Inventory management method for electronic commerce

Abstract

The invention provides an inventory management method for e-commerce, which integrates links of multi-source data analysis and prediction, intelligent replenishment, dynamic inventory tracking and return management, constructs a demand prediction model by using historical sales records, user behaviors and market dynamic information, sets a safe inventory level based on a prediction result, and provides an inventory management method for e-commerce. An automatic purchase request and a quick response replenishment mechanism are implemented, real-time inventory synchronization is achieved through an API, inventory changes are monitored through the Internet of Things technology, a commodity processing decision is guided for a returned commodity development prediction model, unsalable commodities are regularly examined, products close to the expiration date are reasonably handled, and a KPI is set and comprises the inventory turnover rate and the order accuracy rate. And closed-loop control of inventory management is realized.

Description

Technical Field

[0001] The present invention relates to the field of e-commerce, and in particular to an inventory management method for e-commerce. Background Art

[0002] In the e-commerce environment, the supply chain has become more complex and fragmented, involving more suppliers and distribution channels, which increases the difficulty of inventory management. Traditional inventory management usually relies on a single data source, such as historical sales data. However, such data often lags behind and cannot reflect market changes in a timely manner. Traditional forecasting methods based on empirical rules, such as moving average method and exponential smoothing method, although simple and easy to use, can only consider historical sales trends and ignore the impact of factors such as market dynamics and changes in consumer behavior. Market demand fluctuates greatly and is difficult to accurately predict, especially during promotional activities, when demand may suddenly increase, which requires e-commerce companies to have more accurate demand forecasting capabilities.

[0003] Lack of real-time data updates and information sharing may lead to poor inventory visibility, thus affecting the accuracy and timeliness of inventory decisions. Excessive inventory will lead to increased capital occupation and storage costs, while too little inventory may lead to out-of-stock risks and reduced customer satisfaction. The traditional replenishment mechanism often involves regular checks on inventory levels and then manual initiation of purchase orders. This method is not only time-consuming and labor-intensive, but may not respond in time when demand suddenly increases, resulting in inventory shortages. The return process usually involves multiple steps, including product inspection, re-warehousing, refund processing, etc. If there is no systematic approach to manage returns, it will greatly increase operating costs and time consumption. Summary of the invention

[0004] The purpose of the present invention is to provide an inventory management method for e-commerce.

[0005] The problem to be solved by the present invention is: it aims to accurately predict market demand, reasonably set safety inventory levels, realize dynamic adjustment of inventory and efficient replenishment, avoid inventory backlogs or out-of-stock phenomena, reduce unsalable products and properly handle returned goods through the comprehensive use of big data analysis and forecasting models, thereby improving inventory turnover and order accuracy, reducing inventory costs, avoiding out-of-stock or overstock situations, ensuring the smooth operation of the supply chain, optimizing the management of the return process, improving customer satisfaction and reducing operating costs.

[0006] An inventory management method for e-commerce, the technical solution adopted is as follows: S1: Collect internal historical sales records, user browsing records, shopping cart abandonment rates, customer service consultation records, product discussion popularity on social media, and competitor price changes, and establish a forecasting model to generate demand forecast results for a period of time in the future; S2: Combine the historical sales data and demand forecast results in S1 to set the safety stock level, adjust the safety stock every quarter, set trigger conditions based on demand forecast results, and automatically send purchase requests to suppliers when the actual inventory is lower than the predetermined safety stock level. Negotiate with suppliers to determine a quick response mechanism to complete replenishment in a short period of time, predict peak demand, and increase inventory in advance before major promotions; S3: Synchronizes data from all sales channels in real time based on historical sales data in S1, connects to e-commerce platforms through API interfaces, updates inventory status whenever an order is generated or canceled, uses IoT devices to track the movement of goods, and updates inventory quantities; S4: Based on the historical sales data in S1, analyze the historical return data, find out the reasons for returns, build a return prediction model, formulate inspection standards for returned goods, and determine whether they can be directly resold, sold at a discount, or scrapped; S5: Regularly analyze the historical sales data in S1, mark the products that have no sales records for several consecutive weeks, set inspection cycles for products with short shelf lives, and cooperate with charities to donate products that are close to expiration but edible; S6: Use the historical sales data and forecasting model in S1 to set KPIs, including inventory turnover rate and order accuracy, monitor KPI performance, adjust strategies through data feedback, and feed back inventory management results and problems to procurement, sales, and warehousing related departments to form a closed-loop management.

[0007] Furthermore, the prediction model established in S1 generates demand prediction results for a period of time in the future, including: S11: Extract time series features from historical sales records and extract historical sales for a period of time t , the horizontal initial component , initial component of trend , seasonal initial component , the horizontal initial component is the first value of the time series, the trend initial component is the difference between the first two values ​​of the time series, and the seasonal initial component is the monthly average value within a period of time t; Extract the number of visits per day by the user , user daily stay time , the percentage of users abandoning their shopping carts within time t , the number of consultations within time t , consulting keywords within time t , the proportion of positive reviews within time t , the proportion of negative reviews within time t , the number of discussions within time t , competitor price changes in time t ; S12: Update horizontal component , trend component , seasonal component , , , ,in is the smoothing factor of the horizontal component, trend component, and seasonal component, and its value is between 0 and 1. is the cycle length, for The seasonal component in time, for Horizontal component and trend component within time; S13: Construct feature matrix , , set the target variable , the model is trained as ,in is a gradient boosting machine, and the prediction model is ,in is a module operation, h is the predicted future time period, Seasonal component The value in the h time period, is the characteristic matrix in the time period t+h; S14: After the prediction model is established, relevant data is collected in real time according to S11, and corresponding processing is performed in S12, and the processed data is input into the prediction model to obtain corresponding demand prediction results.

[0008] Furthermore, the safety stock level is set in S2, the safety stock is adjusted every quarter, and the trigger conditions are set according to the demand forecast results, including: S21: Use historical sales data and demand forecast results to calculate the standard deviation of sales, and combine the service level and the Z value corresponding to the service level to calculate the safety stock. The service level is the probability of meeting customer demand within a certain period of time, that is, the inventory meets the demand within the probability period. The Z value is obtained from the standard normal distribution table Z corresponding to the probability. The safety stock ,in is the standard deviation of daily demand, is the lead time, which is the time required from order placement to warehouse receipt. For each quarter’s safety stock, the standard deviation of quarterly demand is calculated; S22: When the actual inventory level is lower than the result of safety stock plus average demand multiplied by lead time, a purchase request is automatically sent to the supplier; S23: Based on historical sales data and demand forecasting models, forecast upcoming peak demand, gradually increase inventory in the weeks before major promotions, and purchase from suppliers based on the average weekly ration of inventory to be replenished.

[0009] Furthermore, in S4, a return prediction model is constructed to formulate inspection standards for returned goods and determine whether they can be directly resold, sold at a discount, or scrapped, including: S41: Collect return product types , Return time , price of returned goods , Trading time information , collect customer level information , weather conditions when collecting returns And whether it is a holiday ; Convert the product category into n dummy variables using one-hot encoding , convert the customer level into m dummy variables using one-hot encoding ; Extract temperature T, humidity H, and rainfall R, and mark whether it is a holiday, 0 for no, 1 for yes; S42: Logistic regression model is used as model prediction. ,in ,in represents the vector composed of all input features, is the intercept term, and are the coefficients of product category and customer level, The coefficient of commodity price, is the coefficient of transaction time, are the coefficients of temperature, humidity and rainfall, respectively. is the coefficient of holidays; S43: Whenever there is a return request, the return prediction model is automatically called to calculate the return probability and set the return probability threshold according to business needs , when the return probability is less than , then the product is judged to be put on the shelf for sale again. When the return probability is and If the return probability is greater than , determine that the goods need to be scrapped; S44: When a return order is generated, the corresponding inventory is temporarily deducted. For goods that can be directly resold, their inventory status is restored. For discounted goods, they are transferred to the discounted sales area. For scrapped goods, they are removed from the inventory.

[0010] Furthermore, the performance of KPIs is monitored in S6, and strategies are adjusted through data feedback, and the results and problems of inventory management are fed back to the procurement, sales, and warehousing related departments to form a closed-loop management, including: S61: Calculate inventory turnover by dividing annual sales by average inventory cost, and calculate order accuracy by dividing the number of correct shipments by the total number of shipments; S62: The purchasing department adjusts the purchasing plan based on inventory turnover and demand forecasts. The sales department optimizes the sales process based on order accuracy. The warehousing department optimizes inventory management and order processing based on inventory turnover and order accuracy.

[0011] The beneficial effects of the present invention are as follows: by integrating a demand forecasting model established from multiple data sources, future market demand can be more accurately predicted, reducing inventory backlogs or stockouts caused by forecasting errors; by setting a reasonable and dynamically adjusted safety inventory level and automatically triggering purchase requests based on demand forecasts, inventory costs can be effectively controlled, while ensuring sufficient inventory supply during peak demand periods; Negotiate with suppliers to establish a quick response mechanism to complete replenishment in a short period of time, thereby improving the overall response speed and efficiency of the supply chain. Real-time synchronization of multi-channel sales data and updating of inventory status will help improve the speed and accuracy of order processing and reduce problems caused by lagging inventory information; By building a return prediction model, we can better understand the reasons for returns and make decisions based on this, such as direct resale, discount sales, or scrapping, thereby minimizing the impact of returns on the company and taking timely measures for products that have not been sold for a long time, such as donating near-expiry food; By monitoring KPI performance and feeding back the results of inventory management to relevant parties, information sharing and collaboration between procurement, sales, warehousing and other departments are promoted, forming an effective closed-loop management mechanism. By reducing out-of-stock and delayed shipments, the overall service level is improved, thereby enhancing customer shopping experience and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A flowchart of an inventory management method for e-commerce. DETAILED DESCRIPTION

[0013] The present invention is further clearly and completely described below, but the protection scope of the present invention is not limited thereto.

[0014] An inventory management method for e-commerce, the technical solution adopted is as follows: S1: Collect internal historical sales records, user browsing records, shopping cart abandonment rates, customer service consultation records, product discussion popularity on social media, and competitor price changes, and establish a forecasting model to generate demand forecast results for a period of time in the future; S2: Combine the historical sales data and demand forecast results in S1 to set the safety stock level, adjust the safety stock every quarter, set trigger conditions based on demand forecast results, and automatically send purchase requests to suppliers when the actual inventory is lower than the predetermined safety stock level. Negotiate with suppliers to determine a quick response mechanism to complete replenishment in a short period of time, predict peak demand, and increase inventory in advance before major promotions; S3: Synchronizes data from all sales channels in real time based on historical sales data in S1, connects to e-commerce platforms through API interfaces, updates inventory status whenever an order is generated or canceled, uses IoT devices to track the movement of goods, and updates inventory quantities; S4: Based on the historical sales data in S1, analyze the historical return data, find out the reasons for returns, build a return prediction model, formulate inspection standards for returned goods, and determine whether they can be directly resold, sold at a discount, or scrapped; S5: Regularly analyze the historical sales data in S1, mark the products that have no sales records for several consecutive weeks, set inspection cycles for products with short shelf lives, and cooperate with charities to donate products that are close to expiration but edible; S6: Use the historical sales data and forecasting model in S1 to set KPIs, including inventory turnover rate and order accuracy, monitor KPI performance, adjust strategies through data feedback, and feed back inventory management results and problems to procurement, sales, and warehousing related departments to form a closed-loop management.

[0015] refer to Figure 1 Shown is a flow chart of an inventory management method for e-commerce.

[0016] Furthermore, the prediction model established in S1 generates demand prediction results for a period of time in the future, including: S11: Extract time series features from historical sales records and extract historical sales for a period of time t , the horizontal initial component , initial component of trend , seasonal initial component , the horizontal initial component is the first value of the time series, the trend initial component is the difference between the first two values ​​of the time series, and the seasonal initial component is the monthly average value within a period of time t; Extract the number of visits per day by the user , user daily stay time , the percentage of users abandoning their shopping carts within time t , the number of consultations within time t , consulting keywords within time t , the proportion of positive reviews within time t , the proportion of negative reviews within time t , the number of discussions within time t , competitor price changes in time t ; S12: Update horizontal component , trend component , seasonal component , , , ,in is the smoothing factor of the horizontal component, trend component, and seasonal component, and its value is between 0 and 1. is the cycle length, for The seasonal component in time, for Horizontal component and trend component within time; S13: Construct feature matrix , , set the target variable , the model is trained as ,in is a gradient boosting machine, and the prediction model is ,in is a module operation, h is the predicted future time period, Seasonal component The value in the h time period, is the characteristic matrix in the time period t+h; S14: After the prediction model is established, relevant data is collected in real time according to S11, and corresponding processing is performed in S12, and the processed data is input into the prediction model to obtain corresponding demand prediction results.

[0017] Furthermore, the safety stock level is set in S2, the safety stock is adjusted every quarter, and the trigger conditions are set according to the demand forecast results, including: S21: Use historical sales data and demand forecast results to calculate the standard deviation of sales, and combine the service level and the Z value corresponding to the service level to calculate the safety stock. The service level is the probability of meeting customer demand within a certain period of time, that is, the inventory meets the demand within the probability period. The Z value is obtained from the standard normal distribution table Z corresponding to the probability. The safety stock ,in is the standard deviation of daily demand, is the lead time, which is the time required from order placement to warehouse receipt. For each quarter’s safety stock, the standard deviation of quarterly demand is calculated; S22: When the actual inventory level is lower than the result of safety stock plus average demand multiplied by lead time, a purchase request is automatically sent to the supplier; S23: Based on historical sales data and demand forecasting models, forecast upcoming peak demand, gradually increase inventory in the weeks before major promotions, and purchase from suppliers based on the average weekly ration of inventory to be replenished.

[0018] Furthermore, in S4, a return prediction model is constructed to formulate inspection standards for returned goods and determine whether they can be directly resold, sold at a discount, or scrapped, including: S41: Collect return product types , Return time , price of returned goods , Trading time information , collect customer level information , weather conditions when collecting returns And whether it is a holiday ; Convert the product category into n dummy variables using one-hot encoding , convert the customer level into m dummy variables using one-hot encoding ; Extract temperature T, humidity H, and rainfall R, and mark whether it is a holiday, 0 for no, 1 for yes; S42: Logistic regression model is used as model prediction. ,in ,in represents the vector composed of all input features, is the intercept term, and are the coefficients of product category and customer level, The coefficient of commodity price, is the coefficient of transaction time, are the coefficients of temperature, humidity and rainfall, respectively. is the coefficient of holidays; S43: Whenever there is a return request, the return prediction model is automatically called to calculate the return probability and set the return probability threshold according to business needs , when the return probability is less than , then the product is judged to be put on the shelf for sale again. When the return probability is and If the return probability is greater than , determine that the goods need to be scrapped; S44: When a return order is generated, the corresponding inventory is temporarily deducted. For goods that can be directly resold, their inventory status is restored. For discounted goods, they are transferred to the discounted sales area. For scrapped goods, they are removed from the inventory.

[0019] Furthermore, the performance of KPIs is monitored in S6, and strategies are adjusted through data feedback, and the results and problems of inventory management are fed back to the procurement, sales, and warehousing related departments to form a closed-loop management, including: S61: Calculate inventory turnover by dividing annual sales by average inventory cost, and calculate order accuracy by dividing the number of correct shipments by the total number of shipments; S62: The purchasing department adjusts the purchasing plan based on inventory turnover and demand forecasts. The sales department optimizes the sales process based on order accuracy. The warehousing department optimizes inventory management and order processing based on inventory turnover and order accuracy.

[0020] The present invention provides an inventory management method for e-commerce, which integrates multi-source data analysis and prediction, intelligent replenishment, dynamic inventory tracking, return management and other links, uses historical sales records, user behavior, market dynamics and other information to build a demand forecasting model, sets a safety inventory level based on the forecast results, and implements an automated purchase request and quick response replenishment mechanism, realizes real-time inventory synchronization through API and monitors inventory changes through Internet of Things technology, develops a forecasting model for returns to guide commodity processing decisions, regularly reviews slow-moving products and reasonably disposes of products approaching their expiration dates, establishes KPIs including inventory turnover rate and order accuracy, and realizes closed-loop control of inventory management.

Claims

1. An inventory management method for e-commerce, characterized in that: include: S1: Collect internal historical sales records, user browsing records, shopping cart abandonment rates, customer service consultation records, product discussion popularity on social media, and competitor price changes, and establish a forecasting model to generate demand forecast results for a period of time in the future; S2: Combine the historical sales data and demand forecast results in S1 to set the safety stock level, adjust the safety stock every quarter, set trigger conditions based on demand forecast results, and automatically send purchase requests to suppliers when the actual inventory is lower than the predetermined safety stock level. Negotiate with suppliers to determine a quick response mechanism to complete replenishment in a short period of time, predict peak demand, and increase inventory in advance before major promotions; S3: Synchronizes data from all sales channels in real time based on historical sales data in S1, connects to e-commerce platforms through API interfaces, updates inventory status whenever an order is generated or canceled, uses IoT devices to track the movement of goods, and updates inventory quantities; S4: Based on the historical sales data in S1, analyze the historical return data, find out the reasons for returns, build a return prediction model, formulate inspection standards for returned goods, and determine whether they can be directly resold, sold at a discount, or scrapped; S5: Regularly analyze the historical sales data in S1, mark the products that have no sales records for several consecutive weeks, set inspection cycles for products with short shelf lives, and cooperate with charities to donate products that are close to expiration but edible; S6: Use the historical sales data and forecasting model in S1 to set KPIs, including inventory turnover rate and order accuracy, monitor KPI performance, adjust strategies through data feedback, and feed back inventory management results and problems to procurement, sales, and warehousing related departments to form a closed-loop management.

2. An electronic commerce inventory management method as claimed in claim 1, characterized in that: The prediction model established in S1 generates demand prediction results for a period of time in the future, including: S11: Extract time series features from historical sales records and extract historical sales for a period of time t , the horizontal initial component , initial component of trend , seasonal initial component , the horizontal initial component is the first value of the time series, the trend initial component is the difference between the first two values ​​of the time series, and the seasonal initial component is the monthly average value within a period of time t; Extract the number of visits per day by the user , user daily stay time , the percentage of users abandoning their shopping carts within time t , the number of consultations within time t , consulting keywords within time t , the proportion of positive reviews within time t , the proportion of negative reviews within time t , the number of discussions within time t , competitor price changes in time t ; S12: Update horizontal component , trend component , seasonal component , , , ,in is the smoothing factor of the horizontal component, trend component, and seasonal component, and its value is between 0 and 1. is the cycle length, for The seasonal component in time, for Horizontal component and trend component within time; S13: Construct feature matrix , , set the target variable , the model training is ,in is a gradient boosting machine, and the prediction model is ,in is a module operation, h is the predicted future time period, Seasonal component The value in the h time period, is the characteristic matrix in the time period t+h; S14: After the prediction model is established, relevant data is collected in real time according to S11, and corresponding processing is performed in S12, and the processed data is input into the prediction model to obtain corresponding demand prediction results.

3. The inventory management method of electronic commerce according to claim 1, characterized in that: In S2, the safety stock level is set, the safety stock is adjusted every quarter, and the trigger conditions are set according to the demand forecast results, including: S21: Use historical sales data and demand forecast results to calculate the standard deviation of sales, and combine the service level and the Z value corresponding to the service level to calculate the safety stock. The service level is the probability of meeting customer demand within a certain period of time, that is, the inventory meets the demand within the probability period. The Z value is obtained from the standard normal distribution table Z corresponding to the probability. The safety stock ,in is the standard deviation of daily demand, is the lead time, which is the time required from order placement to warehouse receipt. For each quarter’s safety stock, the standard deviation of quarterly demand is calculated; S22: When the actual inventory level is lower than the result of safety stock plus average demand multiplied by lead time, a purchase request is automatically sent to the supplier; S23: Based on historical sales data and demand forecasting models, forecast upcoming peak demand, gradually increase inventory in the weeks before major promotions, and purchase from suppliers based on the average weekly ration of inventory to be replenished.

4. The inventory management method for e-commerce according to claim 1, characterized in that: In S4, a return prediction model is constructed, inspection standards for returned goods are formulated, and it is determined whether the returned goods can be directly resold, sold at a discount, or scrapped, including: S41: Collect return product types , Return time , price of returned goods , Trading time information , collect customer level information , weather conditions when collecting returns And whether it is a holiday ; Convert the product category into n dummy variables using one-hot encoding , convert the customer level into m dummy variables using one-hot encoding ; Extract temperature T, humidity H, and rainfall R, and mark whether it is a holiday, 0 for no, 1 for yes; S42: Logistic regression model is used as model prediction. ,in ,in represents the vector composed of all input features, is the intercept term, and are the coefficients of product category and customer level, The coefficient of commodity price, is the coefficient of transaction time, are the coefficients of temperature, humidity and rainfall, is the coefficient of holidays; S43: Whenever there is a return request, the return prediction model is automatically called to calculate the return probability and set the return probability threshold according to business needs , when the return probability is less than , then the product is judged to be put on the shelf for sale again. When the return probability is and If the return probability is greater than , determine that the goods need to be scrapped; S44: When a return order is generated, the corresponding inventory is temporarily deducted. For goods that can be directly resold, their inventory status is restored. For discounted goods, they are transferred to the discounted sales area. For scrapped goods, they are removed from the inventory.

5. The inventory management method for e-commerce according to claim 1, characterized in that: The performance of KPIs is monitored in S6, and strategies are adjusted through data feedback. The results and problems of inventory management are fed back to the purchasing, sales, and warehousing departments to form a closed-loop management, including: S61: Calculate inventory turnover by dividing annual sales by average inventory cost, and calculate order accuracy by dividing the number of correct shipments by the total number of shipments; S62: The purchasing department adjusts the purchasing plan based on inventory turnover and demand forecasts. The sales department optimizes the sales process based on order accuracy. The warehousing department optimizes inventory management and order processing based on inventory turnover and order accuracy.

Images