Project definition
The customer company operates a national chain of more than 100 stores offering a menswear collection composed of about 60 different types of garments, and its own garment factory capable of producing over 160 thousand articles per year.
So, it needs a service to automate planning processes for a year ahead and optimize production and sales volumes. The proposed service should include some essential attributes such as known demand, working and reserve stock, schedule for the receipt of goods, as well as the additional specifics of marketability. Based on this information, the service should be able to generate a detailed production plan for each stock keeping unit (SKU) for a year ahead.
Task
The Globus IT team was confronted with the task of developing a service allowing the customer to:
· Enhance planning accuracy by considering and formalizing all key attributes and product range specifics.
· Minimize planning resources while improving the quality of decision-making through automation of the calculation mechanism.
· Create additional analysis tools for working with the product range. Products will be further categorized based on RFM and ABC analysis* with due account for the company’s product range specifics for the purpose of calculation and subsequent plan analysis.
“Strategic Forecasting” module, being part of the developed service, will be used to generate demand and sales forecasts.
The demand forecast allows you to determine the amount and type of goods to be produced or ordered, as well as the ratio in which they will be distributed between the stores, and to minimize the end-of-period surpluses and loss of profit caused by short deliveries.
The sales forecast is indicative of the amount of goods that can be sold with due consideration for the inventory balance and warehouse capacity.
*RFM analysis is essential for determination of customer segments based on their purchasing data.
ABC analysis is used for product ranking based on different attributes.
Yuri Akimov, Globus IT project manager:
“We had to thoroughly analyze historical data contained in the customer’s sales database to attain the expected models’ output. At this stage of analysis, it was critical to identify the key metrics making the greatest impact on the forecast accuracy.
The software solution developed by our company is comprised of Python scripts, bat and json files, and a directory structure with auxiliary data and libraries.
Upon completion of the design phase work, we were able to offer one-year ahead demand and sales forecasting models. As a result, we automated the customer’s production planning process based on the demand and sales forecasts.”
Implementation
The company’s production planning process is an annual procedure applied to lay down an optimized plan for the next year’s production order.
To automate the business-process successfully, the development team had to consider the following major stages of the planning process:
1. Update of demand forecast data
2. Reserve stock calculation
3. Update of keep-up stock data
4. Update of working stock data
5. Update of goods receipt data
6. Product categorization
7. Calculation of customer-specific adjustments
8. Production plan calculation
The solution developed by Globus IT is based on three standalone modules:
1. Product details generation module. This module is intended to convert the accounting details into details with a unique product_id, enabling a more thorough analysis of the product grid along with all the planning procedures required.
2. Product categorization, calculation and RFM-, ABC-tag assignment module. This module is intended for automatic categorization of products based on the analysis of actual sales for previous periods. It will also help the company to structure the product grid and differentiate various product categories.
3. Production plan calculation module. The accuracy of calculation of actual product volumes for the next year’s production orders.
The quality of the result directly depended on the amount of sales data. In case of an insufficient sample size, the forecasting quality could significantly deteriorate. This would require reducing the number of attributes or aggregating of samples by making forecasts for a group of stores, not for a single store.
The first phase of the project was dedicated to product details and identification of attributes to be analyzed for subsequent sales forecasting.
Then we took advantage of the RFM analysis. RFM analysis is a marketing technique used to rank customers or goods based on three parameters: time elapsed since the last purchase (Recency), purchase frequency in each period (Frequency) and customer spending amount in each period (Monetary). Using these sales data the module should automatically identify all the required parameters and split the entire product range into separate groups.
When making calculations, the development team also employed ABC analysis to rank products based on their contribution to total sales. This module can use the dynamics of each product sales for the previous year to automatically divide the products into three groups depending on the type of ABC analysis:
- ABC analysis of units sold
- ABC analysis of net sales
- ABC analysis of total margin
All these data were used to develop the production plan module, allowing to optimize expenses and minimize surpluses, while producing enough goods so that the chain of stores would not face a permanent under-supply.
Results
The solution developed by Globus IT helped to automate the production planning process: through a more accurate analysis of historical data over the past five years the customer company will be able to identify peak seasons in advance, attain higher margins and optimize sales across the entire chain of stores.
Globus IT and the customer plan to develop three other modules to ensure fully automated inventory management.
Stack
Python