysis in reality while accounting for operational nuances at REXX, I established several key assumptions:

Independent Warehouses: Each store operates its own warehouse without inventory pooling among locations.
Unique Transactions: Transactions are uniquely defined by their date and time, with no overlaps.
Transaction Validity: All recorded sales transactions are considered accurate and legitimate.
Immediate Inventory Updates: Post-cancellation, inventory updates occur instantly, making products available for subsequent sales.
Inventory Dynamics: Replenishments can occur at any time after the initial transaction, and inventory reductions directly correspond to sales unless a cancellation or replenishment is recorded.

inventory management insights

I developed and delivered algorithms using Python to detect inventory replenishments and anomalies effectively. Alongside these technical solutions, I provided several insights and strategic enhancements for inventory management, as outlined below:

Transaction-Based Inventory Reduction: I observed that inventory is typically reduced sequentially with each transaction. Due to the potential for aged inventory resulting from technological advancements or damage, I recommend recording the starting inventory for each transaction.
Inventory Tracking for Consumer Electronics: Given the long lead times characteristic of the consumer electronics sector, a comprehensive inventory record is essential. It's beneficial to track inventory as the sum of on-hand and in-replenishment stock, aiding in understanding and planning for replenishment cycles.
Handling Cancellations: When a cancellation occurs, it's often unclear when the product is returned to inventory. To address this, I suggest maintaining a record of adjusted inventory, which includes both on-hand stock and inventory returned due to cancellations.
SKU Analysis for Strategic Planning: To support further analysis, I recommend expanding inventory records to include detailed descriptions of SKUs, encompassing aspects such as product sales group, brand, lifecycle stage, replenishment lead-time, strategic importance, revenue contribution, minimum order quantity, and profit margin.
Maximizing Sales and Forecasting: High gross profit margins in consumer electronics make it crucial to capture lost sales and manage backorder quantities. Keeping detailed customer demand data is vital for enhancing sales strategies and making accurate forecasts.
Cancellation Analysis: For effective root cause analysis of cancellations, it is significant to document the specific reasons for each cancellation, whether it's initiated by the buyer or the seller.